Security, cost benefits drive the rise of bringing AI to data

Copyright TechTarget - All rights reserved https://cyber.law.harvard.edu/rss/rss.html Techtarget Feed Generator en Sun, 11 Oct 2026 22:26:35 GMT https://www.techtarget.com/searchdatamanagement editor@techtarget.com Pipelines that port data into large language models are going the way of the dinosaur when it comes to AI development. Instead, bringing AI models to data in the secure, stable environments where data is created and stored is proving to be more effective, efficient and safe. AI needs <a href="https://www.techtarget.com/data-technologies/news/366646252/Context-is-king-as-agents-evolve-but-problems-persist">relevant data and business logic</a> -- and lots of it -- to be accurate. In the initial rush to build generative AI tools after OpenAI's November 2022 launch of ChatGPT, developers constructed complex <a href="https://www.techtarget.com/ai/definition/What-is-retrieval-augmented-generation-RAG-in-AI">retrieval-augmented generation</a> (RAG) pipelines to find relevant context, including unstructured data, and deliver it to large language models (LLMs). Those pipelines involved exporting data into <a href="https://www.techtarget.com/data-technologies/feature/Vector-search-now-a-critical-component-of-GenAI-development">specialized vector databases</a>, connecting vector databases to LLMs so the models could discover data through similarity searches, and enabling the LLMs to pull in relevant data to inform their outputs. Amid the excitement, the security risks of moving data to public models and high cost of using LLMs were secondary to the potential business benefits of GenAI, according to Sanjeev Mohan, founder and principal analyst at SanjMo. "The best models were only available as hosted APIs, so data had to travel to them," he told TechTarget. "Data platforms lacked … vector search at the time, and urgency won. Teams wanted GenAI demos fast, and copying data into a vector database was the quickest path. Governance came second." However, a confluence of factors led enterprises to reconsider how they built AI workflows. RAG pipelines were not effectively delivering the relevant context required to make AI tools accurate enough to put into production, the <a href="https://www.techtarget.com/ai/feature/2025-will-be-the-year-of-AI-agents">rise of autonomous agents</a> necessitated not only higher degrees of accuracy but also real-time data, and many data platform providers were building vector database capabilities into their broader data management offerings. Meanwhile, with enterprises getting <a target="_blank" href="https://www.deloitte.com/content/dam/assets-zone3/us/en/docs/services/consulting/2026/state-of-ai-2026.pdf" rel="noopener">little return on their AI investments</a>, cost control and security concerns became more prominent. Consequently, expensive and risky ETL pipelines are being phased out in favor of bringing AI to the data instead. "The market evolved in new directions," Kevin Petrie, an analyst at BARC U.S., told TechTarget, citing the evolution of data platforms, growing ability of LLMs to navigate unstructured data without vectorized similarity searches and rising security and regulatory concerns. "All these forces make it more appealing for organizations to bring AI to their data." <section class="section main-article-chapter" data-menu-title="Removing the pipelines, preserving context"> <h2 class="section-title">Removing the pipelines, preserving context</h2> Bringing AI to data means locating capabilities such as model inferencing, vector search and semantic logic within databases and other data management platforms. Critically, by keeping data where it gets created and stored, all of the governance, lineage and business context that gets applied to data in its native environment through tools such as <a href="https://www.techtarget.com/data-technologies/feature/15-top-data-catalog-software-tools-to-consider-using-in-2026">data catalogs</a> and <a href="https://www.techtarget.com/ai/tip/5-advantages-of-a-semantic-layer-for-enterprise-AI">semantic layers</a> remains intact when it is joined with models to build agents and other AI tools. "Bringing AI to data means running AI models and agents where a company's most important information is already stored," Craig Wiley, vice president of AI product at Databricks, told TechTarget. "By operating within a governed environment with existing security and access policies and lineage intact, you remove the need for any data migration or pipelines." By contrast, when data gets copied into models through external pipelines, versions of data need to be created, and governance and business logic get lost along the way, he continued. As a result, developers and engineers need to reconstruct governance as security in each new environment, which is both time-consuming and costly. In addition, Wiley noted that when data, <a href="https://www.techtarget.com/data-technologies/feature/One-year-of-MCP-Support-a-must-for-data-management-vendors">Model Context Protocol servers</a> and AI models are all in different environments with different governance and discovery controls, agents and users accrue <a href="https://www.techtarget.com/it-strategy/feature/Tokenmaxxing-How-CIOs-can-extract-maximum-value-from-AI-tokens">spent tokens</a> and wasted time searching for the right context. "By centralizing discovery, access control and monitoring in a single platform, they accelerate development and reduce time and token costs," he said. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Once companies started asking hard questions about how to deploy AI on their customer data, at scale, without violating their own compliance policies, they realized the 'send your data out' approach wasn't going to work. </figure> <figcaption> Pavan PothukuchiHead of AI platform, Snowflake </figcaption> </div> </blockquote> Databricks is now one of the many data management providers that enables users to bring AI to their data by turning its Unity Catalog a control plane for data, machine learning and AI and <a href="https://www.techtarget.com/data-technologies/news/366631961/Databricks-partners-with-OpenAI-to-boost-AI-development">making models</a> such as OpenAI's GPT-5 natively available in its platform alongside data. Rival Snowflake similarly allows users to bring AI to their data, with Cortex AI serving as a central location and models from Anthropic and Google <a href="https://www.techtarget.com/data-technologies/news/366637132/Snowflake-boosts-Google-partnership-integrates-Gemini-3">natively integrated</a>. Among others, hyperscalers AWS, Google Cloud and Microsoft enable customers to bring AI to their data, as do platform vendors including Cloudera, MongoDB and Teradata. Snowflake head of AI platform Pavan Pothukuchi noted that a critical advantage of bringing AI to data is that the context AI tools require remains intact, which improves the likelihood of an agent or other AI application <a target="_blank" href="https://www.evidentlyai.com/blog/ai-hallucinations-examples" rel="noopener">avoiding hallucinations</a> and performing as intended. "With this architecture, AI can work directly with a company’s unique business information without requiring them to build a separate data environment for each AI application," he said. The rush to build GenAI tools in 2023 resulted in enterprises first exporting data to vector databases and LLMs, which worked well in demos and experiments, Pothukuchi continued. But when it came time to put the applications into production, these projects no longer met most enterprise standards. "Once companies started asking hard questions about how to deploy AI on their customer data, at scale, without violating their own compliance policies, they realized the 'send your data out' approach wasn't going to work," Pothukuchi said. "Building real, production-grade AI systems on proprietary data requires governance, access control and auditability, which can't easily be replicated in new tools." </section> <section class="section main-article-chapter" data-menu-title="Savings and security"> <h2 class="section-title">Savings and security</h2> While preserving context is one benefit of bringing AI to data, <a href="https://www.techtarget.com/ai/tip/AI-inference-cost-optimization-An-enterprise-guide">lower development costs</a> and improved security are others. Moving data into AI models requires building custom storage, governance and ETL pipelines for every AI application. In addition, it racks up <a href="https://www.techtarget.com/data-technologies/definition/What-is-data-egress-How-it-works-and-how-to-manage-costs">data egress</a> fees and generally requires using platforms from multiple providers. "Multiplying that across the multiple AI tools and model providers that most enterprises use, costs compound fast," Wiley said. But by locating vector search, model inferencing and semantic logic within a unified data management environment, those costs get eliminated, according to William McKnight, president of McKnight Consulting. "Bringing AI to data is significantly less expensive," he told TechTarget. In fact, McKnight noted that benchmark tests conducted by his firm showed that bringing AI to data cut the three-year <a href="https://www.techtarget.com/it-infrastructure/definition/What-is-total-cost-of-ownership-TCO">total cost of ownership</a> of AI tools in half when compared with fragmented pipelines that required data egress and ETL workloads. In addition, testing showed that bringing AI to data accelerated development-to-production timelines by over 300%, reduced development complexity by 67%, lowered maintenance complexity by 38% and cut engineering personnel costs in half. "The cost difference comes down to infrastructure that enterprises don't have to build," Pothukuchi said. "Moving data to AI means standing up pipelines, managing ETL jobs, paying cloud egress fees, and maintaining duplicate storage and access control policies. All of that is ongoing. … When AI runs where the data already sits, those costs disappear." While costs disappear, <a href="https://www.techtarget.com/cybersecurity/tip/Guide-to-AI-data-pipeline-security-and-resilience">security improves</a>. Drastically. "Each time AI models or agents consume data, you must maintain careful security controls such as encryption, masking and identity management to ensure you don't expose personally identifiable information (PII) to unauthorized users," Petrie said. "It is easier to maintain such controls when you bring AI to the data rather than vice versa." In fact, while moving data across systems risks exposure at every step since governance has to be repeatedly re-applied, there is 100% perimeter containment when AI is brought to data in a governed data management environment, according to McKnight. "Extracting proprietary data payloads across network boundaries expands the attack surface, exposes sensitive assets to potential breaches, forces fragmented multi-hop security configurations and compromises strict data sovereignty and regulatory compliance mandates," he said. And just as benchmark testing showed the cost savings of bringing AI to data, tests conducted by McKnight Consulting quantify the security benefits of keeping data in place when developing and managing AI tools. Integrated, in-place platforms scored four times higher in <a href="https://www.techtarget.com/it-infrastructure/DellTechnologiesStorage/Is-your-data-security-strategy-ready-for-AI">data security</a> than those that require data movement across external cloud endpoints, according to McKnight. "Keeping enterprise datasets within native, governed database boundaries with automatic row-level security and encryption directly addresses the governance concerns that … data leaders cite as their primary barrier to AI adoption while eliminating the risks of data leaks, intercepted network payloads and regulatory non-compliance," he said. </section> <section class="section main-article-chapter" data-menu-title="Not the be-all-end-all"> <h2 class="section-title">Not the be-all-end-all</h2> Although bringing AI to data is a substantial improvement over exporting data to AI models, it does not reduce <a target="_blank" href="https://www.rand.org/pubs/research_reports/RRA2680-1.html" rel="noopener">all the problems</a> preventing many enterprises from moving AI pilots into production. Even when contextually relevant data and AI models are natively integrated in a secure data management environment, serving agents and other AI applications with appropriate context is challenging, according to Mohan. He pointed out that AI tools need context from many systems, and even if some of those systems are on one platform, others likely aren't. To enable AI tools to <a href="https://www.techtarget.com/data-technologies/news/366646162/Potential-consequences-are-severe-when-AI-agents-lack-context">access context</a> spread across systems and platforms without moving data, federated access through governed interfaces such as MCP and zero-copy data sharing are required. "The winning architecture is one stack with data as the foundation and AI on top," Mohan said. "The real question isn't where AI runs. It's whether it has trusted context when it does." Pothukuchi similarly noted that while bringing AI to data improves context retrieval, it is imperative that enterprises not only make their data available to AI models but ensure that their data is consistent across all systems so it can be understood by agents and models. If two departments within the same organizations define a word such as "revenue" in different ways, an agent won't know which definition is applicable and will produce <a target="_blank" href="https://www.linkedin.com/pulse/why-ai-still-confidently-wrong-rob-hornby-59sxe/" rel="noopener">confident, but wrong, outputs</a>. "Bringing a model next to their data is incredibly valuable, but proximity isn't enough," Pothukuchi said. "Companies also have to make their business knowledge legible to AI, things like how metrics are defined, what the exceptions are, and which data sources are authoritative for which questions." Meanwhile, an AI development and management system's architecture doesn't matter <a href="https://www.techtarget.com/data-technologies/feature/Successful-AI-deployment-rests-on-strong-data-foundation">if the underlying data</a> isn't correctly prepared, according to McKnight. "Bringing AI directly to data solves the architectural tax, pipeline latency, and cloud egress costs, but bringing AI to dirty or uncurated data introduces a far more dangerous problem," he said. Consequently, human oversight to prevent AI from acting on bad data and <a href="https://www.techtarget.com/data-technologies/opinion/Data-observability-for-AI-helps-curb-poor-model-performance">data observability</a> to ensure data quality are also critical aspects of any AI workflow. "The human-in-the-loop control plane and data observability strategy is important," McKnight said. "As enterprise AI evolves from looking up documents taking autonomous action, purely technical prompt engineering loses its competitive advantage to domain expertise and continuous governance." Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. </section> Natively integrating models where data lives has proven safer, more cost efficient and effective than porting data into LLMs when building agents and other AI tools. https://cdn.ttgtmedia.com/rms/onlineimages/code_g136298313.jpg https://www.techtarget.com/data-technologies/news/366651778/Security-cost-benefits-drive-the-rise-of-bringing-AI-to-data Fri, 09 Oct 2026 08:00:00 GMT Security, cost benefits drive the rise of bringing AI to data Data governance can't control what it can't see. Shadow IT bypassed the visibility, accountability and formal controls required for effective governance. Shadow AI presents a similar problem through the ad hoc use of unapproved tools, which can expose enterprise data through activity that conventional governance might not detect. Organizations typically treat it as a <a href="https://www.techtarget.com/cybersecurity/tip/Top-data-loss-prevention-tools-for-2026">data loss prevention issue</a>, but that framing neither captures nor addresses the larger problem.  Each instance of shadow AI is an act of unapproved context assembly: an employee or agent gathers documents, records or other enterprise data and feeds them to an external AI tool to shape its output. It can bypass governance checks for enterprise data, including whether "it was classified correctly, whether access policies were honored, where the data went, whether it was retained by a third party, and how it influenced the final output," said <a href="https://www.linkedin.com/in/guptanitika">Nitika Gupta</a>, product leader for security at <a href="https://www.snowflake.com/en/">Snowflake</a>. Consequently, these governance failures can expose the organization to data breaches, data loss, <a href="https://www.techtarget.com/data-technologies/feature/Why-data-governance-fails-at-compliance-and-how-to-fix-it">regulatory noncompliance</a>, inaccurate model outputs and unauthorized agent actions. Examining where existing controls fail and how enterprises have responded shows which approaches can minimize the frequency and severity of these incidents. <section class="section main-article-chapter" data-menu-title="Why existing controls miss shadow AI"> <h2 class="section-title">Why existing controls miss shadow AI</h2> Several characteristics of shadow AI make it hard to govern with traditional controls, particularly when it's used for context assembly. Conventional governance assumes that organizations can monitor users and their applications. But when an employee uses an unapproved bot or language model outside an organization's purview, "there's no logging or telemetry that goes to the [security information and event management system] that's alerting users, or agents, about it," said Sean Roche, director of value engineering and product marketing at Obsidian Security. Even controls built for <a href="https://www.techtarget.com/whatis/definition/data-in-motion">data in transit</a> tend to watch known channels such as email, downloads and endpoints. With AI agents, however, "data's in motion now, and it connects through OAuth or SaaS integrations," Roche said. "[Data loss prevention] was never made for that. It's looking for chat; it's looking for a download; it's looking for a computer. So, it misses a lot of what happens over the open internet." The <a href="https://www.cybersecuritydive.com/news/shadow-ai-employee-trust-upguard/805280/">distributed nature of shadow AI</a> also undermines existing governance methods, particularly across the many SaaS applications that run language models, agents and retrieval-augmented generation (RAG). When those applications get access to data in frequently used systems, such as Salesforce, "You need protection in the SaaS app to see that, or you never will," Roche said. "There's no network proxy that's in front of that." </section> <section class="section main-article-chapter" data-menu-title="Leakage is only the first failure"> <h2 class="section-title">Leakage is only the first failure</h2> Data leakage is often the first clear sign of shadow AI escaping existing controls and increases an organization's <a href="https://www.techtarget.com/healthtechsecurity/feature/Shadow-AI-in-healthcare-The-hidden-risk-to-data-security">risk of a data breach</a>. That risk is particularly acute with agents "because once AI agents become embedded into workflows and decision making, the attack surface expands very quickly," Gupta said. Agents can also communicate or take actions as though they represent the organization, effectively allowing them to "impersonate the business," Roche said, adding that some of the consequences could result in lawsuits, bad customer service or loss of revenue. Because shadow AI moves, transforms and uses data in external settings, it also raises data sovereignty questions, including where language models are hosted to where data resides, said <a href="https://www.linkedin.com/in/luis-flynn-a0838121">Luis Flynn, AI product marketing strategist </a>at SAS. </section> <section class="section main-article-chapter" data-menu-title="Shadow AI governance requires more than tools"> <h2 class="section-title">Shadow AI governance requires more than tools</h2> Organizations have responded to shadow AI with both traditional and newer controls. "Many enterprises are trying to detect sensitive data leaving the organization, restrict uploads to unsanctioned AI tools, monitor browser or SaaS activity and block the use of certain external services," Gupta said.   Many organizations focus on shadow AI governance at the prompt level. Tools can dynamically parse prompts and model responses in RAG frameworks to detect, obfuscate or redact sensitive information in real time. But as the industry struggles to standardize its approach to shadow AI, organizations <a href="https://www.techtarget.com/cybersecurity/opinion/Too-many-pointless-tools-Platformization-is-better">often turn to point tools</a>, such as prompt governance, rather than addressing the problem at the policy level. It's not uncommon for "people to veer off from the general approach to governance and address it at the tech level and not at the conceptual level," said Terry Dorsey, senior data architect at Denodo. Such efforts can provide some short-term utility, but they are difficult to sustain and scale. They can also <a href="https://www.techtarget.com/ai/tip/How-to-keep-data-silos-from-damaging-your-AI-projects">contribute to new data silos</a> when each one handles access control and governance along a single path rather than applying the organization's broader information policies, Dorsey said. </section> <section class="section main-article-chapter" data-menu-title="How to govern shadow AI"> <h2 class="section-title">How to govern shadow AI</h2> One way to reshape data governance for shadow AI at the policy level is to rearchitect the data estate to "bring your capabilities closer to your data, instead of the other way around," Flynn said.   Organizations might, for example, run language models, including open source ones, locally so sensitive data never leaves the organization. Other approaches include gateways that enforce policies where API calls are made and tools that monitor third-party SaaS applications for user, permission and data governance policy violations. Both also apply to AI agents. The more durable answer is to identify and approve appropriate AI resources already in use and bring them under governance. Jelani Harper is a data industry analyst and journalist covering data management, AI and enterprise IT for more than a decade. He is a research lead at Blue Badge Insights, writing analyst reports for GigaOm and articles for VentureBeat and The New Stack.    </section> Data loss prevention watches known exits, but shadow AI moves data through agents and SaaS integrations. Gateways, local models and sanctioned AI tools close that gap. https://cdn.ttgtmedia.com/rms/onlineimages/folder-files11.jpg https://www.techtarget.com/data-technologies/feature/Why-data-governance-struggles-to-audit-shadow-AI Thu, 08 Oct 2026 14:09:00 GMT Why data governance struggles to audit shadow AI Trust -- or lack thereof -- is one of the main reasons many enterprise AI initiatives never make it past the pilot stage and into production. According to a recent report from SAS and IDC, organizations that engender trust in their AI tools are 15 times more likely to get a strong return on their investments in AI development than those unable to build trust in their AI capabilities. SAS is tackling the onerous and intimidating task of building trust in AI via the launch of AI Navigator. The governance suite, made generally available on the Microsoft Marketplace on Oct. 5, is designed to help SAS customers <a href="https://www.techtarget.com/enterprise-software/feature/AI-could-earn-trust-in-transactional-work-first">build trust</a> in their agents and other AI tools. The new platform enables customers to apply all internal policies and <a href="https://www.techtarget.com/it-strategy/news/366646236/How-CIOs-can-navigate-federal-state-AI-regulation-uncertainty">external regulations</a> to AI tools, features a unified view of an enterprise's entire swath of AI capabilities -- including an inventory of all AI assets and where they are being used -- displays where there is governance risk and sends alerts when such risks are uncovered, and provides policy packs from SAS partners so users can get started with prebuilt capabilities. "AI Navigator is meaningful to SAS customers," Michael Ni, an analyst at Constellation Research, told TechTarget. "SAS is giving customers a central way to inventory models and agents, connect them to business use cases and apply policy across a fragmented AI estate. The question is whether SAS can take the next step and connect that governance to decisions and actions AI systems are actually making." However, because AI Navigator is designed to give an overall view of an enterprise's AI estate rather than provide <a href="https://www.techtarget.com/data-technologies/feature/How-agentic-AI-amplifies-data-management-challenges">fine-grained controls</a> over AI applications -- which SAS offers with other tools -- it alone does not distinguish SAS from other AI governance platforms, he added. <section class="section main-article-chapter" data-menu-title="The trust factor"> <h2 class="section-title">The trust factor</h2> When agents and other AI applications <a target="_blank" href="https://hbr.org/sponsored/2024/09/ai-has-a-trust-problem-heres-how-to-fix-it" rel="noopener">aren't trusted</a>, they go unused. Transparency is a significant aspect of trust. When users can see how AI tools go about making decisions, they're able to trust those decisions more than when there is little insight into the machinations behind the AI tool. SAS and IDC found that 97% of users will override AI-generated decisions when they can't see how the decision was made. