AI agents are quietly transforming daily work at Dos Pinos, a large Costa Rican dairy cooperative. With about 6,000 employees, Dos Pinos is scaling Copilot-built agents across the business, deploying roughly 80 so far for tasks, like catching packaging errors, automating reporting or legal drafting. It’s all part of a push to embed AI into daily work at scale. Dos Pinos’ creative lead, for instance, built an “AI inspector” in Microsoft Copilot Studio to compare packaging labels against technical sheets and flag regulatory discrepancies before files ship, helping drive errors toward nearly zero and easing team stress.

Everything Suarve, an Australian nonprofit, has won accolades for turning young lives around by arming at-risk youth with construction as well as life skills. As the nonprofit grows, it’s turning to Microsoft technology to handle paperwork from managing referrals to securing funding. The end result, says founder Joseph Te Puni-Fromont, is “we have more time to go put back into the young people.”

One of the world’s oldest and most regulated banks is using AI to reshape how its work gets done. BNY is training thousands of employees to design AI solutions and create governance pipelines so its people can be freed up to do higher-value work. Using Microsoft tools and infrastructure, the bank is reworking its org chart and betting that AI will expand, not replace, human potential.

Edge now brings Copilot directly into the browser on desktop and, for the first time, mobile. Microsoft announced new capabilities today, including reasoning across multiple open tabs, more relevant answers shaped by browsing history and past chats, and hands-free use with Voice and Vision. A redesigned new tab page and Journeys are now widely available, while Copilot Mode is retired as these features are now