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> SAS is giving customers a central way to inventory models and agents, connect them to business use cases and apply policy across a fragmented AI estate. The question is whether SAS can take the next step and connect that governance to decisions and actions AI systems are actually making. </figure> <figcaption> Michael NiAnalyst, Constellation Research </figcaption> </div> </blockquote> According to ISG analyst David Menninger, 74% of respondents to his firm's research survey reported that governance, security and privacy concerns are preventing them from <a href="https://www.techtarget.com/ai/feature/10-AI-business-use-cases-that-produce-measurable-ROI">getting value from their AI investments</a>. "AI governance continues to be one of the biggest challenges in getting AI deployments into production," Menninger told TechTarget. "Tools such as SAS AI Navigator can help organizations address these issues." SAS built AI Navigator to enable users to govern and view the entirety of their AI estate so they can not only ensure compliance and accuracy but also <a href="https://www.techtarget.com/searchenterpriseai/opinion/Building-governance-for-machine-speed-The-path-to-trusted-AI-autonomy">build trust in their AI capabilities</a>. "We heard … concerns from customers struggling to inventory AI assets, manage shadow AI, ensure AI use satisfied their internal policies as well as emerging regulations, and establish clear accountability across increasingly complex AI ecosystems," Reggie Townsend, SAS's vice president of AI ethics, governance and social impact, told TechTarget. Beyond customer feedback, SAS's experiences developing and managing its own AI tools provided the impetus for creating AI Navigator, he continued. "The idea for SAS AI Navigator started internally," Townsend said. "As we built and deployed more AI ourselves, we realized we needed a way to scale governance of that usage, not just the AI itself." While a means of viewing an entire <a href="https://www.techtarget.com/ai/essentialguide/Enterprise-AI-accelerates-as-calls-for-a-slowdown-grow">enterprise AI</a> ecosystem, AI Navigator complements pre-existing SAS AI governance capabilities rather than replace them, according to Townsend. "SAS AI Navigator is the lightweight AI system of record for executives who need to understand how AI is used inside their organizations, and ultimately if it's generating expected returns in a risk tolerant way," Townsend said. "Think of SAS AI Navigator as the entry point that provides visibility into how AI models and agents affect business use cases." SAS Viya, <a href="https://www.techtarget.com/data-technologies/news/366642499/Latest-SAS-capabilities-focus-on-fostering-reliable-AI">the vendor's primary platform</a> for data management and analytics, includes capabilities for observing and operating AI applications. In addition, SAS offers specialized tools such as Governance Manager, Model Risk Management and Governance and Compliance Manager for intricate governance of the AI workflow. Like Ni, Menninger added that while AI Navigator is beneficial for <a href="https://www.techtarget.com/data-technologies/feature/British-rowing-team-employs-SAS-analytics-to-prep-for-Olympics">SAS users</a>, it is not a competitive differentiator. "Our analysis of AI software providers earlier in the year indicated that, while governance capabilities were improving, many vendors were still lacking in capabilities," he said. "There have been multiple announcements and improvements during the course of the year, including this announcement, which puts SAS on par with other leaders." </section> <section class="section main-article-chapter" data-menu-title="Engineered to engender conviction"> <h2 class="section-title">Engineered to engender conviction</h2> Given that AI Navigator is designed to provide an overview of an enterprise's AI estate rather than <a href="https://www.techtarget.com/ai/tip/The-best-AI-governance-tools-and-platforms-in-2026">intricate AI governance capabilities</a>, it appears appropriately built for its intended purpose, according to Ni. "From what SAS has shown, AI Navigator is logically put together for the job it is trying to do," he said. Of particular value: the AI governance suite enables users to manage <a href="https://www.techtarget.com/ai/feature/10-AI-business-use-cases-that-produce-measurable-ROI">business use cases</a> rather than just models, Ni continued. "AI risk depends on where and how it is used," he said. "SAS then connects that use case to the models and agents behind it, the people responsible for it and the policies that apply." Menninger similarly noted that that AI Navigator is logically constructed to enable enterprises to track and oversee their AI applications. However, what both AI Navigator from SAS and the <a href="https://www.techtarget.com/data-technologies/news/366651267/Governance-teams-now-emphasize-data-AI-enablement">AI governance tools</a> from competing vendors still need to address is how to oversee AI tools built in other vendor's environments. "AI is evolving toward an amalgamation of capabilities across an entire enterprise IT estate involving many different applications," Menninger said. "The next frontier will be building and supporting a broad ecosystem of enterprise applications." Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. </section> AI Navigator is designed to foster widespread AI adoption by providing transparency into agents' actions and a comprehensive view of an enterprise's entire AI network. https://cdn.ttgtmedia.com/rms/onlineimages/ai_a205627811.jpg https://www.techtarget.com/data-technologies/news/366652036/SAS-launches-new-governance-suite-to-bolster-customer-trust-in-AI Thu, 08 Oct 2026 06:51:00 GMT SAS launches new governance suite to bolster customer trust in AI Precisely is taking a two-pronged approach to fostering AI development with the launch of prebuilt AI components and introduction of a unified data management platform. The vendor's AI Studio, made generally available on Tuesday, is a feature set of pre-built agents, applications and prompts designed to save developers some of <a href="https://www.techtarget.com/enterprise-software/news/366619611/Data-training-complexities-make-DIY-AI-agents-difficult">the complex work</a> that makes it difficult to develop AI tools. Meanwhile, the Precisely Platform, now in preview and planned for GA in early 2027, is a data management suite that unifies capabilities such as data integration and governance on a single foundation to provide <a href="https://www.techtarget.com/data-technologies/feature/Successful-AI-deployment-rests-on-strong-data-foundation">a trusted data layer</a> for AI. AI Studio is a collection of assets designed to give developers and engineers a starting point from which to build agents and other AI tools so don't they have to start AI initiatives from a blank slate. Included are tools aimed at helping users improve their data such as those that simplify configuring <a href="https://www.techtarget.com/data-technologies/definition/data-replication">data replication</a> pipelines and launching governance workflows. Others, meanwhile, build on AI-ready data such as preconfigured agents. The Precisely Platform provides a trusted data foundation on which developers can build AI tools. At its own foundation is a single metadata and <a href="https://www.techtarget.com/ai/tip/5-advantages-of-a-semantic-layer-for-enterprise-AI">semantic layer</a> on top of which capabilities such as data governance, integration and quality along with <a href="https://www.techtarget.com/data-technologies/definition/What-is-master-data-management-MDM">master data management</a> are unified so that customers can improve the readiness of their data for AI without having to migrate it into a single location. In addition, the Precisely Platform features Model Context Protocol servers to connect agents with relevant business logic, data quality and governance scores so users can determine whether data is ready for AI, and built-in data enrichment and location intelligence capabilities. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Precisely is differentiating itself through the combination of unified data management with ready-made AI assets rather than just providing one or the other. Precisely's approach of tightly integrating both layers on a single metadata and semantic foundation … is relatively distinctive. </figure> <figcaption> Stephen CatanzanoAnalyst, Omdia </figcaption> </div> </blockquote> Stephen Catanzano, an analyst at Omdia, a division of Informa TechTarget, noted that the new capabilities are significant for Precisely customers because, in tandem, the prebuilt components and data foundation address some of the problems that plague AI experiments. "These offerings enable users to accelerate development from months to days while ensuring data quality, governance, and enrichment are baked into the foundation rather than bolted on afterward," he told TechTarget. Precisely competitors such as <a href="https://www.techtarget.com/data-technologies/news/366645666/Alation-launches-OS-for-building-governing-AI-agents">Alation</a>, <a href="https://www.techtarget.com/data-technologies/news/366580432/Collibra-launches-AI-Governance-unveils-GenAI-capabilities">Collibra</a> and <a href="https://www.techtarget.com/data-technologies/news/366643437/Informatica-update-aims-to-provide-trust-foundation-for-AI">Informatica</a> also provide data quality and governance capabilities. In addition, vendors including DataRobot and H2O.ai offer prebuilt tools for acceleration AI development. Combining the two, however, could distinguish Precisely among data quality and governance vendors, according to Catanzano. "Precisely is differentiating itself through the combination of unified data management with ready-made AI assets rather than just providing one or the other," he said. "Precisely's approach of tightly integrating both layers on a single metadata and semantic foundation … is relatively distinctive." <section class="section main-article-chapter" data-menu-title="Attacking AI development barriers"> <h2 class="section-title">Attacking AI development barriers</h2> Myriad problems <a target="_blank" href="https://www.rand.org/pubs/research_reports/RRA2680-1.html" rel="noopener">prevent AI projects</a> from moving past the pilot stage, among them the complexity of development and lack of <a href="https://www.techtarget.com/ai/tip/The-role-of-trusted-data-in-building-reliable-effective-AI">a properly prepared data foundation</a>. In conversations with its customers, Precisely found that their customers mentioned how complex and time-consuming it can be to build AI tools, and that <a href="https://www.techtarget.com/data-technologies/opinion/Five-pressures-that-break-AI-programs-after-the-pilot">fragmented data</a> is making AI initiatives risky. "[The products] were driven by two consistent themes we heard from customers," Matt Waxman, the vendor's chief product officer, told TechTarget. With customers consistently bringing up development intricacies and data management difficulties as barriers to AI development, AI Studio and the Precisely Platform are central to Precisely's future product development initiatives as well, he continued. "Our focus is helping organizations make their data AI-ready and easier to use across AI, analytics and operations," Waxman said. "That starts with continuing to expand our unified data management capabilities. We also plan to grow the ecosystem of agents, apps and skills … so organizations can accelerate AI adoption." Like Catanzano, Mike Leone, an analyst at Moor Insights & Strategy noted that the new capabilities are valuable for Precisely users. "Especially for companies already running several Precisely products," he told TechTarget. "Those customers can set a rule or policy once and have it applied across everything from data integration to master data. And AI Studio gives their developers agents and apps to start from, built to work with the AI assistants they already use." The prebuilt agents and unified data management capabilities are becoming common among data platform vendors and hyperscale cloud providers, he continued. However, features such as <a href="https://www.techtarget.com/data-technologies/news/252486088/Geospatial-data-a-means-for-deeper-understanding">location intelligence</a> as part of a unified governance platform help Precisely stand apart from its peers. "Precisely stands apart because it sells its own address, property and location data so that a developer can build an agent that pulls in details like a property's flood risk, which a company usually wouldn't have on its own," Leone. </section> <section class="section main-article-chapter" data-menu-title="Correctly constructed capabilities"> <h2 class="section-title">Correctly constructed capabilities</h2> Precisely's combination of prebuilt tools and a unified data foundation appear logically built to help customers have more success as they try to move AI pilots into production, according to Leone. In particular, he noted that the new capabilities seem set up to enable users to operationalize structured data such as customer, address and product records. However, to better serve its customers, Precisely could add capabilities that allow developers to more easily <a href="https://www.techtarget.com/data-technologies/feature/Unstructured-data-needed-but-often-untapped-for-agentic-AI">include unstructured data</a> when serving AI tools. "Ideally, Precisely extends its quality and governance scores to the documents agents read, like contracts and manuals," Leone said. Catanzano likewise noted that while appropriately constructed for their intended purpose, there are <a href="https://www.techtarget.com/it-infrastructure/definition/What-are-ModelOps-model-operations-analytics-models">model operations</a> tools that Precisely could add to AI Studio and the Precisely Platform to make the new capabilities more effective. "To more fully enable users, Precisely could consider adding native model performance tracking, drift detection and automated retraining workflows that close the loop between AI deployment and the data quality issues that emerge in production," he said. "More sophisticated cost optimization tools for managing data operations at scale would also strengthen the offering." Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. </section> The vendor's AI Studio and unified data management suite are designed to simplify development on top of a trusted data foundation https://cdn.ttgtmedia.com/rms/onlineimages/iot_g871472636.jpg https://www.techtarget.com/data-technologies/news/366651681/New-Precisely-capabilities-address-common-AI-development-hurdles Tue, 06 Oct 2026 09:00:00 GMT New Precisely capabilities address common AI development hurdles A mere eight days after its acquisition by Progress Software closed, Domo on Wednesday launched a platform update highlighted by an updated version of Magic ETL, the vendor's no-code tool for data preparation. The new features in Magic ETL are designed to enable data teams to more efficiently build, monitor and fix <a href="https://www.techtarget.com/searchenterpriseai/tip/Tools-and-techniques-for-optimizing-AI-data-pipelines">data pipelines</a>. In addition, Domo's platform update includes new capabilities in AI Chat, the vendor's <a href="https://www.techtarget.com/data-technologies/feature/AI-in-business-intelligence-How-to-manage-it-effectively">natural language interface</a> for querying data, that allow users to analyze multiple datasets and documents simultaneously. "Trusted AI starts in the pipeline," Devin Pratt, an analyst at IDC, told TechTarget. "This release helps teams see what is happening and fix it before it reaches the dashboard." Capabilities such as AI-powered data analysis and AI-assisted data preparation are now common among data and analytics vendors, he continued. But where Domo is separating itself from competing vendors is by combining AI-powered analysis with <a href="https://www.techtarget.com/data-technologies/feature/Data-and-AI-governance-must-team-up-for-AI-to-succeed">concrete governance</a> and embedding AI directly into ETL workspaces. "Conversational analytics is table stakes now," Pratt said. "Domo's edge is wrapping it in governance and pushing AI into the ETL canvas itself." Staying the course There are no immediate plans to alter Domo's product development roadmap or change the cadence of its updates following Progress' <a href="https://www.techtarget.com/data-technologies/news/366646124/Domo-acquired-for-400M-sells-to-Progress-Software">$400 million acquisition</a> of Domo's technology, according to Jason Longhurst, Progress' senior director of product marketing. "Providing consistency in delivery and honoring recent commitments is important to Progress," he told TechTarget. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Trusted AI starts in the pipeline. This release helps teams see what is happening and fix it before it reaches the dashboard. </figure> <figcaption> Devin PrattAnalyst, IDC </figcaption> </div> </blockquote> Meanwhile, a combination of customer feedback and the need to keep pace with broader market trends drove Domo to develop the new features included in its latest platform update, Longhurst continued. "Customer feedback was the main motivation for the updates we are making to Magic ETL," he said. Conversely, Longhurst noted that <a href="https://www.techtarget.com/data-technologies/feature/One-year-of-MCP-Support-a-must-for-data-management-vendors">Model Context Protocol</a> (MCP) toolkits that enable authorized AI assistants to perform operational work across Domo's environment -- a feature not yet generally available -- are being developed in response to where the vendor sees the market headed. Domo first launched Magic ETL in 2017 before introducing a no-code version of the <a href="https://www.techtarget.com/data-technologies/feature/How-to-prepare-enterprise-data-for-advanced-AI">data preparation</a> tool in 2021. Now, Magic ETL includes an expanded ability to see and address data pipeline problems, heatmaps that surface jobs with significant runtime effects and versioning to provide <a href="https://www.techtarget.com/it-infrastructure/definition/read-only">read-only</a> access to pipeline configurations so developers can compare current versions of pipelines with earlier ones. In addition, the Magic ETL update features redesigned canvas to improve navigating large workflows, an <a href="https://www.techtarget.com/enterprise-software/tip/How-AI-inferred-sentiment-analysis-unlocks-customer-insights">AI-powered Sentiment Analysis</a> Tile to evaluate text from customer surveys and support tickets, and an AI-powered Classification Tile that automates categorizing unstructured text. "The latest Domo updates deliver several improvements that are individually incremental but add up to significant improvements to the overall user experience for developing and managing data pipelines," Matt Aslett, an analyst at ISG, told TechTarget, adding that the Magic ETL and AI Chat updates represent the vendor delivering some of the capabilities it unveiled during its user conference in March. Beyond new Magic ETL and AI Chat capabilities, among a spate of other new capabilities, Domo's platform update features the MCP toolkits, an MCP trigger in Data Workflows that tracks when an agent or AI assistant begins a process and support for Snowflake Semantic Views so customers can use <a href="https://www.techtarget.com/whatis/definition/business-logic">business logic</a> stored in Snowflake semantic models without having to copy data. "The announcements reinforce that … Domo should be thought of as providing much more than a business intelligence platform, with additional capabilities addressing data preparation and integration, as well as business workflow execution," Aslett said. Regarding whether the platform update includes any hints about Progress' long-term plan for Domo, Pratt noted that the substantive nature of the release hints that Progress is committed to keeping Domo competitive.  "A first release with this much substance is the clearest sign yet that Progress is committed to Domo," Pratt said.  Next for Domo to do As Domo <a href="https://www.mascience.com/interviews/why-is-m-a-integration-so-hard">becomes part</a> of Progress' broader platform, Longhurst identified three areas where Progress could focus: Enabling customers to build data and AI tools on top of trusted data, improving the speed at which customers can act on insights and ensuring that <a href="https://www.techtarget.com/ai/tip/The-best-AI-governance-tools-and-platforms-in-2026">AI is appropriately governed</a>. "Customers are seeking help with making AI more practical and accountable," he said. "You'll continue to see us focus on simplifying data preparation, enabling more intuitive analytics experiences and providing the governance capabilities required to deploy AI at enterprise scale to support their needs." Longhurst also said it's a priority to move MCP toolkits and support for Snowflake's <a href="https://www.techtarget.com/data-technologies/feature/Why-enterprise-AI-depends-on-the-semantic-layer">semantic models</a> into public release. And that is critical, according to Pratt. He noted that while the new Magic ETL and AI Chat capabilities now generally available are significant, the features that might ultimately be the most valuable are the ones that are still in the testing stages. "The best of this release is still in preview," Pratt said. "Getting it into customers' hands is the obvious next move." Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. While the vendor's capabilities are integrated into Progress Software's broader offering, new capabilities demonstrate that Domo remains competitive. https://cdn.ttgtmedia.com/rms/onlineimages/code_g1289411982.jpg https://www.techtarget.com/data-technologies/news/366651502/Magic-ETL-update-marks-first-Domo-platform-release-since-acquisition Wed, 30 Sep 2026 14:28:00 GMT Magic ETL update marks first Domo platform release since acquisition MongoDB on Tuesday unveiled new capabilities that advance the company's evolution from database specialist to data management and AI platform provider, including a memory and governance layer for AI application and management and an architecture for MongoDB Atlas. Atlas Agent Engine, currently in public preview, is a unified memory and governance layer that is designed to equip users with a single environment for building and managing AI applications, so developers don't have to piece together <a href="https://www.techtarget.com/data-technologies/feature/How-to-build-an-all-purpose-big-data-pipeline-architecture">complex pipelines</a> featuring tools from various vendors. Meanwhile, MongoDB 9.0, now generally available, is the latest version of the vendor's database engine, and Atlas Infinite, now in public preview, is a new architecture for MongoDB Atlas that <a href="https://www.techtarget.com/it-infrastructure/tip/A-primer-on-disaggregated-storage">separates compute and storage</a> to substantially improve performance. With MongoDB 9.0 the database foundation, Atlas Infinite enabling greater scale and Atlas Agent Engine enabling users to deploy agents on top, the new capabilities are logically built to advance MongoDB's evolution, according to Stephen Catanzano, an analyst at Omdia, a division of Informa TechTarget. "The new capabilities appear well-constructed to support MongoDB's positioning as an intelligent data platform for the AI era, particularly because they address the full stack from foundational performance through elastic scaling to agent execution and governance," he told TechTarget. However, MongoDB's new capabilities do not include certain key elements of AI-native development and management such as built-in model <a href="https://www.techtarget.com/ai/definition/What-is-fine-tuning-in-machine-learning-and-AI">fine-tuning</a>, automated <a href="https://www.techtarget.com/data-technologies/definition/What-is-feature-engineering">feature engineering</a> or support for emerging AI technologies beyond agents. "Competitors might offer [such capabilities] to differentiate themselves as purpose-built AI data platforms rather than traditional databases adapted for AI workloads," Catanzano said. One day before revealing the new capabilities, MongoDB CEO Chirantan "CJ" Desai stepped down to become chief enterprise platform officer at Meta. Dev Ittycheria, who served as the vendor's president and CEO from 2014 to 2025, was named interim president and CEO. MongoDB's <a target="_blank" href="https://www.google.com/search?q=mongod+stock&sca_esv=6f88a670a7218a83&sxsrf=APpeQnukyqgcFnbhrgPF09BHlzzz67agrQ%3A1790680024197&source=hp&ei=2Ju7aq3vCcj_ptQPue2F4A0&iflsig=ABILxe8AAAAAarup6OIW1UVNH7Lk9q_nCAGSgRkLJeAS&ved=0ahUKEwitu_z10pOXAxXIv4kEHbl2AdwQ4dUDCC4&uact=5&oq=mongod+stock&gs_lp=Egdnd3Mtd2l6Igxtb25nb2Qgc3RvY2syDBAAGIAEGAoYCxixAzIPEAAYgAQYChgLGLEDGIMBMgkQABiABBgKGAsyDBAAGIAEGAoYCxi0BzIJEAAYgAQYChgLMgkQABiABBgKGAsyCRAAGIAEGAoYCzIJEAAYgAQYChgLMgkQABiABBgKGAsyCRAAGIAEGAoYC0i1F1AAWN0TcAB4AJABAJgBXKABmAeqAQIxMrgBA8gBAPgBAZgCDKACuAfCAgQQIxgnwgIKEAAYgAQYigUYQ8ICEBAuGIAEGIoFGEMYxwEY0QPCAgoQLhiABBiKBRhDwgIKECMY8AUYyQIYJ8ICDRAAGIAEGIoFGEMYsQPCAggQLhiABBixA8ICBRAAGIAEwgINEC4YgAQYxwEY0QMYCsICBxAAGIAEGAqYAwCSBwIxMqAH4G6yBwIxMrgHuAfCBwYwLjExLjHIBxOACAE&sclient=gws-wiz" rel="noopener">stock fell</a> over 70 points from Friday's closing price of $410.44 per share to under $340 following the news. From database to data platform MongoDB's evolution echoes a broader industry trend. In addition to MongoDB, Couchbase and Redis have expanded beyond their database roots to become broader platforms for managing data and AI, while platform vendors including Databricks and Snowflake have made AI development and management focal point in addition to managing data. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> The new capabilities appear well-constructed to support MongoDB's positioning as an intelligent data platform for the AI era, particularly because they address the full stack from foundational performance through elastic scaling to agent execution and governance. </figure> <figcaption> Stephen CatanzanoAnalyst, Omdia </figcaption> </div> </blockquote> However, while remaining in step with the market, and despite widespread competition, MongoDB is somewhat distinguished by its longstanding operational, document-native database capabilities and <a href="https://www.techtarget.com/data-technologies/news/252515720/Gartner-Augmented-analytics-ecosystem-for-BI-now-key">emphasis on interoperability</a> with other systems, according to William McKnight, founder and president of McKnight Consulting. "MongoDB faces immense competition [but] positions itself as an open, operational runtime that consolidates memory, retrieval and governance without introducing architecture sprawl," he told TechTarget. So far in 2026, MongoDB has <a href="https://www.techtarget.com/data-technologies/news/366637414/MongoDB-launches-latest-Voyage-models-to-aid-AI-development">introduced new models</a> to improve data retrieval for agents, <a href="https://www.techtarget.com/data-technologies/news/366642768/MongoDB-adds-new-vector-performance-capabilities-to-aid-AI">improved vector search and storage</a> capabilities, and added numerous tools such as a Model Context Protocol server aimed at aiding connecting operational data with AI development tools. Now, with the launch of MongoDB 9.0 and introductions of Atlas Agent Engine and Atlas Infinite, the vendor aims to enable its users to construct a unified AI workflow. "I've had countless customer conversations, and lately they all come back to same theme," Ben Cefalo, MongoDB's chief product officer of core products, said on Sept. 26 during a virtual press conference. "Customers are building applications that are more demanding than anything we've ever seen before." The new features are designed to work in conjunction with one another to enable users to build, deploy and manage those demanding AI applications, he continued. "We're not bolting AI capabilities onto a legacy system," Cefalo said. "We built out from a database that was designed for this change." Atlas Agent Engine is designed to address some of <a target="_blank" href="https://www.rand.org/pubs/research_reports/RRA2680-1.html" rel="noopener">the problems</a> that prevent enterprises from putting agents into production, including an absence of governance, agents that lack memory capabilities and vendor lock-in to one model or AI framework. New features include data retrieval powered by MongoDB's Voyage AI embedding and reranking models, built-in memory so agents get more accurate over time, <a href="https://www.techtarget.com/it-strategy/feature/The-AI-agent-governance-gap-How-CIOs-can-gain-control">prebuilt governance</a> that oversees every action agents take and an open design that runs on any cloud or self-managed environment. "Moving AI agents from proof-of-concept into production is a major bottleneck in the enterprise. Atlas Agent Engine addresses this by consolidating execution, persistent memory and governance directly into MongoDB Atlas," McKnight said. Of particular value to McKnight: Atlas Agent Engine enables users to deploy enterprise-grade retrieval and state management directly where their operational data already by natively integrating Voyage AI models as well as allowing for flexible data and AI frameworks. He noted that the new capabilities improve MongoDB's architecture for managing the entire data and AI lifecycle by addressing key problems that many enterprises face. However, McKnight added that because MongoDB remains focused on real-time operational execution rather than broader AI lifecycle capabilities such as native model hosting and <a href="https://www.techtarget.com/ai/tip/When-to-use-prompt-engineering-vs-fine-tuning">fine-tuning</a>, it is not yet a full-featured platform for AI. "It remains to be seen if partner integrations with Voyage AI and Frontier Labs meet the needs of the enterprise," McKnight said. MongoDB 9.0 is designed to improve performance so that users can scale AI workloads without sacrificing efficiency. In tests conducted by MongoDB, version 9.0 of its database engine led to faster queries and throughput when compared to the previous iteration of MongoDB's database engine. In addition, the new version features add security measures and <a href="https://www.computerweekly.com/feature/Why-AI-is-forcing-enterprises-to-rethink-observability">observability capabilities</a>. Meanwhile, Atlas Infinite represents a new architecture for the MongoDB platform, further addressing performance by separating compute and storage. In tests conducted by MongoDB, throughput nearly doubled when compared to workloads using Atlas Core, which is the vendor's existing cloud database service. "MongoDB 9.0 and Atlas Infinite are both significant additions, with MongoDB 9.0 delivering substantial performance improvements [and] Atlas Infinite addressing a different but equally important need by providing extreme elasticity that can absorb unpredictable demand spikes in real time," Catanzano said. And though <a href="https://www.techtarget.com/data-technologies/news/366646161/Data-management-vendors-race-to-connect-AI-with-context">competing vendors</a> could distinguish themselves from MongoDB by including advanced AI-native development and management features, MongoDB is similarly finding ways to stand apart, he continued. "MongoDB appears to be differentiating itself through its open, multi-model approach that avoids vendor lock-in and its focus on unified governance and memory for production agents, whereas many competitors either lock customers into their own AI stack or require more manual integration work," Catanzano said. The road ahead for MongoDB As MongoDB makes product development plans for the remainder of 2026, it could best serve existing users and appeal to potential new ones by adding <a href="https://www.techtarget.com/ai/feature/Key-enterprise-strategies-for-AI-agent-observability">agent observability</a> and evaluation tools to Atlas so developers can trace <a href="https://www.techtarget.com/ai/tip/Managing-drift-in-AI-models-and-data">context drift</a>, according to McKnight. In addition, expanding Atlas' storage and processing capabilities as well as adding new search tools would be beneficial, he continued. "MongoDB needs to expand Atlas's capabilities to store and process a variety of tensors, matrices and scalars for high-dimensionality data, [and] enhancing MongoDB Search with greater search variety such as native spellcheck and search recommendations would strengthen its competitive showing," McKnight said. Catanzano, meanwhile, suggested that MongoDB reduce burdens places on development teams by adding AI-powered optimization capabilities that address <a href="https://www.techtarget.com/it-infrastructure/tip/The-main-types-of-database-testing-explained">database performance</a> and resource allocation based on their organization's specific application patterns. "Additionally, deeper native integrations with emerging AI development paradigms such as automated reasoning systems, multi-agent orchestration frameworks, and real-time model performance monitoring would help MongoDB move from being a platform that supports AI applications to one that actively accelerates AI development," he said. Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. A memory and governance layer and an updated database engine further the vendor's progress toward becoming a platform provider for data and AI while helping it remain competitive. https://cdn.ttgtmedia.com/rms/onlineimages/code_g1019737194.jpg https://www.techtarget.com/data-technologies/news/366651436/Evolving-MongoDB-targets-managing-agents-performance-for-AI Tue, 29 Sep 2026 08:30:00 GMT Evolving MongoDB targets managing agents, performance for AI Microsoft continues to convert Fabric, its combined data management and analytics platform, into a foundation for AI. On Tuesday at the European Microsoft Fabric + SQL Community Conference, a user conference in Barcelona, Microsoft revealed an integration between Fabric IQ -- a <a href="https://www.techtarget.com/data-technologies/news/366646252/Context-is-king-as-agents-evolve-but-problems-persist">context layer</a> within Microsoft IQ that serves AI with consistent <a href="https://www.techtarget.com/whatis/definition/business-logic">business logic</a> -- and Microsoft Copilot. In addition, Microsoft introduced IQ sharing so users can share governed data and context across their organization and with third parties, and capabilities in Fabric Apps that allow builders to connect their applications directly to Fabric warehouses, SQL databases, lakehouses and Power BI semantic models without duplicating. "Microsoft is turning Fabric from a data and analytics platform into an operating foundation for enterprise AI," Michael Ni, an analyst at Constellation Research, told TechTarget. "Power BI, Fabric IQ, real-time intelligence, agents, applications and observability are increasingly running against the same governed context." It remains to be seen whether Fabric IQ and a new <a href="https://www.techtarget.com/data-technologies/tip/Data-ontologies-are-foundational-for-usable-AI-outputs">ontology layer</a> can truly become an enterprise context layer, he continued. However, if Microsoft can integrate platforms such as Power BI and Copilot into one operating environment, it could help differentiate the tech giant from other platform vendors aiming to provide context layers for AI. "Microsoft is not alone in trying to become the context layer for enterprise AI," Ni said, naming Databricks and Snowflake among Microsoft's competitors. "Microsoft's superpower will be its distribution. It can connect Power BI, OneLake, SQL, Copilot and Fabric into one operating environment building from solutions that … already contain businesses metrics and business definitions." Becoming the Fabric for AI Over the past year, <a href="https://www.techtarget.com/data-technologies/news/366646161/Data-management-vendors-race-to-connect-AI-with-context">the dominant trend</a> among data management vendors has been facilitating the delivery of contextually relevant business logic to agents. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Microsoft is turning Fabric from a data and analytics platform into an operating foundation for enterprise AI. </figure> <figcaption> Michael NiAnalyst, Constellation Research </figcaption> </div> </blockquote> AWS, Databricks, Google Cloud, Snowflake and Teradata have all introduced tools designed to serve agents and other AI applications with context, as have more specialized data management and analytics providers such as Alteryx, Informatica, Tableau and ThoughtSpot. Microsoft is likewise making <a href="https://www.techtarget.com/data-technologies/news/366646162/Potential-consequences-are-severe-when-AI-agents-lack-context">context for AI</a> a key tenet of its data management strategy. "Everything is centered around the concept of context," Amir Netz, Microsoft's chief technology officer for Azure Data, told TechTarget. "Context allows organization to make agents the same way they expect employees to work. That's what makes agents aware of the special sauce every organization has." The nature of the enterprise is changing, he continued. Once entirely human, the workforce is going to become a hybrid mix of humans and agents. Microsoft's aim with features such as Fabric IQ and Copilot are designed to enable <a target="_blank" href="https://www.linkedin.com/pulse/four-key-requirements-building-reliable-agents-quinton-wall-93sdc/" rel="noopener">agents to become as reliable</a> as their human counterparts. "You do that by curating context," Netz said. "Fabric IQ is about providing that dimension of context that explains to the agent what is going on right now in the business. It's almost like virtual reality for agents. We have to take what's happening in the real world, record it and re-create reality for agents." Beyond the integration between Fabric IQ and Copilot, IQ Sharing and new Fabric Apps capabilities, Microsoft introduced Observability in Fabric to better enable visibility into customers' Fabric estates, updates to databases that underpin AI workflows and a data engineering agent, among a spate of other new capabilities. "The biggest thing holding back AI at work is context, meaning how well the AI understands a specific company's business," Mike Leone, an analyst at Moor Insights & Strategy, told TechTarget. "These announcements put the business knowledge companies already built in Power BI and Fabric to work across Copilot, apps and agents." From a competitive standpoint, the individual new features are largely in line with what peers are providing. However, Leone noted that the <a href="https://www.techtarget.com/ai/tip/5-advantages-of-a-semantic-layer-for-enterprise-AI">semantic models</a> that Microsoft customers have been building for years in Power BI give the hyperscaler an advantage when it comes to connecting agents with context. "Most big data platforms are building semantic layers and agents right now, so Microsoft is in step with its peers on direction," he said. "But … Microsoft knows semantics as well as anyone, since they've managed more than 20 million customer semantic models. Another way Microsoft pulls ahead is by connecting Fabric to Copilot. Very few data platforms also own the AI assistant people use at work." Meanwhile, a combination of customer feedback and Microsoft's vision led to the spate of new features being revealed in Barcelona, according to Netz. Customer interactions enabled Microsoft to understand where <a target="_blank" href="https://www.rand.org/pubs/research_reports/RRA2680-1.html" rel="noopener">enterprises were having trouble</a> as they tried to move AI agents into production, while the tech giant's vision translated the feedback into new capabilities. "When you get feedback, what you have to really listen to is what problems are out there, and [figure out] what the customer can't even imagine that we can bring in," Netz said. The future of Fabric's evolution As Microsoft plans product development for the final months of 2026, building a complete platform for <a href="https://www.microsoft.com/en-us/ai/frontier-transformation">its Frontier Transformation strategy</a> -- evolving toward a hybrid human/agent workforce -- is its focus, according to Netz. Toward that end, it will continue to improve the front-end user experience through Microsoft Copilot, the Microsoft IQ suite to provide context, and Microsoft 365 to <a href="https://www.techtarget.com/ai/feature/How-to-ensure-AI-transparency-explainability-and-trust">address trust</a>, security and governance. "It's about helping organizations, end-to-end, on their transformation journey," Netz said. "It's something [every enterprise] has to go through. … We need to help them because AI is not easy." Ni advised Microsoft to push the interoperability of Fabric IQ, making it easier for enterprises to use Fabric IQ to connect data and AI estates spread across multiple platforms. "Making semantics, ontology and governed context easy to consume from third-party tools would make Fabric more valuable to existing customers and lower the barrier for organizations that aren't all-in on Microsoft.," he said. "They seem to be heading in this direction, but to easily be part of an enterprise context layer when Microsoft doesn't own the entire stack would make Fabric much more interesting." Leone, meanwhile, suggested that Microsoft move pieces of Fabric to Azure Local so that customers can better ensure that they remain compliant with AI regulations and <a href="https://www.techtarget.com/searchenterpriseai/tip/How-to-navigate-data-sovereignty-for-AI-compliance">data sovereignty</a> rules by using more of Fabric in their own data center. "A lot of companies … [want to] know that if they have to make a move due to a regulation, Microsoft provides an easy path to do so," he said. "They want the option to bring their data and AI in-house without rebuilding anything." Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. An integration between Fabric IQ and Copilot helps turn Fabric into a context layer for agents and demonstrates the interconnectedness of Microsoft's many platforms. https://cdn.ttgtmedia.com/rms/onlineimages/cloud_g1223481405.jpg https://www.techtarget.com/data-technologies/news/366651515/Microsoft-transforming-Fabric-into-AIs-foundational-layer Tue, 29 Sep 2026 02:30:00 GMT Microsoft transforming Fabric into AI's foundational layer Writing data governance policies that promise compliance is far easier than achieving actual compliance outcomes. That gap is a primary challenge for many organizations trying to align governance and compliance. Closing it starts with understanding why governance policies often fail to deliver real compliance protection, and how to evolve both governance and compliance architectures to translate policy into practice. <section class="section main-article-chapter" data-menu-title="Why governance fails to align with compliance"> <h2 class="section-title">Why governance fails to align with compliance</h2> For most organizations, identifying compliance goals and <a href="https://www.techtarget.com/it-strategy/definition/What-is-a-data-governance-policy">writing governance policies</a> that align with them is straightforward. Problems arise when those policies go into practice. <h3>1.     Governance doesn't reflect real-world conditions</h3> A common pitfall is writing governance policies that fail to capture actual data flows, resource states and processes. For instance, a governance policy might state that the business will back up its databases hourly, but some <a href="https://www.techtarget.com/data-technologies/tip/6-best-practices-on-data-governance-for-big-data-environments">databases are so large</a> that copying them takes more than an hour, making it impossible to fully implement this policy. Avoiding this shortcoming requires assessing actual business conditions, which requires policy authors to engage with the stakeholders who work with the business's resources daily. They're the ones who know what is achievable from a governance perspective. <h3>2.     Unclear stewardship</h3> Governance policies are of little value if no one implements them, and the policies don't assign clear ownership or stewardship of tasks. As a result, no one carries out the policy because everyone assumes someone else will. The fix is to include <a href="https://www.techtarget.com/data-technologies/tip/Enhance-data-governance-with-distributed-data-stewardship">clear stewardship designations</a> in governance policies that state which teams or roles are responsible for meeting governance obligations. <h3>3.     Inadequate resources for policy enforcement</h3> Applying governance policies requires staff and funding. Organizations that fall short <a href="https://www.techtarget.com/data-technologies/feature/Which-data-management-costs-should-survive-budget-cuts">can't put policies into practice</a>. Adequate funding for the teams responsible for governance enforcement is one way to mitigate this challenge, as are automations that minimize the time and effort required to operationalize governance. <h3>4.     Sluggish governance operations</h3> Similarly, businesses may struggle to operationalize governance policies quickly enough to meet compliance requirements. For instance, they might not perform audits as frequently as a compliance framework requires, or they may fail to update governance processes quickly enough when they roll out new software. Increased funding and <a href="https://www.techtarget.com/data-technologies/tip/How-to-use-AI-to-enforce-data-governance-policies">automation investments</a> can help here as well, giving teams the tools they need to ensure that governance keeps pace with compliance mandates. <h3>5.     Insufficient incentives</h3> Sometimes, teams have the time and resources to pursue governance goals, but they fail to do so because they don't deem them important enough. The underlying cause is often cultural, arising in businesses that view governance and compliance as box-ticking exercises rather than a way to create real value. The remedy is to invest in education that helps stakeholders understand the real impact healthy governance and compliance practices have on the organization. For instance, effective data security protections not only help the business meet its compliance obligations but also mitigate the stress and disruption employees would face in the event of a data breach. <h3>6.     Changing compliance mandates</h3> Compliance frameworks change often, and <a href="https://www.techtarget.com/data-technologies/feature/Where-data-teams-fall-short-on-governance-for-the-EU-AI-Act">new rules roll out regularly</a>. If organizations fail to update their governance operations accordingly, they risk falling out of compliance. The risk isn't just that an organization won't update its policies fast enough. Policies can change while practices and tooling don't, leaving the updated policy unenforced in practice. </section> <section class="section main-article-chapter" data-menu-title="How to keep data governance and compliance aligned"> <h2 class="section-title">How to keep data governance and compliance aligned</h2> Mitigating these challenges requires investments in modern governance and compliance processes, including the following: <ul class="default-list"> <li>Automated controls. The fewer governance and compliance obligations employees must handle manually, the lower the risk of compliance failure. Automated controls, or <a href="https://www.techtarget.com/it-infrastructure/definition/What-is-compliance-automation">tools that automatically enforce governance rules</a> without human intervention, go far toward aligning governance and compliance.</li> <li>Continuous monitoring. Similarly, detecting governance failures in real time helps businesses identify issues and fix compliance problems rapidly.</li> <li>Automated report generation. The easier it is for teams to generate reports about governance and compliance status, the more likely they are to share this information with other stakeholders. Tools that automatically compile evidence and generate reports aren't a substitute for human involvement, but they streamline the tedious parts of reporting, leading to more efficient and impactful governance insights.</li> <li>Clear accountability guidelines. Establishing clear rules about who is responsible for implementing governance policies prevents situations in which no one acts because everyone thinks compliance is someone else's job. Ideally, tools automate accountability by reminding stakeholders of their obligations and following up if they don't fulfill them.</li> <li>Regulatory change management automation. Regulatory change management tools help automate the discovery of new or updated compliance rules that affect a business, reducing the risk of gaps in governance and compliance strategy.</li> </ul> Maximizing the value of governance operations requires embedding governance controls and processes into business systems rather than defining policies and leaving employees to enforce them. That automation is even more important as privacy, retention and AI regulations intensify, making traditional, manual compliance architectures impractical. Chris Tozzi is a freelance writer, research adviser, and professor of IT and society who has previously worked as a journalist and Linux systems administrator. </section> Governance policies break down in practice due to unclear ownership, thin resources and weak incentives. Automated controls and monitoring help align them with compliance. https://cdn.ttgtmedia.com/rms/onlineimages/folder-files11.jpg https://www.techtarget.com/data-technologies/feature/Why-data-governance-fails-at-compliance-and-how-to-fix-it Thu, 24 Sep 2026 12:05:00 GMT Why data governance fails at compliance and how to fix it Teradata aims to turn its Tera environment into an agentic AI-powered colleague for data management, insight generation and orchestrating agents. First <a href="https://www.techtarget.com/data-technologies/news/366642649/Teradatas-latest-targets-putting-agentic-AI-into-production">introduced in May</a> and made generally available in July, Tera enables users to interact with data and AI agents using natural language. Now, with capabilities including Tera Context Engine to enable agents to call on data across disparate data platforms, Tera Harness to automatically route workflows to the appropriate next step, and Agent Skills that feature task-specific AI tools built on Teradata's years of domain expertise, Teradata is introducing features that advance Tera beyond <a href="https://www.techtarget.com/data-technologies/tip/How-much-does-conversational-analytics-cost-in-2026">a natural-language interface</a>. All are scheduled for general availability by the end of the year. Collectively, the new tools introduced on Tuesday are designed to provide a single workspace for business users and trained experts alike to analyze data, build applications, oversee infrastructures and automate business workflows. As a result, they are valuable additions for Teradata customers, according to Stephen Catanzano, an analyst at Omdia, a division of TechTarget. "The Tera transformation addresses the important enterprise challenge of fragmented tools and skills gaps while introducing vendor-neutral capabilities that work across existing data environments, not just Teradata systems," he told TechTarget. In addition, the task-specific agents and vendor-neutral capabilities that enable users to work with data and AI beyond <a href="https://www.techtarget.com/data-technologies/news/366639802/Teradata-updates-vector-indexing-suite-to-aid-AI-development">Teradata's environment</a> to include competing platforms provide competitive differentiation, Catanzano continued. "Teradata differentiates through its vendor-neutral positioning, [which sets] it apart from cloud platform providers like Snowflake and Databricks whose AI capabilities favor their own ecosystems," he said. "The depth of enterprise data management expertise and prebuilt industry knowledge represents decades of domain expertise that newer entrants cannot easily replicate." <section class="section main-article-chapter" data-menu-title="Firming up Tera"> <h2 class="section-title">Firming up Tera</h2> Many enterprises are <a target="_blank" href="https://www.deloitte.com/content/dam/assets-zone3/us/en/docs/services/consulting/2026/state-of-ai-2026.pdf" rel="noopener">still struggling</a> to move AI initiatives into production. Reasons they're often failing in the pilot stage include <a href="https://www.techtarget.com/data-technologies/feature/Successful-AI-deployment-rests-on-strong-data-foundation">insufficient data foundations</a> and inadequate data and AI infrastructures. In addition, data management platforms that don't work well together and a dearth of <a href="https://www.techtarget.com/ai/tip/How-to-build-AI-skills-across-your-workforce">workers with the skills</a> to build and use AI tools contribute to the low success rate of AI projects. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> The Tera transformation addresses the important enterprise challenge of fragmented tools and skills gaps while introducing vendor-neutral capabilities that work across existing data environments, not just Teradata systems. </figure> <figcaption> Stephen CatanzanoAnalyst, Omdia </figcaption> </div> </blockquote> The new Tera capabilities are designed to help Teradata customers overcome some of these barriers. In particular, they address platforms that aren't interoperable with its vendor-neutral Context Engine and the skills gap with Agent Skills. "I'd describe the earlier Tera as a capable assistant. It could help you do things, but you were still doing most of the driving," Sumeet Arora, Teradata's chief product officer, told TechTarget. "[Now, it] can hold an objective, figure out what needs to happen next, coordinate across different tools and data sources, and carry the work through to a governed outcome without someone babysitting every step." To <a href="https://www.techtarget.com/data-technologies/news/366646252/Context-is-king-as-agents-evolve-but-problems-persist">serve agents with appropriate context</a> from disparate data sources, the Tera Context Engine includes a context graph that connects business logic, neurosymbolic models that combine <a href="https://www.techtarget.com/ai/tip/5-advantages-of-a-semantic-layer-for-enterprise-AI">semantic mapping</a> with Industry Knowledge Models and governance capabilities such as validation controls. Tera Harness, which is built on a Go-native engine and the <a href="https://www.techtarget.com/it-infrastructure/definition/gRPC">gRPC</a> call framework, routes workflows by loading relevant memory and pre-planning work, monitoring progress and recovering from failures rather than abandoning tasks, and applying limits to unproductive iterations so AI tools don't get stuck in a failure loop that drives up costs. Finally, Agent Skills includes a set of Platform Agents for managing the Teradata environment and a different set of Analytics Agents for working with data across platforms, both of which Tera automatically routes to the correct models and tools. Like Catanzano, Devin Pratt, an analyst at IDC, called out the significance of enabling organizations to <a href="https://www.techtarget.com/data-technologies/news/366646162/Potential-consequences-are-severe-when-AI-agents-lack-context">connect agents with context</a> stored across data platforms. "Enterprises already know they can prototype an agent," he told TechTarget, noting that his firm's research shows that fragmented, hard-to-access data is frequently cited as a reason organizations struggle to put agents into production. "What they are paying for is one that survives contact with production, and that means building it where the governed data already lives." Numerous vendors provide the same individual layers Teradata is planning to deliver once the new Tera capabilities are GA, Pratt continued, specifically naming AWS, Databricks, Google Cloud, Microsoft and Snowflake. However, Teradata's <a href="https://www.techtarget.com/it-strategy/opinion/Rethink-the-AI-stack-Why-lean-open-architecture-is-the-answer">open framework for AI</a> is a distinguishing characteristic. "The building blocks are everywhere now," he said. "What is rare is running governed agents directly on top of the mission-critical data that never leaves the building." </section> <section class="section main-article-chapter" data-menu-title="Missing links"> <h2 class="section-title">Missing links</h2> Tera Context, Tera Harness and Agent Skills are logically designed to carry out Teradata's aim of turning Tera from a natural language assistant into a system of tools that can autonomously work alongside business users. However, despite Teradata Harness including <a href="https://www.computerweekly.com/feature/Why-AI-is-forcing-enterprises-to-rethink-observability">observability</a> into how agents complete their work and capabilities that enable agents to recover from failures, Catanzano noted that what's missing is a feature that enables humans to correct and teach Tera. "The three capabilities are well-architected for the agentic coworker vision, but there appears to be a gap in bidirectional learning mechanisms where users can naturally correct and teach Tera when outputs miss the mark or business context shifts," Catanzano said. "A more robust human-in-the-loop framework with contextual learning would strengthen the coworker positioning beyond task execution." Pratt likewise noted that Teradata's three new Tera capabilities logically cover the AI lifecycle and further the vendor's aim of turning Tera into a coworker. Nevertheless, he says, Teradata could more fully deliver on its goal by adding testing, rollback when bugs or errors occur and versioning capabilities. In addition, <a href="https://www.techtarget.com/enterprise-software/feature/The-hidden-cost-of-AI-automation-Preserving-organizational-expertise">more human involvement</a> would be beneficial. "The lifecycle is in the right shape," he said. "The next layer to build out is the control that lets an enterprise trust an agent with real responsibility." </section> <section class="section main-article-chapter" data-menu-title="Up next: more Agent Skills, more customer progress"> <h2 class="section-title">Up next: more Agent Skills, more customer progress</h2> As Teradata plots its next product development initiatives, helping customers move more of their AI pilots into production by providing AI services is a priority, according to Arora. "A huge amount of enterprise AI investment is still sitting in experimentation," he said. "One of our biggest priorities is helping our customers graduate from 'we tried this and it was interesting,' to 'this is running in production and creating measurable value.' That means … helping people identify where governed context will make a difference to a business outcome and getting there faster." In addition, Teradata is focused on adding more Agent Skills to aid organizations that <a href="https://www.techtarget.com/enterprise-software/feature/The-AI-skills-gap-is-getting-worse-Why-and-what-to-do">don't have enough employees</a> with the experience to build and use AI tools, Arora added. "The agent skills we're shipping at GA are a starting point, not a ceiling," he said. Pratt noted that the biggest step Teradata needs to take next is to prove that the new capabilities work as intended. That includes showing how named customers are using Tera to run critical AI workloads. Catanzano, meanwhile, suggested that Teradata expand Tera by adding industry-specific agent skills that accelerate AI adoption and making Tera a means of orchestrating workflows that spread across different <a href="https://www.techtarget.com/ai/feature/AI-engineer-vs-data-scientist-Whats-the-difference">data and AI personas</a> within organizations. "Teradata should expand Tera into cross-functional workflow orchestration that facilitates handoffs between business analysts, data engineers and data scientists who speak different languages and work in different tools," he said. Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. </section> Teradata is introducing features that advance Tera beyond a natural-language interface, including vendor-neutral capabilities that work across existing data environments. https://cdn.ttgtmedia.com/rms/onlineimages/storage_g1197646065.jpg https://www.techtarget.com/data-technologies/news/366650846/With-new-context-engine-Teradata-aims-to-be-an-AI-colleague Tue, 22 Sep 2026 12:18:00 GMT With new context engine, Teradata aims to be an AI colleague Meet the streamhouse. Created in response to the demand for AI agents to be served with the context they require to perform properly in production, this portmanteau of <a href="https://www.techtarget.com/data-technologies/opinion/Real-time-data-streaming-for-AI-invest-where-it-matters">streaming data</a> and <a href="https://www.techtarget.com/data-technologies/news/366545117/Lakehouse-architecture-the-best-fit-for-modern-data-needs">data lakehouse</a> builds on the data lakehouse foundation to address AI agents' need for continuous, real-time data. Aiven, Confluent, Redpanda, StreamNative and Ververica recently established the Streamhouse Working Group and introduced an open definition of the new streamhouse architecture and a list of its key attributes. Lakehouse architectures enable users to store and analyze large volumes of historical data. Streamhouse architectures similarly enable access to relevant data. But where streamhouses differ from lakehouses is in the data timeframe: rather than house historical data, streamhouses simplify and standardize serving real-time data spread across myriad systems to agents so they're informed by <a href="https://www.techtarget.com/data-technologies/news/366646252/Context-is-king-as-agents-evolve-but-problems-persist">the fresh context</a> that keeps agentic workflows current and reliable. "Publishing a standardized architecture matters, even if only because it lays down a shared language and a common starting point for the industry," Donald Farmer, founder and principal of TreeHive Strategy, told TechTarget. "The streamhouse definition provides a vendor-neutral blueprint." Kevin Petrie, an analyst at BARC U.S., similarly noted that defining a new architecture -- including naming supporting technologies such as change data capture, open table formats and data catalogs -- is valuable for the data and AI industries. However, he noted that <a target="_blank" href="https://streamhouse.com/definition/" rel="noopener">the working definition</a> of a streamhouse is nascent and specifics such as how the architecture supports other open standards are scarce. "I fully support the concept and believe this working group calls out the right supporting technologies," he told TechTarget. "With that said, the Streamhouse website documentation is pretty sparse so far and would benefit from a more detailed reference architecture based on customer implementations." <section class="section main-article-chapter" data-menu-title="A need for something new"> <h2 class="section-title">A need for something new</h2> Agents are designed to make decisions and execute business processes such as supply chain management and fraud detection as operational events take place. To perform as intended in production, agents require not only high-quality data that can be trusted, but also fresh, governed business context delivered continuously and on a production scale. Without streaming data, they lack <a href="https://www.techtarget.com/data-technologies/news/366646161/Data-management-vendors-race-to-connect-AI-with-context">real-time context</a> and can miss the triggers they need to carry out business processes as intended. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Publishing a standardized architecture matters, even if only because it lays down a shared language and a common starting point for the industry. The streamhouse definition provides a vendor-neutral blueprint. </figure> <figcaption> Donald FarmerFounder and principal, TreeHive Strategy </figcaption> </div> </blockquote> However, enterprise data is often distributed across a morass of SaaS applications and stored in databases, data warehouses, data lakes and data lakehouses on multiple clouds. As a result, building pipelines that connect data spread across so many systems and delivering it to agents in real-time -- with the appropriate governance -- is too complex for some organizations, according to Farmer. "The ad-hoc plumbing most enterprises had in place just wasn't good enough," he said. A new way of serving agents the <a href="https://www.techtarget.com/data-technologies/feature/6-top-business-benefits-of-real-time-data-analytics">real-time data</a> they require is therefore needed, Farmer continued. "The streamhouse model specifically engineers streams for production environments, adding the necessary transformation and governance that ad-hoc messaging pipelines have not delivered," he said. "It also enables organizations to avoid the friction of consolidating all enterprise data into a single system before use." Petrie pointed out that while streaming data is important for agents, as enterprises race to develop and deploy autonomous AI networks, not enough emphasis has been placed on how agents access real-time data. "Amidst the rush to embrace agents, lakehouses, and documents, I think a lot of organizations overlooked the need to provide their agents with a standard, open method of accessing data on a real-time basis," he said. Streaming architectures such as <a target="_blank" href="https://kafka.apache.org/" rel="noopener">Apache Kafka</a>, which is the technology <a href="https://www.techtarget.com/data-technologies/news/366643312/Latest-from-Confluent-streamlines-use-of-streaming-for-AI">Confluent</a> is built on, offer one way of doing so, Petrie continued. The streamhouse is now another. "Nearly all agents require real-time access to data, and streaming is an ideal way to provide real-time access and processing of tabular or semi-structured logs," Petrie said. Aiven, Confluent, Redpanda, StreamNative and Ververica define a streamhouse as an architecture that enables enterprises to continuously capture, govern and serve <a target="_blank" href="https://mitsloan.mit.edu/ideas-made-to-matter/4-capabilities-a-real-time-business" rel="noopener">the current state of their operations</a> to agents and other applications so they can analyze and act upon the information in real time. Key attributes of the streamhouse architecture include the following: <ul type="disc" class="default-list"> <li>Continuous capture, processing and delivery of data so it is made available in real time as events occur rather than delivered via <a href="https://www.techtarget.com/it-infrastructure/definition/What-is-a-batch-file">batch file</a> processing.</li> <li>Production-grade engineering so that business-critical applications, including agents, and key analyses can depend on continuous delivery.</li> <li>Decentralization so agents can <a href="https://www.techtarget.com/searchdatacenter/opinion/Distributed-computing-The-infrastructure-shift-AI-demands">access data where it lives</a> rather than wait for it to be packaged and consolidated in a single location.</li> </ul> Alex Gallego, co-founder and CEO of <a href="https://www.techtarget.com/data-technologies/news/366640882/Redpanda-launches-streaming-engine-optimized-for-AI">Redpanda</a>, noted that the vendors came together to define the streamhouse after observing customers originate from different starting points to build agent workflows that ultimately looked similar, each featuring real-time data and production-grade engineering that connect decentralized data estates. "The data can't be hours old, and the system can't depend on a batch job catching up before an agent makes a decision," Gallego told TechTarget. "We were all hearing versions of that from customers. When we compared notes, it was clear this … was the market converging on the same problem. That convergence is what made a shared, vendor-neutral definition worth publishing." While the key attributes named by the Streamhouse Working Group are valuable, Farmer noted that more capabilities are also critical to developing an effective pipeline to serve agents with streaming data. In particular, governance, security and interoperability that enable streamhouses to exchange data with lakehouses using open table formats such as <a href="https://iceberg.apache.org/">Apache Iceberg</a> and <a href="https://www.techtarget.com/data-technologies/news/366542953/Databricks-introduces-Delta-Lake-30-to-help-unify-data">Delta Lake</a> are also vital. "If an agent is taking autonomous action, the data feeding it must be strictly audited, lineage-tracked, and access-controlled," Farmer said.  Petrie noted that the Streamhouse Working Group's correctly emphasizes certain governance capabilities. However, like Farmer, he advised the consortium to add more specifics regarding how streamhouses can keep streaming data secure and <a href="https://www.techtarget.com/ai/opinion/How-to-manage-the-gap-between-enterprise-AI-use-and-AI-regulation">compliant with evolving regulations</a>. "The working group rightly names governance as a must-have characteristic, but I think users need more detail about how best to meet specific requirements for identity and access management, personally identifiable information obfuscation, and protection of intellectual property," he said. "They also need more information about how catalogs and meta stores can organize the supporting metadata." </section> <section class="section main-article-chapter" data-menu-title="The streamhouse needs additions"> <h2 class="section-title">The streamhouse needs additions</h2> After publishing the initial definition of a streamhouse, the Streamhouse Working Group plans to maintain the definition and alter it as technology evolves, according to Gallego. In addition, the group is working to improve the standards and practices that allow organizations to build streamhouses using both <a href="https://www.techtarget.com/ai/feature/Open-source-AI-What-it-means-for-enterprise-innovation">open-source</a> and proprietary tooling and opening participation in the Streamhouse Working Group to additional members. "Streamhouse is open by design," Gallego said. "No one company owns the definition, and we welcome anyone building in this space to contribute and be part of it. That's better for customers -- they get choice, competition, and innovation instead of another category controlled by one vendor." Gallego acknowledged that the streamhouse architecture addresses only part of the problem that prevents organizations from putting agents in production. While the architecture provides the data infrastructure, a <a href="https://www.techtarget.com/data-technologies/feature/Why-AI-forces-security-first-governance">governance and security</a> layer is still needed to control what agents access. An additional missing piece that Farmer advised the working group to broach is how streamhouses interface with the memory systems of agents, including <a href="https://www.techtarget.com/data-technologies/feature/Vector-search-now-a-critical-component-of-GenAI-development">vector databases</a> and <a href="https://www.techtarget.com/ai/definition/What-is-retrieval-augmented-generation-RAG-in-AI">retrieval-augmented generation</a> systems, so data is prepared in real time. "Transporting the data is only half the battle," he said. "Formatting and indexing the streaming data so an LLM can query it -- in the milliseconds needed for real-time -- is a critical hurdle they do not address." Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. </section> A group of vendors is joining forces to define an architecture for fueling AI with real-time context to aid enterprises struggling to feed agents the streaming data they need. https://cdn.ttgtmedia.com/rms/onlineimages/ai_a264431831.jpg https://www.techtarget.com/data-technologies/news/366650920/Streamhouse-A-new-way-to-serve-AI-with-real-time-context Tue, 22 Sep 2026 08:00:00 GMT Streamhouse: A new way to serve AI with real-time context There is no AI without a strong data foundation. If agents and other AI tools can't access governed, <a href="https://www.techtarget.com/data-technologies/feature/GenAI-demands-greater-emphasis-on-data-quality">high-quality data</a>, they'll deliver inaccurate, untrustworthy results and AI projects will <a target="_blank" href="https://www.deloitte.com/content/dam/assets-zone3/us/en/docs/services/consulting/2026/state-of-ai-2026.pdf" rel="noopener">never make it past the pilot stage</a> and into production. <div class="imagecaption alignLeft"> <img src="https://cdn.ttgtmedia.com/rms/onlineimages/mohan.sanjeev.jpg" alt="Sanjeev Mohan, founder and principal analyst at SanjMo">Sanjeev Mohan </div> In his new book titled "Designing the AI-Driven Data Foundations: Architecture, Principles, and Practice," veteran analyst Sanjeev Mohan, who headed Gartner's data and analytics research team before founding SanjMo in 2021, lays out the critical nature of a proper data foundation for AI and how organizations can build that bedrock. AI, he states, is fundamentally just the most modern use case for data. For decades, data in the enterprise was kept under wraps by specialists and largely used to inform business intelligence tools such as reports and dashboards. In recent years, the paradigm has evolved. Now, data is earmarked to <a href="https://www.techtarget.com/data-technologies/feature/8-proactive-steps-to-build-trusted-data-for-analytics-and-AI">underpin agents and other AI tools</a>, and it is becoming the infrastructure for widespread AI-driven analysis and process automation. In the coming years, data will be the foundation for technologies such as robotics and quantum computing. But just as BI without reliable data couldn't be used to inform decisions, AI without a strong data foundation can't be trusted to generate insights and run business processes. In a recent interview with TechTarget, Mohan discussed the essential nature of a strong data foundation for AI, what makes a good data foundation and how AI can play a role in preparing data for AI as well as benefitting from it. In addition, he spoke about <a href="https://www.techtarget.com/data-technologies/feature/Data-and-AI-governance-must-team-up-for-AI-to-succeedhttps:/www.techtarget.com/data-technologies/feature/Data-and-AI-governance-must-team-up-for-AI-to-succeed">data governance's role</a> in the evolving AI workflow, the barriers that still prevent some enterprises from building a strong data foundation and how data for AI will continue to evolve. Editor's note: This Q&A has been edited for clarity and conciseness. Before we get into the details of what makes a good data foundation for AI, tell us why a strong data foundation is so important for AI. Sanjeev Mohan: When I watch vendors demo new capabilities these days, they all open a natural-language interface, they ask a question, and then an answer magically appears. But that's a fallacy. It makes it look like AI has overpowering capabilities that wash over the cesspool of data that [organizations] have collected. Your AI is only as good as your data, which is <a href="https://www.techtarget.com/it-strategy/feature/What-enterprises-are-getting-wrong-about-AI-data-readiness">easier said than done</a>. Data doesn't have to be pristine [for AI], but you at least have to know what you have. Why don't many organizations know what they have and instead have a cesspool of data? Mohan: Data estates are <a target="_blank" href="https://explodingtopics.com/blog/data-generated-per-day" rel="noopener">exploding at their seams</a>. Most organizations today have over 100 SaaS products. All of them collect data, but they don't share it because there is no mechanism [to connect them], and there are no standards to make it consistent across applications. Instead, there are all these SaaS products with local copies of data -- maybe some gets moved into data warehouses -- leading to too many data siloes. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> I'm a very strong believer that if we can fix the data foundation, we can derive the success from AI that has so far eluded us. </figure> <figcaption> Sanjeev MohanFounder and principal analyst, SanjMo </figcaption> </div> </blockquote> In organizations, sales and marketing should be connected at the hip. Instead, sales has its ecosystem of data, and marketing has its ecosystem of data. As a human consumer in a senior position, if I need to know how marketing is leading to sales, I can connect them. But if AI agents are analyzing data, how in the world will an agent know that 'X' in one system is the same as 'Y' in another system? There are no standards, and the two are coded differently. I'm a very strong believer that if we can fix the data foundation, we can derive the success from AI that has so far eluded us. Fixing the data foundation to get value from AI seems obvious, so what are you experiencing when you talk to organizations about doing so? Mohan: If I talk too much about data, people start shaking their heads and say, 'What are you talking to us about data? Talk to us about AI.' When I tell them that AI is a use case of data, some take offense. To them, AI is a whole, gorgeous, brand-new world. But it's a use of data. Tomorrow it will be robotics, then it will be <a href="https://www.techtarget.com/whatis/definition/quantum-computing">quantum computing</a>, and then it will be something we haven't even thought about. All of these things are sitting on top of a substrate, and that substrate is a data foundation. Before the AI era began a few years ago, what did organizations need to do to build a proper data foundation? Mohan: That process was determined by technology. When computers came in during the 1960s, 70s and 80s, memory was very expensive, so when networks arose, they were very slow. Data foundations evolved from that era. It was almost always structured data,it was all <a href="https://www.techtarget.com/it-infrastructure/definition/What-is-a-batch-file">batch file</a> processing, there was no concept of real time and there were a lot of manual checks with humans in the loop. Automation was something only enabled by the cloud. How has AI changed what the data foundation looks like -- how is it different than it was just a few years ago? Mohan: With AI, there is an amazing ability to detect similarities across any data source. Now, I can say, 'Give me information structured sources such as databases and CRM applications and do sentiment analysis on customer support calls for a given customer.' In one query, AI can join audio transcripts, structured data and <a href="https://www.techtarget.com/enterprise-software/tip/How-AI-inferred-sentiment-analysis-unlocks-customer-insights">sentiment analysis</a>. There is data [that informs] AI, and AI [that helps manage] data. AI plays two roles. It fixes the data foundation, and it helps derive new insights from that data. If you were advising an enterprise, how would you tell them to build a data foundation for AI -- what tools do they need from start to finish to have a proper data and AI stack? Mohan: I tell them to first pick one business process that they're trying to turn into an agent, then lay out the data floor from ingestion, through integration and transformation, to observability and reporting. Once they understand their data process, they can look to see where they can insert AI into the process, such as at the ingestion level. And once they understand the process, they can turn their natural language documentation into a skill. The first skill becomes ingestion so they can instruct their AI to ingest, classify, profile and find anomalies. Others are data integration, data quality, and so on, but start with ingestion and automate it, and have a <a href="https://www.techtarget.com/enterprise-software/feature/The-hidden-cost-of-AI-automation-Preserving-organizational-expertise">human in the loop</a> to check if the data was ingested correctly and whether there were duplicates. In AI, the most important thing is evaluations, and what you're doing in this process is creating relevant context for large language models (LLMs) to act on your behalf. What does an LLM acting on a user's behalf look like? Mohan: The final thing is to be able to tell your application something such as, 'Build me a dashboard for SEC filings,' or, 'Do my taxes.' That combines all the other skills. To do my taxes, first the LLM needs to get my 1099s and W2s -- that's ingestion and integration. Then a human needs to cross-check to make sure it was correct. Ingestion is not an outcome. Integration is not an outcome. These are means to an end. AI should be considered a new paradigm, a new way of doing things. In the past, humans built their own reports but were limited. Now, AI can access more information. Where do capabilities such as data governance enforcement fit in now? Mohan: People ask how to go from data governance to <a href="https://www.techtarget.com/ai/tip/The-best-AI-governance-tools-and-platforms-in-2026">AI governance</a>, and I tell them that there is a problem in how they're asking the question. You're not going from one to the other. You're upscaling from data to AI governance. You still need all the lineage, business glossaries and catalogs at the data level. But now, with AI, everything changes. The idea of data governance was to govern source data and build deterministic applications on top that would always run the same way. AI breaks that thinking because <a target="_blank" href="https://www.youtube.com/watch?v=U8kuVAvam50" rel="noopener">AI is probabilistic</a>. Even with highly governed data, AI can misbehave and provide wrong answers. Now, you have to govern the output. That's why humans need to be in the loop. You have to govern your input, which is data governance. You have to govern your output, which is AI governance. And with agents, you have to govern all the things that are happening in a multi-agent systems. Data governance is still needed, but you have to upskill agents so that before something bad happens, it can be stopped. As you interact with organizations, are you seeing that most now have their data in order for AI, or do many still not have proper data foundations? Mohan: It's getting better. What I mean by that is that when I started my career, all the data was proprietary. SQL was the open thing, but everyone had their own flavor, so it wasn't really open. Today, there are open standards throughout the data stack, such as object store with Apache Parquet files and Apache Iceberg as a table format. Even vendors such as Snowflake with proprietary formats <a href="https://www.techtarget.com/data-technologies/news/366641473/Snowflake-broadens-open-source-embrace-ups-Iceberg-support">are embracing Iceberg</a>. That allows users in Snowflake to directly see what's going on in SAP, which is being published to the outside world in Iceberg. Now, you also have standards for semantic layers with <a href="https://www.techtarget.com/data-technologies/news/366631576/New-consortium-to-aid-AI-by-standardizing-semantic-modeling">the Open Semantic Interchange</a>, for agents with <a href="https://www.techtarget.com/data-technologies/feature/One-year-of-MCP-Support-a-must-for-data-management-vendors">Model Context Protocol</a>. We've broken free of proprietary technologies, and for the first time, we are seeing standards span the entire stack. … The data foundation is a living, breathing organism where cells die and new ones generate. What are some of the biggest hurdles preventing organizations from building a data foundation for AI? Mohan: The hurdles that organizations face are in keeping up with technology in a manner in which their investment is safeguarded, and that is very difficult. For example, 15 years ago, we thought <a href="https://www.techtarget.com/data-technologies/definition/Hadoop">Hadoop</a> was a savior. So, now we ask how to know whether agents will be successful. Right now, they're running haywire. Knowing what to do is a big challenge. A second challenge is how to do it. Everyone is <a href="https://www.techtarget.com/data-technologies/news/366646252/Context-is-king-as-agents-evolve-but-problems-persist">talking about context</a>, but how many are talking about how to build that context? Two people building context might have very different opinions about what context is. When you haven't even agreed upon a definition of context, how can you build it. The idea is great, but we don't know how to do it yet. Looking into the future, how will the data foundation evolve, or do you think that AI's current data requirements will remain somewhat static over the next few years? Mohan: I think the data foundation will evolve tremendously. In my book, I introduce a concept which should be the next version of the extract, transform and load process (ETL). I call it ECL -- extract, context, link. ETL was great for batch jobs, for moving data. But why are we moving data? If you tell AI where the data is, where the context is for a query, let the AI go where the data is. You don't want to move data unnecessarily because <a href="https://www.computerweekly.com/news/366649643/Rising-token-bills-force-rethink-of-AI-cost-management">it's expensive</a>, it opens up security holes and there's latency. What changes is that you need access to contextual metadata, and your sources are [unstructured data]. … And you need the context to be linked in a <a href="https://www.techtarget.com/data-technologies/feature/Use-knowledge-graphs-with-databases-to-uncover-new-insights">knowledge graph</a> or some other way. This is how AI changes the ETL paradigm into ECL where context is created in real time. But there is work to be done. All this tooling still has to be developed. A lot of enterprises are still trying to solve that context layer. Beyond figuring out context, is there anything you want enterprises to be able to do to build a data foundation that still isn't possible? Mohan: I was very hopeful that by now, four years into this, there would be an ability for organizations to train their own models, but we haven't gotten there. And so far, it doesn't look close. There are frontier models that are trained on common data, but enterprise data is locked away. That's why we started doing <a href="https://www.techtarget.com/ai/definition/What-is-retrieval-augmented-generation-RAG-in-AI">retrieval-augmented generation</a> and other experiments. Eventually, every company might have its own model that is constantly refreshed, and that might change things again. Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. Just as BI was built on a bedrock of high-quality, relevant data, AI tools are only as trustworthy as the data that informs them. https://cdn.ttgtmedia.com/rms/onlineimages/ai_a252657224.jpg https://www.techtarget.com/data-technologies/feature/Successful-AI-deployment-rests-on-strong-data-foundation Mon, 21 Sep 2026 10:28:00 GMT Successful AI deployment rests on strong data foundation Governance programs are standard now, but many of them cost more than the problems they prevent are worth. McKinsey research shows <a href="https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/designing-data-governance-that-delivers-value">30% of total enterprise time</a> is being lost to poor data quality. Gartner expects <a href="https://www.gartner.com/en/newsroom/press-releases/2024-02-28-gartner-predicts-80-percent-of-data-and-analytics-governance-initiatives-will-fail-by-2027-due-to-a-lack-of-a-real-or-manufactured-crisis-">80% of governance programs to fail</a> by 2027, and the cost of getting it wrong is rising fast. The<a href="https://www.edpb.europa.eu/news/news/2023/12-billion-euro-fine-facebook-result-edpb-binding-decision_en"> </a>Irish Data Protection Commission's €1.2 billion GDPR fine on Meta in 2023 wasn't an outlier. It was a preview of what regulators are willing to do. Governance and speed are often treated as opposites. That framing is wrong, and it costs organizations the speed they're trying to protect. The better framing is whether the governance most leaders have built is the kind that works at the speeds they are chasing, and for most organizations, the answer is no. Well-designed governance covers four areas: access, quality, lineage and deployment. When all four work, teams operate within clear boundaries rather than starting every decision from scratch. The payoff is faster decisions and trust that doesn't break when regulators or executives ask hard questions. <section class="section main-article-chapter" data-menu-title="Access"> <h2 class="section-title">Access</h2> Access controls determine who can use which data and under what terms. This is where most day-to-day friction happens. Poorly designed access treats every request the same way. Every request goes through the same review, no matter how simple it is, and the approval queue gets longer over time. The problem gets worse when a single data set has multiple owners. A request must clear each owner in turn, and what should be a one-day approval becomes a three-week chain. Marketing teams can't run experiments fast enough to learn from their campaigns, and finance teams wait weeks to add a new analyst to quarterly reporting. By the time approval comes through, the team that needed the data has moved on. Well-designed access is mostly <a href="https://www.techtarget.com/data-technologies/feature/Data-governance-for-self-service-analytics-best-practices">self-service within set rules</a>, and ownership stays tight. Only the most critical stakeholders are listed as owners, so approval doesn't require chasing five people across three teams. Capital One does this with its <a href="https://www.capitalone.com/tech/cloud/data-management-at-scale/">"You Build, Your Data" framework</a>, a federated model that spreads ownership across the business, while a central team enforces shared standards. Teams move fast within their lane when ownership is clear. </section> <section class="section main-article-chapter" data-menu-title="Quality"> <h2 class="section-title">Quality</h2> Quality controls catch flawed data before it reaches a dashboard, a report or a model. This is where the gap between the visible cost and the hidden one is widest. Poorly designed quality controls assume the data is fine, letting problems show up later. A finance team finds its quarterly numbers were off two weeks before earnings. A marketing dashboard reports the wrong numbers because something changed upstream last month. The model trains on six months of bad data that nobody notices until it's in production. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Well-designed guardrails don't slow innovation. They're what let finished work ship. </figure> </div> </blockquote> Well-designed <a href="https://www.techtarget.com/data-technologies/tip/How-to-create-a-data-quality-management-process-in-5-steps">quality controls run automatically</a> at every stage.<a href="https://medium.com/airbnb-engineering/data-quality-at-airbnb-e582465f3ef7"> </a>Airbnb does this through its <a href="https://medium.com/airbnb-engineering/data-quality-at-airbnb-e582465f3ef7">Midas certification process</a>, which requires key datasets to meet defined standards for accuracy, ownership and documentation before the company trusts them. Quality badges are visible in every internal tool, so a data scientist or BI analyst can tell at a glance whether a data set is safe to build on.<a name="_p045llis4qcb"></a> </section> <section class="section main-article-chapter" data-menu-title="Lineage"> <h2 class="section-title">Lineage</h2> Lineage tracks where data came from and how it moved through the pipeline. It's the quietest of the four areas, and AI has raised its profile since training data provenance is now a compliance question. Outside of AI, many teams still treat lineage as optional until someone asks how a number was calculated. Auditors and regulators need to know where a number came from, and so do executives probing a quarterly result and CFOs trying to reconcile two reports that should match but don't. In<a href="https://www.sec.gov/newsroom/press-releases/2024-144"> </a>September 2024, the <a href="https://www.sec.gov/newsroom/press-releases/2024-144">SEC fined 12 financial firms a combined $88 million</a> for widespread failures to preserve and produce electronic communications when regulators asked for them. Stifel, Invesco and CIBC each paid penalties of $12 million or more. Well-designed lineage produces the answer in minutes. Brian Dummann, chief data officer and vice president of insights and technology at AstraZeneca, framed the broader principle in an October 2025 <a href="https://www.cdomagazine.tech/aiml/inside-astrazenecas-ai-strategy-cdo-brian-dummann-on-innovation-governance-and-speed">CDO Magazine interview</a>: "We're systematically looking at where we can redefine processes to go quicker. It's about removing bottlenecks without removing accountability." </section> <section class="section main-article-chapter" data-menu-title="Deployment"> <h2 class="section-title">Deployment</h2> Deployment controls decide what gets to production, and how. People often think of deployment as a model problem, but the same pattern applies to dashboards, reports and analyses that go to the board. Poorly designed deployment treats every release as a new negotiation. Every dashboard, every report, every model waits its turn while the backlog grows. Models that should be in production get stuck in review, and dashboards the business needs sit in draft. The work is done, but it never ships. The opposite risk is a model that scales before anyone understands its limits. Zillow's 2021 <a href="https://www.sec.gov/Archives/edgar/data/0001617640/000161764021000085/q32021991.htm">shutdown of its iBuying business</a> shows what that costs. Once the pricing algorithm was buying thousands of homes in a volatile housing market, Zillow found it was paying more for homes than it expected to sell them for. The result was a $304 million inventory write-down and layoffs affecting 25% of Zillow's workforce. Well-designed deployment follows a known path. Automated checks run in parallel, and teams have tested their rollback procedures. Routine releases don't restart the approval process, and monitoring catches a model or report that drifts after launch. Finished work reaches the business faster, and problems surface while they're still inexpensive to fix. </section> <section class="section main-article-chapter" data-menu-title="Why most programs fall short"> <h2 class="section-title">Why most programs fall short</h2> Organizations usually build governance in response to a problem that has already caused damage, but the controls often don't align with the actual risk. A single audit causes a heavy review of everything, even low-risk work. Controls are also often run by people rather than software because manual processes are easier to set up than automated ones. Humans end up reviewing routine decisions that don't need human judgment. This works with 20 practitioners, but it breaks with 200. The result is a governance team that acts as a gatekeeper, measured by what it blocks rather than what it enables. The relationship with the rest of the data team becomes tense, and teams treat even good controls as obstacles to be worked around. When teams complain that governance is slowing them down, the first reaction is usually to ask whether the controls are needed at all. That's not the right fight. Well-designed guardrails don't slow innovation. They're what let finished work ship. Data in successful organizations is often subject to strict governance, but they've made theirs fast, and that builds the trust the rest of the business depends on. Mostafa Ibrahim is a freelance software engineer and technical writer who covers machine learning, data engineering and cloud infrastructure for AI and SaaS companies. </section> Governance and speed are treated as opposites, but the real variable is design. Four control domains determine whether data guardrails help or hinder. https://cdn.ttgtmedia.com/rms/onlineimages/health%20security_g1418645103.jpg https://www.techtarget.com/data-technologies/feature/Does-setting-data-guardrails-slow-innovation Thu, 17 Sep 2026 12:42:00 GMT Does setting data guardrails slow innovation? Once primarily a data catalog provider for fueling analytics, Alation is turning its platform into a base for agentic AI. In July, Alation <a href="https://www.techtarget.com/data-technologies/news/366645666/Alation-launches-OS-for-building-governing-AI-agents">launched its AIOS</a> to provide Alation users with a dedicated environment for building and governing AI tools. It featured Agent Studio for development and AI Governance to keep agents' actions in <a href="https://www.techtarget.com/it-strategy/news/366646236/How-CIOs-can-navigate-federal-state-AI-regulation-uncertainty">compliance with AI regulations</a>. Two months later  during its annual revAlation user conference in Chicago, the vendor introduced <a href="https://www.techtarget.com/data-technologies/news/366646252/Context-is-king-as-agents-evolve-but-problems-persist">contextual data governance</a> features for its Alation Intelligence Operating System (AIOS). New capabilities, among others, include lineage tracing in AI Governance for visibility into the data that guides each agent's actions, native integrations with AI models and semantic layers, and Ontologies to better enable agents to understand an organization's unique characteristics. "The new features start to shift their governance from passive catalog documentation into an active, runtime enforcement system for enterprise data, context and AI agents," William McKnight, president of McKnight Consulting, told TechTarget. However, as the new capabilities become generally available -- most are in early access -- some will be limited in their scope, and humans will still need to be involved to oversee agent interactions with ontologies, McKnight continued. Alation's push toward agentic governance is in line with what <a href="https://www.techtarget.com/data-technologies/feature/15-top-data-catalog-software-tools-to-consider-using-in-2026">peers</a> such as Atlan, Collibra and <a href="https://www.techtarget.com/data-technologies/news/366643437/Informatica-update-aims-to-provide-trust-foundation-for-AI">Informatica</a> are doing rather than a competitive differentiator. "The entire data catalog market is pivoting from passive documentation for humans toward becoming the governed context layer for autonomous AI agents," McKnight said. "While Alation's execution, ecosystem neutrality, and analyst-friendly curation provide strong competitive leverage, several major vendors … are offering capabilities to solve the exact same AI trust gap." <section class="section main-article-chapter" data-menu-title="A governance suite for agents"> <h2 class="section-title">A governance suite for agents</h2> Lack of governance is one of <a target="_blank" href="https://www.growthaccelerationpartners.com/blog/why-agentic-ai-projects-never-make-it-to-production" rel="noopener">the main reasons</a> many AI projects still fail to make it into production. Alation built the AIOS to give customers a way to oversee AI tools so they can be trusted to perform as intended in production rather than produce <a target="_blank" href="https://www.evidentlyai.com/blog/ai-hallucinations-examples" rel="noopener">confident but incorrect answers</a> that can lead to financial and reputational harm, Alation co-founder and CEO Satyen Sangani told TechTarget in July. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> The new features start to shift their governance from passive catalog documentation into an active, runtime enforcement system for enterprise data, context and AI agents. </figure> <figcaption> William McKnightPresident, McKnight Consulting </figcaption> </div> </blockquote> The new features are designed to further that aim of engendering trust in agents and other AI applications, according to GT Volpe, Alation's vice president of product management. "Before, Alation could tell you what your data meant and whether it was trustworthy," he told TechTarget. "Now, it can make that meaning executable by agents on any platform, test it against real agent consumption, and trace every agent's risk back to the live state of the data underneath it." Alation's recognition of <a target="_blank" href="https://www.gartner.com/en/articles/genai-project-failure" rel="noopener">what hinders AI initiatives</a> also played a part in developing the new capabilities, Volpe continued. "The direct signal came from the customers furthest along in deploying agents on AIOS, [but] the broader signal was watching where AI programs stall," he said. "It's almost never the model. It's that the business meaning and process required to effectively use the data was never made explicit. ... That told us the problem enterprises actually have is a governance and context problem." Specifically, the new AIOS features include the following: <ul type="disc" class="default-list"> <li>Expanded lineage in AI Governance so users can better understand every agent's regulatory risk, including the data it consumes.</li> <li>Semantic Model Mastering, a tool that ingests <a href="https://www.techtarget.com/ai/tip/5-advantages-of-a-semantic-layer-for-enterprise-AI">semantic models</a> from Databricks and Snowflake, governs them in Alation, and syncs the enriched semantic definitions between the platforms.</li> <li>Native connectors in AI Governance to bring models from AWS, Databricks, Microsoft and Snowflake into a cross-platform registry so customers can connect an agent's compliance standing to the policies and quality of its underlying data.</li> <li>Console, a natural language interface that automatically routs user's requests to AIOS capabilities without forcing them toggle between different environments.</li> <li>Governed Collections, a feature that automatically turns <a href="https://www.techtarget.com/data-technologies/feature/Unstructured-data-needed-but-often-untapped-for-agentic-AI">unstructured data</a> from applications such as SharePoint and Confluent, into governed catalog objects so they can inform agents.</li> <li>Intelligent Feeds to automatically deliver <a href="https://www.techtarget.com/whatis/definition/business-logic">business logic</a> to teams making operating decisions.</li> <li>Guides for agents called Ontologies that enable them to understand and execute business processes.</li> </ul> The new AI Governance and Semantic Model Mastering capabilities are now generally available, while the other capabilities are in early access. Semantic Model Mastering is perhaps the most valuable of the new capabilities because it helps enterprises prevent their <a href="https://www.techtarget.com/data-technologies/feature/Data-governance-metrics-Measure-success-identify-issues">metrics</a> and definitions from "turning into the next enterprise mess" as agentic AI spreads business logic across disparate systems, according to Michael Ni, an analyst at Constellation Research. Meanwhile, Alation's additions to the AIOS are significant for users given that they demonstrate the vendor's shift toward governing actions, he continued. "Alation has made a meaningful market move with their latest release," Ni told TechTarget. "They are messaging their shift from governance as documentation to governance that guides actions. If agents are going to act on enterprise data, they need current definitions, lineage, policy and business context at the moment of use." Collectively, the new features provide <a href="https://www.techtarget.com/data-technologies/tip/9-metadata-management-standards-that-guide-success">a metadata foundation</a> for AI, he added. However, Ni noted that it remains to be seen how completely they enable agents to understand what matters in the moment and what agents are allowed to do at a given instant. Meanwhile, from a competitive standpoint, Alation's additions advance its AIOS but don't represent differentiation, according to Ni. "Alation is not alone in moving toward governed context for agents," he said, naming Databricks, Snowflake and Microsoft as competitors also offering such capabilities. "Alation could differentiate as the neutral layer across them. Semantic Model Mastering is a good example where, instead of creating another semantic layer, Alation is trying to reconcile the ones enterprises already have." McKnight highlighted Ontologies as the most important of Alation's AIOS additions because it helps connect agents with the business logic they need to avoid confidently wrong responses. Meanwhile, though it includes key features, he added that capabilities such as automated LLM evaluation, <a href="https://www.techtarget.com/ai/tip/Managing-drift-in-AI-models-and-data">model drift monitoring</a> and real-time execution guardrails would make AIOS a more thorough foundation for agents. "Evaluating it as a complete agent foundation highlights key operational gaps," McKnight said. "What Alation is delivering here is an essential contextual and regulatory layer, but enterprises must pair it with runtime guardrails, evaluation suites, and orchestration tools to complete their agent stack." </section> <section class="section main-article-chapter" data-menu-title="Looking ahead"> <h2 class="section-title">Looking ahead</h2> As Alation plots its product development plans for the remainder of 2026, its priority is to get Ontologies, Governed Collections, Intelligent Feeds and Console ready for general availability, according to Volpe. Turning Console into <a href="https://www.computerweekly.com/blog/Data-Matters/Get-set-for-interfaces-that-know-what-youre-doing-before-you-do">the interface</a> for AIOS is a focal point. "Console is the biggest change to how people experience Alation since the [data] catalog itself," he said, noting that the intent is to turn Console into a single entry point that interprets a user's intent and takes them to the proper workspace. "The focus is broadening what Console can route to, so the operating system behaves like a single integrated system the same way the intelligent applications built on it do." McKnight suggested that Alation can continue its platform's transition from a passive data catalog to an active data intelligence layer by automating <a href="https://www.techtarget.com/data-technologies/opinion/Data-observability-for-AI-helps-curb-poor-model-performance">data and AI observability</a>, and deepening metadata governance across streaming and unstructured data. In addition, he noted that Alation has room to improve the AI capabilities within its platform as well as augment its querying capabilities with <a href="https://www.techtarget.com/data-technologies/feature/Improve-data-visualizations-become-a-better-storyteller">visualizations</a> that enable users to better understand their data. "It can elevate its AI capabilities by embedding native reasoning models that explain logic alongside agents that automatically resolve pipeline issues, [while] enriching its core querying features with dynamic, accompanying visualizations will give users deeper, immediate insights into their underlying data assets," he said. Ni, meanwhile, advised Alation to use the knowledge it has regarding which agents get used and what <a href="https://www.techtarget.com/data-technologies/news/366646161/Data-management-vendors-race-to-connect-AI-with-context">sources they call upon</a> to provide more comprehensive direction for agents. "Alation can build from knowing which data gets used, which sources tend to be used together, and who are the experts," he said. "What I expect next from Alation is using those signals to provide dynamic guidance to agents beyond just the right data, but also what matters now and which policies apply given a specific intent and situation." Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. </section> Lineage tracing in AI Governance and governed access to unstructured data advance the platform and keep the vendor in step with what competitors are offering. https://cdn.ttgtmedia.com/rms/onlineimages/folder-files13.jpg https://www.techtarget.com/data-technologies/news/366650554/New-features-position-Alations-AIOS-as-AI-management-layer Thu, 17 Sep 2026 09:30:00 GMT New features position Alation's AIOS as AI management layer Graph database specialist Neo4j is upping its fight against financial crime. Just over one month after completing the acquisition of GraphAware, which was first <a href="https://www.techtarget.com/data-technologies/news/366643806/Neo4js-GraphAware-acquisition-targets-new-customer-segment">agreed upon in June</a>, Neo4j launched a new suite for detecting and stopping fraud that includes capabilities it inherited through the purchase. Neo4j GraphAware Financial Crime Intelligence is a <a href="https://www.techtarget.com/data-technologies/news/366593101/Graph-technology-helps-battle-election-misinformation">graph-native</a> tool featuring a graph-based knowledge layer purpose-built for banks and insurance companies to use AI to detect and prevent financial crime. Neo4j has provided fraud detection capabilities since launching its first graph database in 2010. However, after detecting potential incidents, customers had to carry out their investigations and analyses using other platforms. GraphAware Financial Crime Intelligence, which combines Neo4j's existing <a href="https://www.techtarget.com/data-technologies/news/252507769/Gartner-predicts-exponential-growth-of-graph-technology">graph-based</a> fraud detection capabilities with complementary capabilities from GraphAware, encompasses the entire financial crime investigation cycle from detection through alerting and investigation to decision. "Detection was only ever half the job," Devin Pratt, an analyst at IDC, told TechTarget. "Putting the full investigation on one graph-native stack removes the handoffs -- and the risk -- that live in the gaps between systems." Stephen Catanzano, an analyst at Omdia, a division of TechTarget, likewise noted that the significance of Financial Crime Intelligence is that it expands on Neo4j's <a href="https://www.techtarget.com/cybersecurity/feature/Why-companies-should-use-AI-for-fraud-management-detection">fraud detection capabilities</a>.  "What makes it particularly valuable is the addition of a comprehensive workflow … underpinned by a reusable knowledge layer for trustworthy AI that enables explainable decision-making throughout the entire detection, investigation and prevention process," he told TechTarget. <section class="section main-article-chapter" data-menu-title="A complete crime-fighting lifecycle"> <h2 class="section-title">A complete crime-fighting lifecycle</h2> Neo4j is not the only graph database specialist to offer fraud detection capabilities. TigerGraph, PuppyGraph and Memgraph do as well. In addition, broader-based data management vendors such as AWS and Oracle provide fraud detection tools as part of their database offerings. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Detection was only ever half the job. Putting the full investigation on one graph-native stack removes the handoffs -- and the risk -- that live in the gaps between systems. </figure> <figcaption> Devin PrattAnalyst, IDC </figcaption> </div> </blockquote> However, Catanzano noted that Neo4j is going beyond what most <a href="https://www.computerweekly.com/news/252524802/How-graph-technology-is-making-a-dent-in-the-database-market">competing vendors</a> offer by combining fraud detection with investigative and decision-making capabilities in a single feature. "While other vendors may offer fraud detection tools, Neo4j's approach of combining graph-native storage with a reusable knowledge layer that continuously enriches context and enables multi-hop reasoning across the entire investigation workflow positions it distinctively in the market," he said. Pratt similarly stated that Neo4j's differentiation lies in the depth and completeness of Financial Crime Intelligence. In particular, he highlighted the inclusion of an explainable <a href="https://www.techtarget.com/data-technologies/tip/5-knowledge-graph-use-cases-in-data-fabric-architecture">knowledge layer</a> that provides a map of connections between data points and shows how and why the tool's AI capabilities deemed something suspicious and made the recommendations it did. "Plenty of vendors do graph, or do fraud," Pratt said. "Neo4j's bet is owning the whole workflow, with explainability built in." While acquiring GraphAware enabled Neo4j to build a complete financial crimes detection suite, the vendor's impetus for developing Financial Crime Intelligence was also a response to the growing need for customers to show they are <a target="_blank" href="https://www.aciworldwide.com/blog/2026-fraud-trends-banks-must-prepare-for" rel="noopener">taking appropriate steps to fight fraud</a>, according to Michael Down, Neo4j's global head of financial solutions. "With the acquisition, we have the investigation and knowledge layer capabilities to package a … decision workflow, rather than requiring institutions to build it themselves," he told TechTarget. "The timing reflects both the acquisition coming together and a market under real pressure, with regulators pushing harder on institutions to prove proactive prevention, not just after-the-fact detection." </section> <section class="section main-article-chapter" data-menu-title="Yearend plans"> <h2 class="section-title">Yearend plans</h2> Neo4j's product development plans over the final months of 2026 include expanding and improving its knowledge layer for AI, adding agentic AI capabilities to its platform to aid users, and enabling customers to remain compliant with <a href="https://www.techtarget.com/data-technologies/feature/Data-sovereignty-expands-beyond-compliance-boundaries">data sovereignty</a> regulations by enabling them to control where and how their data models run, according to Down. "Financial Crime Intelligence is an early proof point of this approach [to data sovereignty], applying what we built with GraphAware to a specific vertical," he said. "We expect the same pattern to extend into other industries over time." Turning its knowledge layer into a foundation for other industry-specific capabilities is a wise strategy, according to Pratt, who noted that financial services is not the only vertical that benefits from <a href="https://www.techtarget.com/ai/feature/How-to-ensure-AI-transparency-explainability-and-trust">explainable decisions</a> based on connected data. "Build the knowledge layer once for fraud, and the payback comes from every regulated decision it can serve next," he said. Catanzano, meanwhile, suggested that Neo4j could improve its fraud detection and financial crime fighting prowess by adding AI capabilities such as predictive risk scoring and investigation recommendations to help analysts battle growing <a target="_blank" href="https://kpmg.com/ca/en/insights/2026/03/fraud-in-the-age-of-ai.html" rel="noopener">AI-powered fraud tactics</a>. In addition, like Pratt, he advised Neo4j to expand crime detection capabilities beyond financial services. "Neo4j might extend the platform to regulated industries facing similar networked crime challenges, such as healthcare fraud or supply chain integrity, thereby appealing to new market segments while deepening its position as the go-to graph intelligence platform for complex, relationship-driven risk detection across enterprises," he said. Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. </section> The vendor is adding tooling from its GraphAware acquisition to go beyond fraud detection and differentiate itself from those providing only part of the crime-stopping ecosystem. https://cdn.ttgtmedia.com/rms/onlineimages/security_a283199860.jpg https://www.techtarget.com/data-technologies/news/366649996/Neo4j-intros-graph-based-suite-for-fighting-financial-crime Wed, 16 Sep 2026 03:00:00 GMT Neo4j intros graph-based suite for fighting financial crime In early September, an outside group gained access to the network of Springfield Public Schools, the second-largest school district in Massachusetts, and cut staff off from online platforms the district depends on to run its schools. Among them was the system that tells nurses which children need medication or have documented allergies; and without it, Superintendent Sonia Dinnall canceled classes for four days, calling the decision a matter of student safety. On the day Springfield's closure began, the city of Everett, on the opposite side of the state, closed City Hall to the public after discovering a separate cybersecurity incident on its internal network two days earlier. Police, fire, public works and the city's school kept operating. Essential staff moved to another city building to maintain basic services, while residents continued to pay bills online through third-party systems that remained available. Whether the school district or the city lost data remains under investigation. What both lost, immediately and visibly, was the ability to operate normally. Cybersecurity planning tends to center on keeping attackers out and keeping data in, yet the damage that closed Springfield's schools<a href="https://www.techtarget.com/data-technologies/answer/Where-does-security-fit-into-a-business-continuity-plan"> came from lost access</a>. The functions each kept running relied on measures already in place, from separate networks to alternate ways of reaching families. The public sector is often where these disruptions are most visible. What Springfield and Everett illustrate about prevention, operational priorities and recovery planning applies to any organization that relies on networked systems and third-party platforms to do its core work. <section class="section main-article-chapter" data-menu-title="What the response revealed"> <h2 class="section-title">What the response revealed</h2> The details worth studying concern what each organization could still do once the attack was discovered. Springfield's IT team activated the district's cyberincident response plan as soon as it identified malicious traffic on the network. At a <a href="https://www.wbur.org/news/2026/09/09/springfield-cyber-attack-school-closures">Sept. 8 news conference</a>, CIO Robert St. Lawrence said the team's primary work was identifying, containing and eradicating the threats, with restoration of services to follow once that was confirmed. The district told staff to stay off its network and students to stop using school laptops to <a href="https://www.techtarget.com/cybersecurity/tip/12-common-types-of-malware-attacks-and-how-to-prevent-them">prevent the malware from spreading</a>, and officials said classroom instruction would return, even if it had to be conducted on paper. Everett kept essential services running while its IT staff, police department and outside cybersecurity specialists worked to restore the city's internal network. The city had appointed a new CIO in August as part of Mayor Robert Van Campen's push to modernize City Hall's technology, so that investment was already underway when the incident occurred. Officials have given no indication that the two incidents are connected, and that makes the timing more instructive: two separate organizations lost the use of core systems in the same week, adding to a string of similar incidents across Massachusetts over the past several years. For any enterprise, the useful question is whether it could <a href="https://www.techtarget.com/cybersecurity/tip/CISOs-guide-to-building-a-strong-cyber-resilience-strategy">keep operating through the same kind of disruption</a>. </section> <section class="section main-article-chapter" data-menu-title="Limit what an attacker can reach"> <h2 class="section-title">Limit what an attacker can reach</h2> Keeping attackers out is the first layer of security, and these incidents show why it won't be the last: once an intruder is inside, the damage depends on how much of the organization the intruder can reach. Everett's school district shows the safety effects of limiting reach. <a href="https://www.boston25news.com/news/local/everett-city-hall-closed-tuesday-after-cybersecurity-incident/KOUAMZDVIFHEBPTLZWSMN6PNVI/">In a statement</a>, Everett Public Schools Superintendent William D. Hart said the breach did not affect the school district, which runs on networks separate from the city's. Confining an intrusion to the environment where it started is the purpose of <a href="https://www.techtarget.com/it-infrastructure/tip/How-to-implement-network-segmentation-for-better-security">network segmentation</a>, and Everett's schools, on their own networks, kept operating throughout. Segmentation works best alongside measures that make it harder for an intruder to move once inside, including multifactor authentication, disciplined access controls and tight limits on administrative privileges. Christopher Smith, director of legislative and external affairs at Massachusetts' Executive Office of Technology Services and Security, said the state leads with those measures in its guidance to municipalities. This is baseline work. The steady run of incidents across Massachusetts suggests it is not yet in place as widely as it needs to be. </section> <section class="section main-article-chapter" data-menu-title="Spend against the right failure"> <h2 class="section-title">Spend against the right failure</h2> Every organization carries more risk than it has budget to address, and the instinct is often to spread resources broadly. A more effective approach is to <a href="https://www.techtarget.com/revcyclemanagement/news/366614016/How-healthcare-leaders-can-mitigate-concentration-risk">concentrate on the exposures</a> that would do the most operational damage if exploited. That starts with knowing what assets the organization has. An inventory is simple, but without one, there is no way to assess which systems are most exposed or which data matters most to daily operations. Springfield's closure hinged on student medical records, with transportation and food data also out of reach of outsiders, and an organization that has mapped its dependencies knows where that single point of failure sits before an incident exposes it. From there, the order of work is clear: assess where the highest-risk vulnerabilities sit, fix those first and resist the pressure to patch everything evenly. Smith said the state advises municipalities to "prioritize their most critical findings rather than trying to solve everything at once," and the same logic holds for enterprises. Everett's appointment of a new CIO about a month before the incident illustrates a priority that often loses out at budget time. Investment in the people who do this work is harder to justify in a funding cycle than a new one, and harder to make up for once it has been deferred too long. </section> <section class="section main-article-chapter" data-menu-title="Plan for operating without core systems"> <h2 class="section-title">Plan for operating without core systems</h2> Recovery is where the gap between planning and reality tends to be widest. In a <a href="https://www.techtarget.com/it-strategy/feature/The-incident-response-mistake-73-of-leaders-make">2026 survey of 600 senior security decision-makers</a> by incident response firm Sygnia, 99% said their organizations had a formal incident response plan, yet 73% said they would not be fully ready to execute it under real-world attack conditions. Springfield is a useful example because the district got the fundamentals right. It had a plan and activated it as soon as the threat was clear, and when email systems went down, staff kept families informed through text messages and social media channels. "We have highly skilled educators and teachers who are able to instruct without technology," Dinnall said at the Sept. 8 news conference. That is what functioning continuity looks like under pressure, and every enterprise should ask whether its own team could make the same transition. Beyond these two cases, a less visible recovery risk sits in the backup environment. When an attacker has reached the network, backups tied to the same credentials or infrastructure may already be exposed. In a 2024 <a href="https://www.sophos.com/en-us/blog/the-impact-of-compromised-backups-on-ransomware-outcomes">Sophos-commissioned survey</a> of 2,974 IT and cybersecurity professionals at organizations hit by ransomware, 94% said attackers had tried to compromise their backups, and 57% of those attempts succeeded. Where they did, median recovery costs reached $3 million, eight times the $375,000 at organizations whose backups were unaffected, and just 26% of those organizations fully recovered within a week, compared with 46% of those whose backups held. Smith said the state emphasizes continually testing and refining <a href="https://www.techtarget.com/cybersecurity/tip/How-to-create-an-incident-response-playbook">incident response plans</a> through simulation exercises and after-action reviews. Rehearsal is what shows whether a plan still holds when the network is down, the phones are offline, and the team is making decisions under pressure. Three priorities belong at the top of any organization's recovery preparation. The first is testing the plan under conditions that resemble a real attack, including the loss of email and phones. The second is verifying that backups are isolated from the credentials and infrastructure that an intruder is most likely to reach, and the third is deciding in advance how the organization will communicate internally and externally when its normal channels are compromised. </section> <section class="section main-article-chapter" data-menu-title="The real measure"> <h2 class="section-title">The real measure</h2> Springfield set Sept. 14 as the date for students to return, with district staff first reviewing records on allergies, medications and transportation. In Everett, Van Campen has pledged to strengthen and secure the city's systems, building on modernization work his administration began before the incident. Both organizations responded with competence and adaptability under difficult circumstances. The lessons from both cases follow a specific order. The first is knowing which systems and data the operation depends on, and the next is separating the most critical functions from the general network and directing limited resources toward the exposures that pose the greatest operational risk. The last is a recovery plan that assumes full compromise and has been tested before an incident occurs. Prevention decides whether an attacker gets in. Preparation determines what the organization can still do while an attacker is inside. Recovery planning decides how quickly it gets back to normal once the attacker is out. Scott Thompson is a site editor for TechTarget's Data Technologies group. </section> Cyberattacks on Springfield Public Schools and Everett City Hall show what it takes to keep operating when an intrusion cuts off access to core systems. https://cdn.ttgtmedia.com/rms/onlineimages/security_a135187239.jpg https://www.techtarget.com/data-technologies/feature/Cyber-recovery-plans-must-cover-lost-access-not-just-data Fri, 11 Sep 2026 16:09:00 GMT Cyber recovery plans must cover lost access, not just data The latest features from Alteryx advance its transformation from a data preparation specialist toward becoming a connective layer between enterprise data and AI tools. Unveiled on Sept. 9, the new capabilities include <a href="https://www.techtarget.com/data-technologies/feature/One-year-of-MCP-Support-a-must-for-data-management-vendors">a Model Context Protocol (MCP) server</a> and studio for transforming trusted datasets into agents. Collectively, they are designed to link AI applications with the information, including enterprise rules and previous analyses, that gives AI tools <a href="https://www.techtarget.com/data-technologies/news/366646252/Context-is-king-as-agents-evolve-but-problems-persist">the context</a> to perform properly. As a result, as enterprises <a target="_blank" href="https://www.deloitte.com/content/dam/assets-zone3/us/en/docs/services/consulting/2026/state-of-ai-2026.pdf" rel="noopener">increasingly try to move past experimenting</a> with agents and put AI tools into production, Alteryx's new capabilities are valuable for the vendor's users, according to Stephen Catanzano, an analyst at Omdia, a division of TechTarget. "These capabilities represent strong additions because they shift Alteryx … to becoming connective infrastructure between AI agents and enterprise business logic," he told TechTarget. Of particular importance is that the new features don't require Alteryx customers to create a new context layer each time they build an AI application, <a href="https://www.techtarget.com/enterprise-software/feature/AI-feature-spend-is-the-new-software-cost-control-problem">saving them both cost and effort</a>, Catanzano continued. "They [address] the practical challenge of organizations recreating the same business rules across multiple AI implementations while maintaining security, auditability and cost efficiency," he said. <section class="section main-article-chapter" data-menu-title="Changing with the times"> <h2 class="section-title">Changing with the times</h2> Before the AI era, enterprises often relied on specialists to build their data workflows, including vendors such as Alteryx for <a href="https://www.techtarget.com/data-technologies/feature/Top-data-preparation-challenges-and-how-to-overcome-them">data preparation</a>, Fivetran and Informatica for <a href="https://www.techtarget.com/data-technologies/feature/8-data-integration-challenges-and-how-to-overcome-them">data integration</a>, and Qlik and Tableau for data analysis. Now, AI tools can perform many data and analytics tasks on their own, reducing the need for specialized platforms. To remain relevant to existing customers and attempt to appeal to potential new ones, many data and analytics providers are evolving. For example, longtime self-service analytics specialists <a href="https://www.techtarget.com/data-technologies/news/366640059/GoodDatas-Context-Management-aims-to-make-AI-trustworthy">GoodData</a> and <a href="https://www.techtarget.com/data-technologies/news/366642778/Tableau-repositions-for-AI-unveils-new-knowledge-layer">Tableau</a> are re-positioning themselves as underlying layers for AI. Additionally, whether database providers such as <a href="https://www.techtarget.com/data-technologies/news/366649199/MongoDB-intros-latest-features-to-fuel-AI-development">MongoDB</a>, data platform vendors like <a href="https://www.techtarget.com/data-technologies/news/366644813/Databricks-intros-new-Genie-data-management-tools-to-aid-AI">Databricks</a> and <a href="https://www.techtarget.com/data-technologies/news/366643795/Snowflake-barrage-adds-more-AI-development-analysis-tools">Snowflake</a>, or governance layers including <a href="https://www.techtarget.com/data-technologies/news/366645666/Alation-launches-OS-for-building-governing-AI-agents">Alation</a> and <a href="https://www.techtarget.com/data-technologies/news/366580432/Collibra-launches-AI-Governance-unveils-GenAI-capabilities">Collibra</a>, AI has forced data technology vendors to adjust. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> These capabilities represent strong additions because they shift Alteryx … to becoming connective infrastructure between AI agents and enterprise business logic. </figure> <figcaption> Stephen CatanzanoAnalyst, Omdia </figcaption> </div> </blockquote> Alteryx likewise doing so with its latest capabilities. David Menninger, an analyst at ISG, noted that enterprises and software vendors are attempting to figure out the best mix of deterministic processing that relies on fixed rules and probabilistic processing that uses statistical patterns when building AI tools. For agents that execute tasks such as data engineering and analysis, deterministic processing, which results in a consistent output when using the same input, is generally preferred. "This is where Alteryx has earned its bread and butter," Menninger told TechTarget. "The new Alteryx features provide the glue between the data and analytics processes it manages and the AI agents enterprises are deploying to help with automation and enhanced decision making." Specifically, Alteryx's latest new capabilities include the following: <ul type="disc" class="default-list"> <li>Agent Studio, an environment where business users can transform existing, governed datasets into agents they can query using natural language.</li> <li>Alteryx MCP Server, a governed connection based on <a target="_blank" href="https://modelcontextprotocol.io/docs/2026-07-28/getting-started/intro" rel="noopener">the MCP open standard</a> that enables AI tools to interact with Alteryx to discover and operationalize relevant data.</li> <li>An updated version of Ask Alteryx that turns the conversational tool from an embedded assistant into the main interface for <a href="https://www.techtarget.com/data-technologies/news/366623973/Alteryx-One-launch-aims-to-unify-simplify-vendors-platform">the vendor's Alteryx One platform</a>.</li> <li>Alteryx Insights for OpenAI, an integration that enables users to derive insights in ChatGPT based on proprietary data and business logic previously approved by their organization's analysts. Similar integrations with Claude, Gemini, Slack and Microsoft Teams are being developed, according to Alteryx.</li> <li>Alteryx Skills, a <a href="https://www.techtarget.com/it-infrastructure/definition/GitHub">GitHub</a> integration that teaches agentic interfaces such as OpenAI Codex and Claude Code to develop Alteryx assets similar to the way Ask Alteryx builds them.</li> </ul> All but Alteryx Skills, which is in local preview on GitHub, are now generally available. User feedback led Alteryx to develop the new features, according to Ben Canning, the vendor's chief product officer. Through conversations with customers, Alteryx gleaned that enterprises were having trouble getting consistent answers to the same queries from agents built with different development tools such as Microsoft <a href="https://www.techtarget.com/enterprise-software/news/366633987/Microsoft-opens-Copilot-agent-building-to-office-rank-and-file">Copilot</a>, Salesforce <a href="https://www.techtarget.com/enterprise-software/news/366640897/Salesforce-Agentforce-head-discusses-future-of-AI-agents">Agentforce</a> and Amazon <a href="https://www.techtarget.com/ai/definition/What-is-Amazon-Bedrock-AWS-Bedrock">Bedrock</a>. "You end up with [different] answers to the same question and no way to tell which is right, and every agent is paying tokens to re-derive logic the business settled years ago," Canning told TechTarget. "That is the problem we built for." With the business logic that agents need to deliver consistent outputs living in existing Alteryx workflows, the vendor developed tools that connect agents with Alteryx's environment, he continued.  "The gap was that no agent could call them, so we built the layer that lets it," Canning said. "The business builds the logic once, governs it once, and exposes it to any AI system that asks." Menninger noted that Agent Studio is perhaps the most valuable of the new capabilities because it enables users to extend the value of their existing Alteryx workflows to agents. In addition, Alteryx's longstanding position as a connector between an organization's data and its analytics team is helping it stand apart from competing vendors as all data and analytics vendors <a href="https://www.techtarget.com/data-technologies/feature/Evolving-Alteryx-focusing-on-AI-ready-data-logic-for-agents">adjust to AI</a>, Menninger continued. "That unique position gives Alteryx building blocks to bring together the world of AI with data and analytics in ways that others can't necessarily do as easily," he said. Catanzano highlighted the MCP server because it enables agents to interact with Alteryx workflows in a governed and secure manner. Meanwhile, like Menninger, he noted that by positioning its capabilities between agents and enterprise data, Alteryx is carving out a somewhat unique position in <a href="https://www.computerweekly.com/blog/CW-Developer-Network/Series-brief-AI-workflow-platforms-tools">the AI workflow</a>. "These features position it as middleware or a business logic layer between AI agents and enterprise data rather than competing directly as another AI agent platform, which is a distinct architectural approach," he said. Other providers such Databricks, Microsoft and Snowflake provide AI-powered analytics, but they keep connections between data and AI functionality within their own environments, Catanzano continued. "Alteryx is explicitly designing for cross-platform governance and interoperability, which could provide competitive advantage for enterprises already committed to multiple AI tools," he said. </section> <section class="section main-article-chapter" data-menu-title="Alteryx's next steps in its ongoing evolution"> <h2 class="section-title">Alteryx's next steps in its ongoing evolution</h2> Over the final months of 2026, Alteryx is focused on increasing the <a href="https://www.techtarget.com/whatis/definition/transparency">transparency</a> of agent workflows to engender more trust in AI outputs, according to Canning. That includes explaining to an end user why an output is reliable and enabling agents to better understand the information they're acting upon. In addition, Alteryx plans to add capabilities that enable agents to know how well their results hold up over time. "Connecting an agent to governed logic is the part we have solved," Canning said. "Knowing what happened after it ran is the part nobody has." Menninger suggested that Alteryx could benefit from better messaging. He noted that the new capabilities serve existing customers, but the vendor needs to show potential new ones how it fits in <a target="_blank" href="https://mitsloan.mit.edu/ideas-made-to-matter/openness-control-and-competition-generative-ai-marketplace" rel="noopener">the evolving paradigm created by AI</a>. "The key thing Alteryx needs to do to attract new customers is help the market understand better where they fit and how they can complement enterprise AI deployments," he said. "Its existing customers already know this." Catanzano, meanwhile, advised Alteryx to build <a href="https://www.techtarget.com/it-infrastructure/feature/How-providers-industry-specific-cloud-offerings-impact-IT">industry-specific capabilities</a> that enable governed workflows to easily work with the operational systems where organizations execute their business processes. "Additionally, developing more sophisticated workflow templates or a marketplace where organizations can share anonymized, governed workflow patterns for common business processes like supply chain reconciliation or compliance reporting could help attract new users," he said. Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. </section> New features such as an MCP server and a studio for transforming datasets into agents aid the vendor's transformation and help distinguish itself from competing providers. https://cdn.ttgtmedia.com/rms/onlineimages/ai_a199952058.jpg https://www.techtarget.com/data-technologies/news/366650345/Alteryx-evolving-to-become-connective-tissue-for-AI Fri, 11 Sep 2026 06:50:00 GMT Alteryx evolving to become connective tissue for AI Concerns about losing control of data fed into AI models are preventing some enterprises from putting AI tools into production. Cloudera's new partnership with Mistral AI, unveiled on Thursday, will provide customers with tools to exert better command over the data that feeds agents and other AI applications. The partnership allows Cloudera users to integrate <a href="https://www.techtarget.com/ai/news/366625822/What-differentiates-Mistral-AI-reasoning-model-Magistral">Mistral's models</a> directly into their governed data in Cloudera's secure environment, effectively enabling developers to bring AI to their data. Historically, enterprises have had to move data out of their secure data management systems and combine it with models to feed AI <a href="https://www.techtarget.com/data-technologies/news/366646252/Context-is-king-as-agents-evolve-but-problems-persist">the context it needs</a> to deliver accurate outputs. Now, however, as <a target="_blank" href="https://www.deloitte.com/content/dam/assets-zone3/us/en/docs/services/consulting/2026/state-of-ai-2026.pdf" rel="noopener">more enterprises</a> put AI tools in production environments, some AI initiatives are stalling as users get more apprehensive about exposing sensitive information to external services, violating local regulations and boosting deployment costs beyond expectations. Data management providers such as Databricks, IBM and Snowflake have responded by integrating with model providers to reduce certain risks by letting customers bring AI to their data, including data kept in private clouds and on-premises environments. Cloudera's partnership with Mistral does the same. "This integration is significant because it underscores the popularity of data and AI sovereignty," Kevin Petrie, an analyst at BARC U.S., told TechTarget. "Volatile geopolitics, regulatory uncertainty and security threats are prompting AI adopters across the globe to reconsider their commitment to the cloud." Cloudera's partnership with Mistral, which has its headquarters in Paris, particularly benefits enterprises with operations in Europe that want to reduce their presence in the United States where AI regulations are <a target="_blank" href="https://www.twobirds.com/en/insights/2026/comparing-us-and-eu-ai-legislation-divergent-regulatory-approaches-and-practical-governance-implicat" rel="noopener">less stringent</a> than in Europe, he continued.  "As a French company, Mistral appeals in particular to European organizations and business units that want to limit their exposure to U.S. technology," Petrie said. <section class="section main-article-chapter" data-menu-title="Minimizing movement"> <h2 class="section-title">Minimizing movement</h2> Moving data costs money. Constantly doing so as agents call on data to carry out their tasks can lead to <a href="https://www.techtarget.com/enterprise-software/feature/AI-feature-spend-is-the-new-software-cost-control-problem">unexpected spending</a>. Meanwhile, moving data across environments can mistakenly violate <a href="https://www.techtarget.com/whatis/definition/data-sovereignty">data sovereignty</a> regulations, and there is the always risk of accidental exposure when shifting data from one system to another. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> This integration is significant because it underscores the popularity of data and AI sovereignty. Volatile geopolitics, regulatory uncertainty and security threats are prompting AI adopters across the globe to reconsider their commitment to the cloud. </figure> <figcaption> Kevin PetrieAnalyst, BARC U.S. </figcaption> </div> </blockquote> To bring AI capabilities into data management environments rather than force users to move data to build AI, Databricks now has <a href="https://www.techtarget.com/data-technologies/news/366631961/Databricks-partners-with-OpenAI-to-boost-AI-development">a native integration with OpenAI</a> that makes models available to users in the Databricks Intelligence Platform. In addition, among other vendors, IBM enables customers to integrate data and AI in <a href="https://www.techtarget.com/ai/news/366537216/IBM-launches-Watsonx-a-new-generative-AI-platform">watsonx</a>, and Snowflake makes models such as Google's Gemini 3 <a href="https://www.techtarget.com/data-technologies/news/366637132/Snowflake-boosts-Google-partnership-integrates-Gemini-3">natively available</a> in Cortex AI. "When you move data into an AI environment, you make a second copy that has to be secured, governed and kept current all over again," Mike Leone, an analyst at Moor Insights & Strategy, told TechTarget. "Bringing AI to the data avoids most of that overhead. ... And for regulated data that legally can't move, running the model in place is the only option." Now, by partnering with Mistral, Cloudera -- which in August introduced Cloudera Anywhere to enable users to build and run AI across multi-cloud and on-premises systems -- is similarly enabling customers to bring AI to their data. "It matters because Mistral's models can be downloaded and run inside a customer's own walls," Leone said. "Most frontier model providers only rent access through their own service. For Cloudera customers in banking or government working disconnected from the internet, a model they can run themselves is the only kind they can use." Cloudera, like many data management vendors that now provide AI development environments, makes a variety of AI models available to users, including <a href="https://www.techtarget.com/cybersecurity/video/Cutting-through-the-Claude-Mythos-hype-What-it-means-for-CISOs">Claude from Anthropic</a> and those that are part of <a href="https://www.techtarget.com/ai/news/366608795/Nvidia-launches-NIM-Agent-Blueprints-to-speed-AI-use">Nvidia's NIM microservices suite</a>. However, the vendor's partnership with Mistral marks the first time it is natively integrating AI models with its platform to eliminate the need to move data and enable users to build AI in private environments. "Model access alone doesn’t solve the challenge of putting AI into production," Abhas Ricky, Cloudera's chief business officer and GM of applied AI, told TechTarget. "Enterprises increasingly need a more complete approach that connects the model to their governed data. ... It's about moving beyond access to a model and giving customers what they need to turn that model into a production AI capability." While Cloudera's move to partner with Mistral enables customers -- especially those in Europe -- to more safely and <a href="https://www.computerweekly.com/news/366649643/Rising-token-bills-force-rethink-of-AI-cost-management">cost effectively</a> build and deploy AI tools, Petrie noted that there is a compromise to using Mistral's models. Because Mistral's default is to route and store data within the European Union, it simplifies staying compliant with Europe's data and AI regulations. However, not all AI models are created equally, and Mistral's, though more cost effective, do not perform as well as those from Anthropic, Google and OpenAI in <a target="_blank" href="https://artificialanalysis.ai/leaderboards/models" rel="noopener">benchmark testing</a> for reasoning. "There are tradeoffs," Petrie said. "While Mistral helps non-U.S. companies increase their sovereignty, its models generally do not perform as well as [leading] LLMs." Leone, meanwhile, noted that bringing AI to data -- though not unique -- remains rare enough that Cloudera's partnership with Mistral somewhat distinguishes Cloudera from other <a href="https://www.techtarget.com/data-technologies/opinion/The-race-to-build-the-ultimate-data-platform">data platform providers</a>. "Model partnerships are close to table stakes across data platforms now, and some already stretch to disconnected environments," he said. "The training piece is the sharper part here, since a Cloudera customer could build a custom model on their own data without moving that data. </section> <section class="section main-article-chapter" data-menu-title="Looking ahead"> <h2 class="section-title">Looking ahead</h2> As Cloudera looks ahead to the final months of 2026, enabling customers to run their data and AI wherever they need is a focal point, according to Ricky. "The Mistral partnership is an extension of that strategy because it brings advanced AI capabilities into the private, sovereign and air-gapped environments where many of our customers keep their most valuable data," he said. "We want customers to have consistent control over their data and AI regardless of where either is deployed." Another wise area to concentrate product development efforts would be <a href="https://www.techtarget.com/data-technologies/feature/Data-and-AI-governance-must-team-up-for-AI-to-succeed">AI governance</a>, according to Leone. He noted that Cloudera's security and data governance capabilities are highly regarded. Extending those <a href="https://www.techtarget.com/ai/feature/Agentic-governance-must-go-beyond-traditional-IT-practice">to agents</a> on an individualized basis would benefit Cloudera's existing customers and appeal to potential new ones. "Cloudera already has the tooling for building and running agents, so the next step is every agent carrying its own identity and permissions like an employee does," Leone said. "Governance is Cloudera's strongest ground, and agents are being viewed as the hardest thing to govern." Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. </section> The integration enables Cloudera customers to keep data secure when building and deploying agents and helps differentiate the vendor from some of its competitors. https://cdn.ttgtmedia.com/rms/onlineimages/code_g1019737194.jpg https://www.techtarget.com/data-technologies/news/366648767/Cloudera-partners-with-Mistral-to-bring-AI-to-users-data Thu, 10 Sep 2026 07:00:00 GMT Cloudera partners with Mistral to bring AI to users' data Databricks is once again aiming to improve the accuracy of data retrieval systems for AI. In January, Databricks <a href="https://www.techtarget.com/data-technologies/news/366637142/New-Databricks-tool-aims-to-up-agentic-AI-response-accuracy">launched Instructed Retriever</a>, an alternative to the traditional <a href="https://www.techtarget.com/ai/definition/What-is-retrieval-augmented-generation-RAG-in-AI">retrieval-augmented generation</a> (RAG) pipelines that many enterprises deploy to try to find and feed data to AI agents. With <a target="_blank" href="https://www.deloitte.com/content/dam/assets-zone3/us/en/docs/services/consulting/2026/state-of-ai-2026.pdf" rel="noopener">AI projects failing</a> far more than they succeeded heading into 2026, RAG, which searches for data based on users' queries, showed it was not good enough on its own at providing agents with relevant data. Databricks designed Instructed Retriever, powered by the Instructed-Retriever-1 model, to discover more contextually appropriate data by augmenting user queries with additional parameters to make data retrieval more precise. To further improve data retrieval for AI, Databricks is introducing Adaptive Instructed-Retriever, a new retrieval model specifically designed for complex searches. Instructed Retriever, which deploys parallel single-step retrieval capabilities to simultaneously search multiple parameters, works well for one-step queries but does not perform as fast or accurately when responses require multi-step searches. Adaptive Instructed-Retriever adds sequential search to parallel single-step retrieval to enable multi-step queries without adding <a href="https://www.techtarget.com/whatis/definition/latency">latency</a>. "Adaptive Instructed-Retriever is a significant upgrade because it enables agents to efficiently locate relevant … enterprise assets across large workspaces," William McKnight, president of McKnight Consulting, told TechTarget. "Its key advantage is it can use fast, single-step retrieval for straightforward requests while escalating to multi-step sequential search for more complex ones." Maintained speed, more relevant search results and <a href="https://www.techtarget.com/ai/tip/AI-inference-cost-optimization-An-enterprise-guide">lower costs</a> could all result from using the new model, he continued. "Unlike queries that are focused due to a single-source -- like a data lake query -- agent data retrieval is harder because it can be for anything … and the ways of ensuring which is the 'right data' for a particular agent task are nascent," McKnight said. "Adaptive retrieval addresses this by matching retrieval depth to query complexity to improve the accuracy, speed and cost of agent-driven data discovery." <section class="section main-article-chapter" data-menu-title="Improving accuracy"> <h2 class="section-title">Improving accuracy</h2> When Databricks launched Instructed Retriever, it benchmarked the tool against <a href="https://www.techtarget.com/ai/tip/How-to-prepare-data-for-your-RAG-pipeline">traditional RAG pipelines for AI</a> to demonstrate its proficiency. Adaptive Instructed-Retriever was similarly measured against competing capabilities to show its effectiveness. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Adaptive Instructed-Retriever is a significant upgrade because it enables agents to efficiently locate relevant … enterprise assets across large workspaces. </figure> <figcaption> William McKnightPresident, McKnight Consulting </figcaption> </div> </blockquote> In tests conducted by Databricks -- third-party evidence was not provided by the vendor -- the new model's recall capabilities were in line with those of models from <a href="https://www.computerweekly.com/blog/Cliff-Sarans-Enterprise-blog/Anthropic-Claude-incidents-show-the-limits-of-AI-guard-rails">Anthropic</a>, <a href="https://www.techtarget.com/cybersecurity/tip/What-CISOs-should-know-about-DeepSeek-cybersecurity-risks">DeepSeek</a> and <a href="https://www.techtarget.com/searchsecurity/news/366646755/What-CISOs-can-learn-from-the-Hugging-Face-OpenAI-incident">OpenAI</a>. However, Adaptive Instructed-Retriever executed its searches significantly faster. "This release is incremental for users but significant for the product," Donald Farmer, founder and principal of TreeHive Strategy, told TechTarget. "Adaptive Instructed-Retriever adds multi-step search for questions that the previous Instructed Retriever would get wrong, so a user gets better answers with a fixed latency budget at low cost." Farmer noted that enterprises continue to struggle to discover and operationalize the right data for agents because finding information <a href="https://www.techtarget.com/data-technologies/feature/Controlling-data-sprawl-requires-governance-discipline">across disparate systems</a> is difficult, especially when query instructions can differ from user to user. As a result, agile searches rather than rigid ones are beneficial. "In the enterprise, retrieval is not one problem," Farmer said. "Data is spread over tables, notebooks, dashboards and documents, and getting the right answer often depends on instructions that require interpretation. Plain similarity search can't help." </section> <section class="section main-article-chapter" data-menu-title="Differentiating with governance"> <h2 class="section-title">Differentiating with governance</h2> Although beneficial for users and an important advancement of Databricks' data retrieval capabilities for AI, Adaptive Instructed-Retriever itself is not a competitive differentiator for the vendor, according to McKnight. For example, Snowflake similarly offers a model that delivers text-to-SQL and <a href="https://www.techtarget.com/ai/tip/5-advantages-of-a-semantic-layer-for-enterprise-AI">semantic-based retrieval</a> across widespread enterprise data estates. However, what will appeal to Databricks' users is Adaptive Instructed-Retriever synchronization with Unity Catalog, the vendor's centralized governance layer. "Databricks differentiates [by anchoring] dynamic search, predicate extraction and governance directly into Unity Catalog, which serves as the optimized backbone for its conversational data tools, including Databricks Genie, Databricks Assistant and the Mosaic AI Agent Framework," McKnight said. Farmer likewise noted that Databricks is not the first vendor to provide a small model specialized for data retrieval. Google Cloud and Microsoft -- among others -- offer tools that plan and run parallel subqueries to discover and <a href="https://www.techtarget.com/data-technologies/news/366646252/Context-is-king-as-agents-evolve-but-problems-persist">operationalize relevant data</a>. Databricks, however, goes further with Adaptive Instructed-Retriever by training the model to know when to stop searching, according to Farmer. "When an AI agent searches for data, it can take one look and answer, or it can keep looking," he said. "More looks usually give a better answer, but each one takes time and costs money. The hard part is knowing when to stop. Databricks trained a small model to make the stopping decision itself." In addition, Farmer posited that the savings which the new model could lead to might force <a href="https://www.techtarget.com/data-technologies/opinion/The-race-to-build-the-ultimate-data-platform">competing vendors</a> to respond. "Customers do not choose a data platform for a retrieval model, and the competitors' own retrieval layers are probably good enough, but the argument for cost-control is compelling," he said. "So, every vendor will need a cheap, fast retrieval layer for agents. But they'll do it their own way, I expect." </section> <section class="section main-article-chapter" data-menu-title="Next steps"> <h2 class="section-title">Next steps</h2> Databricks developed Adaptive Instructed-Retriever after observing customers that have begun to put agentic AI tools into production ask increasingly intricate questions of those agents, according to Michael Bendersky, the vendor's director of research. "Agentic search requires multiple rounds of retrieval using tools, reviewing the information and then searching again," he told TechTarget. "In these environments, the agents search, retrieve, and then examine again until they find the right answer. At the same time, you don't want the user to wait longer than necessary for the system to finish the search." Over the next few months, Databricks will add serverless <a href="https://www.techtarget.com/enterprise-software/definition/What-is-a-graphics-processing-unit-GPU">GPU</a> compute capabilities -- now in public preview -- to Adaptive Instructed-Retriever to optimize the model's usage by automatically scaling up when needed and scaling down when workload demands decrease, Bendersky continued. In addition, he said that Databricks plans to continue to reduce barriers enterprises face when moving AI into production by addressing <a href="https://www.techtarget.com/data-technologies/news/366646161/Data-management-vendors-race-to-connect-AI-with-context">context for agents</a>, AI governance, cost control and the openness of its platform. Continuing to address the quality and performance of existing capabilities to enable customers to build and deploy reliable AI capabilities is a wise strategy for Databricks, according to McKnight. Specifically, improving the accuracy rate of its model serving capabilities and simplifying the use of <a href="https://www.techtarget.com/data-technologies/feature/Databricks-lakehouse-a-secret-weapon-for-WNBAs-Fever">its platform</a> would be beneficial. "To support enterprise-grade AI agents, Databricks must resolve its real-time concurrency bottlenecks and high model serving failure rates to ensure stable, low-latency agent execution loops," McKnight said. "Additionally, they need to simplify the overall user experience by shifting toward a zero-administration serverless platform and offering prebuilt operational templates … to accelerate business ROI." Farmer, meanwhile, suggested that Databricks further improve its data retrieval capabilities by better enabling them to choose which tables, metrics and semantic definitions to use when searching for <a href="https://www.techtarget.com/ai/post/AI-agents-are-only-as-smart-as-the-data-that-feeds-them">which assets to feed AI tools</a>. "The hard problem for data agents is not finding documents, but choosing among thousands of tables, metrics and semantic definitions," he said. "I think that's a future demand [from customers]." Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. </section> Adding sequential search to improve the speed and accuracy of responses to complex queries evolves existing capabilities and keeps Databricks in step with competitors. https://cdn.ttgtmedia.com/rms/onlineimages/machine%20learning_g1186820873.jpg https://www.techtarget.com/data-technologies/news/366650164/Databricks-launches-model-to-boost-data-retrieval-for-AI Thu, 10 Sep 2026 06:56:00 GMT Databricks launches model to boost data retrieval for AI Boomi is betting big on a control layer for AI. The vendor, which began as <a href="https://www.techtarget.com/data-technologies/news/366554135/New-Boomi-AI-tool-enables-natural-language-data-integration">a data integration specialist</a> but now focuses on enabling customers to activate their data for AI, on Tuesday launched the Agent Control Plane. The new layer brings previously disparate Boomi capabilities together in a single environment and is designed to provide customers with capabilities that enable them to connect AI agents with business systems, govern agents in production and <a href="https://www.techtarget.com/enterprise-software/feature/AI-feature-spend-is-the-new-software-cost-control-problem">rein in spending on AI</a>. Keeping control over agents and the cost of doing so are rising concerns for many enterprises as they try to <a href="https://www.deloitte.com/content/dam/assets-zone3/us/en/docs/services/consulting/2026/state-of-ai-2026.pdf">move past experiments with AI</a> and put agents into production. Boomi's Agent Control Plane is therefore a valuable addition for the vendor's users, according to Michael Ni, an analyst at Constellation Research. "Boomi is making a bid for one of the more valuable control points in enterprise AI, the layer between agents and the systems where business actually gets done," he told TechTarget. "As companies add more models and agents, they need a way to control what those agents can access, what they can do and what they cost." Boomi is not the only vendor adding tools aimed at giving customers better control over their agents, Ni continued, noting that AWS, Salesforce and ServiceNow are among the others that also provide features for orchestrating agents. However, Boomi's integration capabilities give it a foundation from which to build <a href="https://www.techtarget.com/it-strategy/feature/The-AI-agent-governance-gap-How-CIOs-can-gain-control">a control plane for agents</a> that could help it compete for market share. "The idea of an Agent Control Plane is quickly becoming a crowded but strategic land grab," Ni said. "Boomi's advantage is … to turn its footprint into a neutral data activation layer where enterprise data can be contextualized, governed and acted upon across models and platforms." <section class="section main-article-chapter" data-menu-title="Control vs. chaos"> <h2 class="section-title">Control vs. chaos</h2> Agents that are improperly overseen can cause serious problems. For example, an agent powered by Anthropic's Claude large language model <a target="_blank" href="https://www.theguardian.com/technology/2026/apr/29/claude-ai-deletes-firm-database" rel="noopener">deleted a company's entire production database</a> last April. Two months earlier, another agent allegedly <a target="_blank" href="https://sumsub.com/media/news/developer-warns-ai-agent-defamation-post-shows-risks-of-autonomous-ai/" rel="noopener">wrote a blog post</a> defaming a developer after the developer rejected the agent's code contribution. Beyond headline-grabbing mishaps, agents that aren't closely controlled can drive up an organization's spending by getting caught in recursive loops and consuming unlimited amounts of tokens. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> Boomi is making a bid for one of the more valuable control points in enterprise AI, the layer between agents and the systems where business actually gets done. As companies add more models and agents, they need a way to control what those agents can access, what they can do and what they cost. </figure> <figcaption> Michael NiAnalyst, Constellation Research </figcaption> </div> </blockquote> Gartner <a target="_blank" href="https://www.gartner.com/en/newsroom/press-releases/2026-05-26-gartner-says-applying-uniform-governance-across-ai-agents-will-lead-to-enterprise-ai-agent-failure" rel="noopener">predicts</a> that by the start of 2027, 40% of organizations will demote or decommission agents due to a lack of proper governance. Boomi's Agent Control Plane is designed to provide the governance enterprises need to trust that their AI tools appropriately connect to enterprise systems such as Salesforce and Workday, act only as intended in production, and do so within an expected spending range. As a result, though other vendors are similarly addressing AI governance, Boomi's control layer for agents is significant for its customers, according to Stephen Catanzano, an analyst at Omdia, a division of TechTarget. "The Agent Control Plane represents a big addition because it addresses what Boomi identifies as the largest unmanaged risk in enterprise technology today, [which is] giving AI agents the deep access they need to transactional systems while maintaining governance, cost control, and security that most organizations currently lack," he told TechTarget. Meanwhile, there is some potential differentiation between Boomi's approach to governing AI and competing capabilities from Salesforce through <a href="https://www.computerweekly.com/blog/CW-Developer-Network/MuleSoft-Omni-Gateway-offers-route-to-agentic-visibility-consistent-governance">MuleSoft</a>, AWS and specialists such as Apigee, Catanzano continued. "Boomi seems to be betting that their unified approach sitting between any agent and any system, rather than being tied to a single cloud ecosystem, provides the differentiation," he said. Boomi's Agent Control Plane is designed to be flexible, running across public clouds, customers' virtual private clouds and on-premises environments. In addition, it supports data and <a href="https://www.techtarget.com/cybersecurity/feature/The-push-for-digital-sovereignty-What-CISOs-need-to-know">digital sovereignty</a> to ensure that customers remain compliant with the rules and regulations where their information is created. The Agent Control Plane combines pre-existing Boomi capabilities to provide the following: <ul class="default-list"> <li>A single, centralized location for overseeing agents and tools across platforms, including putting limits on <a href="https://www.techtarget.com/whatis/definition/token">token</a> consumption and stopping high-risk actions until they have human approval.</li> <li>Enforcement through the Boomi AI Gateway, which combines MCP Gateway and LLM Gateway in one control panel and is built on the technology the vendor acquired when <a href="https://www.constellationr.com/insights/news/boomi-buy-lunardev-eyes-intelligent-prompt-and-model-routing">it bought Lunar.dev</a>.</li> <li>Governed connectivity through Boomi Connect MCP servers so that agents can safely call on enterprise systems to inform their actions.</li> <li>Deployment flexibility via a hybrid runtime architecture to adhere to data sovereignty rules, secure sensitive information and control spending.</li> <li>Data semantics and lineage to ground agent reasoning with approved data.</li> <li>Natural language processing capabilities in Boomi Orchestrate so that non-technical teams can build agentic AI workflows.</li> </ul> "This came directly from what we were hearing in a critical mass of customer conversations, not from a single feature request," Ed Macosky, Boomi's chief product and technology officer, told TechTarget. User feedback consistently included concerns over <a href="https://www.computerweekly.com/news/366645085/Boomi-CEO-shares-vision-of-AI-cost-management">increasing spending on AI</a>, difficulty connecting agents with disparate transactional systems and fear of losing control over intellectual property once it's accessed by agents, he continued. "The Agent Control Plane exists because enterprises shouldn't have to choose between moving fast on AI and keeping control of their data and infrastructure," Macosky said.  Meanwhile, Boomi's control layer for AI appears logically constructed by combining centralized visibility with enforcement capabilities, deployment flexibility and <a href="https://www.techtarget.com/data-technologies/opinion/Tracing-data-lineage-in-AI-systems">data lineage</a>, according to Catanzano. "The architecture directly maps to the three problems Boomi highlights -- governance gaps that cause production incidents, runaway token costs that have made AI spend management the most in-demand FinOps skill, and trust deficits," he said. Ni likewise noted that the Agent Control Plane features the capabilities needed to provide Boomi's customers with a single environment for <a href="https://www.techtarget.com/data-technologies/feature/Data-and-AI-governance-must-team-up-for-AI-to-succeed">governing AI</a>. In addition, it does so by combining tools that have already been proven to work. "Boomi has put the major pieces together and built on areas where it already has known strengths," Ni said. "Boomi's next step will be to go deeper into runtime context and managing agents across their full lifecycle. ... That becomes increasingly important as enterprises move from experimenting with agents to operating them at scale." </section> <section class="section main-article-chapter" data-menu-title="Looking ahead"> <h2 class="section-title">Looking ahead</h2> After launching the Agent Control Plane , Macosky said that Boomi's product development plans for the rest of 2026 focus on improving AI governance across providers and models, broadening its partnership network to add openness, and enhancing basic platform features such as <a href="https://www.techtarget.com/data-technologies/feature/Why-AI-forces-security-first-governance">security</a> and reliability. One specific way the vendor could better serve its customers would be to add tools that show whether agents in action delivered their desired outcome, according to Ni. "Boomi already connects the data, exposes the tools, governs the agent and orchestrates the action. Now it needs to show whether that action delivered the desired business outcome, then use that feedback to improve the next decision," he said. "That would make Boomi’s positioning around data activation more tangible to CIOs and business leaders." Catanzano similarly suggested that Boomi expand beyond governance and connectivity to measure and improve the <a href="https://www.techtarget.com/enterprise-software/feature/Software-value-needs-outcome-measures">business outcomes</a> generated by agents. "Boomi could continue serving users and attract new ones by … helping organizations not just control what their agents do, but systematically improve how well they do it through analytics that connect agent actions to actual business metrics," he said. Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. </section> The Agent Control Plane combines previously disparate capabilities in one environment to help users better observe and command their fleets of agents. https://cdn.ttgtmedia.com/rms/onlineimages/machine_learning_g1303163039.jpg https://www.techtarget.com/data-technologies/news/366649690/Boomi-intros-layer-for-controlling-AI-costs-and-connections Wed, 02 Sep 2026 09:49:00 GMT Boomi intros layer for controlling AI costs and connections Vector database specialist Qdrant on Tuesday launched Fineweb-10B, a public dataset for enterprises to use when benchmarking their vector retrieval capabilities. The dataset uses information from FineWeb, an open-source dataset created by <a href="https://www.techtarget.com/whatis/definition/Hugging-Face">Hugging Face</a> that contains more than 15 trillion pieces of text for training large language models. Vector embeddings have become <a href="https://www.techtarget.com/data-technologies/feature/Vector-search-now-a-critical-component-of-GenAI-development">a critical part of the AI pipeline</a>. Vectors are algorithmically assigned numerical representations that make data -- including <a href="https://www.techtarget.com/data-technologies/post/Unstructured-data-is-the-bottleneck-for-agentic-AI-success">unstructured data</a> -- easy to search and discover. Once discovered, whether by agents or human users, vectorized data can be used to inform analytics and AI tools. However, many organizations don't fully understand the speed and accuracy of their vector retrieval systems at the scale now demanded by AI development, and whether their vectorized data can be trusted to inform applications in production environments. As a result, Qdrant's launch of a massive public dataset for testing vector retrieval speed and accuracy at enterprise scale is valuable, according to William McKnight, president of McKnight Consulting. "Having done a number of benchmarks evaluating [large-scale retrieval], I can attest there is a massive compute barrier needed to calculate exact ground truth across 10 billion vectors," he told TechTarget. "This dataset enables organizations to accurately measure recall … on realistic, non-synthetic web data under true production scale." In addition, building the dataset with capabilities from Hugging Face, the <a target="_blank" href="https://commoncrawl.org/" rel="noopener">Common Crawl</a> Foundation, Vultr and Alibaba lends it credibility, McKnight continued. "Because enterprise IT teams lack the dedicated methodology and neutral vantage point to properly configure and stress-test these massive workloads and vendors lack independence, partnering with an independent, reputable third party is essential to obtain credible, production-grade evaluations free from vendor bias," he said. <section class="section main-article-chapter" data-menu-title="Open-source value"> <h2 class="section-title">Open-source value</h2> Qdrant's origins are in open-source technology. Founded in 2021 and based in Berlin and New York City, the vendor began as an open-source project published on GitHub before launching a proprietary version of its vector database in 2023. Meanwhile, with both versions of its vector database built with the open-source <a href="https://rust-lang.org/">Rust programming language</a>, Qdrant remains closely tied to the open-source community. <blockquote class="main-article-pullquote"> <div class="main-article-pullquote-inner"> <figure> This dataset enables organizations to accurately measure recall … on realistic, non-synthetic web data under true production scale. </figure> <figcaption> William McKnightPresident, McKnight Consulting </figcaption> </div> </blockquote> Developing a public dataset for benchmarking vector retrieval capabilities reflects the vendor's recognition that public benchmarks that measure the rising demands being placed on vector retrieval systems are needed, according to Nathan LeRoy, a research engineer at Qdrant Labs. "Vector search workloads have already hit billion vector scale. Enterprises need benchmarks to reflect this so that they can make informed decisions at this scale," he told TechTarget. "Ideally, such benchmarks and their datasets are public and accessible, so it's easy to make comparisons and draw conclusions across search systems." While Qdrant recognizes the value of the open-source tools, McKnight noted that vendors sometimes release open-source capabilities even if they aren't closely aligned with the open-source community. Google, for example, developed the Agent2Agent protocol in May 2025 to provide an open standard for connecting and <a href="https://www.techtarget.com/enterprise-software/news/366636690/Agentic-orchestration-the-next-AI-issue-for-CIOs-to-tackle">orchestrating agents</a> and donated it to the Linux Foundation. However, the reasons vendors develop open-source capabilities are not always magnanimous, McKnight continued. While it shows that the vendor is committed to its market and is aligned with cutting-edge thought, it also can lead to greater visibility and attract new potential customers. "It isn't uncommon for vendors to release open-source benchmarking datasets and tools, especially ones they are likely good at competitively," he said. Michael Ni, an analyst at Constellation Research, similarly noted that Qdrant's launch of an open-source dataset for benchmarking vector retrieval is beneficial to enterprises as they <a href="https://www.techtarget.com/ai/tip/Citizen-developers-are-redefining-enterprise-AI-development">build AI tools,</a> as well as the vendor itself. "A public dataset like Fineweb-10B is an ecosystem and category-building move," he told TechTarget. "Qdrant is effectively subsidizing the cost of evaluation … and position[ing] Qdrant as central to that evaluation. Qdrant gains mindshare, creates de facto standards and evaluation criteria, and increases the odds [potential customers] ask the kinds of questions Qdrant is already optimized to answer." </section> <section class="section main-article-chapter" data-menu-title="Up to the test"> <h2 class="section-title">Up to the test</h2> Qdrant Fineweb-10B is aimed at enabling organizations to benchmark vector retrieval at enterprise scale. Toward that end, Qdrant built the dataset using vector embeddings of real documents rather than <a href="https://www.techtarget.com/searchenterpriseai/tip/What-to-know-about-synthetic-data-as-a-business-advantage">synthetic data</a>, and a corpus -- a collection of data -- of 10 billion documents so it is large enough to ensure that size-dependent behavior shows up during testing. In addition, the dataset includes a grounded truth query set, which is a list of search prompts paired with their ideal responses against which an enterprise can measure the accuracy of its benchmark tests. Ni noted that Fineweb-10B is not the first public dataset for benchmarking vector retrieval. However, the size of the new dataset differentiates it from others. "This is an important extension of existing benchmark work, not a brand-new category of benchmarking," he said. "Qdrant has contributed a much larger, more application-like retrieval workload with known correct answers that allow engineering teams' tests to go beyond throughput to how accurately the system behaves as corpus size, filtering and retrieval complexity increase." Meanwhile, the dataset is appropriately constructed to aid organizations seeking to better understand the effectiveness of their vector retrieval systems, Ni continued, while adding that it is only one part of what should be <a href="https://www.techtarget.com/it-infrastructure/tip/How-effective-is-your-AI-agent-9-benchmarks-to-consider">a broader evaluation process</a>. "Qdrant-Fineweb-10B certainly solves the problem Qdrant says it is solving, measuring whether vector retrieval remains accurate and performant at very large scale," Ni said. "At the same time, companies still need to test against their own data, business filters, re-ranking and downstream agent outcomes as part of a total enterprise retrieval approach versus just testing a retrieval engine." McKnight similarly noted that Qdrant-Fineweb-10B is well constructed, though limited in scope. "The dataset seems well-designed for pairing realistic web-text distributions with exact ground truth across dense, sparse and filtered queries," he said. "However, fully leveraging it seems to require substantial compute and storage infrastructure to ingest and index 10 billion records, and it does not test real-time streaming updates." <figure class="main-article-image full-col" data-img-fullsize="https://www.techtarget.com/rms/onlineimages/traditional_search_vs_vector_search-f.png"> <img data-src="https://www.techtarget.com/rms/onlineimages/traditional_search_vs_vector_search-f_mobile.png" class="lazy" data-srcset="https://www.techtarget.com/rms/onlineimages/traditional_search_vs_vector_search-f_mobile.png 960w,https://www.techtarget.com/rms/onlineimages/traditional_search_vs_vector_search-f.png 1280w" alt="A graphic shows the differences between keyword search and vector search. " data-credit="Informa TechTarget " height="319" width="560"> <div class="main-article-image-enlarge"> </div> </figure> </section> <section class="section main-article-chapter" data-menu-title="Looking ahead"> <h2 class="section-title">Looking ahead</h2> As Qdrant plans product development, speed, accuracy and openness are focal points, according to LeRoy. In April, the vendor's <a href="https://www.techtarget.com/data-technologies/news/366642580/Qdrant-boosts-performance-reliability-to-meet-AI-needs">Qdrant Cloud update</a> included features aimed at improving performance. "As 2026 draws to a close, we’ll continue to push the limits of vector search for speed, accuracy, and being deployable anywhere," LeRoy said. Focusing on speed and accuracy is wise, according to Ni. He noted that vector database specialists such as Qdrant and Pinecone now face competition from broad-based data management providers, including hyperscale cloud vendors <a href="https://www.techtarget.com/data-technologies/news/366577632/Vector-search-and-storage-key-to-AWS-database-strategy">AWS</a> and <a href="https://www.techtarget.com/data-technologies/news/366583139/Oracle-adds-vector-search-capabilities-to-database-platform">Oracle</a>, that provide vector search. As a result, specialists need to distinguish themselves. "Vector search is becoming table stakes," Ni said. "Most enterprises can get it from databases they already own, so specialists like Qdrant have to drive what makes them distinct and valuable to the market by delivering better retrieval quality, latency and cost at scale." McKnight, meanwhile, suggested that better integrating <a href="https://www.techtarget.com/searchitchannel/definition/feedback-loop">feedback loops</a> and adding to its <a href="https://www.techtarget.com/data-technologies/tip/Top-5-metadata-management-best-practices">metadata management</a> capabilities would help Qdrant better serve existing customers and perhaps draw in new ones. "By natively integrating generative feedback loops, [Qdrant would remove] the need for teams to build these pipelines via external APIs," he said. "Additionally, broadening its rich scalar metadata capabilities to include multidimensional matrices and tensors would enhance support for complex data structures." Eric Avidon is a senior news writer for Informa TechTarget and a journalist with more than three decades of experience. He covers analytics and data management. </section> With AI tools requiring massive volumes of vectors to perform properly, the public dataset is designed to help organizations test their systems in real-world conditions. https://cdn.ttgtmedia.com/rms/onlineimages/code_g1195673150.jpg https://www.techtarget.com/data-technologies/news/366649627/Qdrant-builds-dataset-to-benchmark-vector-retrieval-at-scale Tue, 01 Sep 2026 09:00:00 GMT Qdrant builds dataset to benchmark vector retrieval at scale Search Data Management Resources and Information from TechTarget 60 webmaster@techtarget.com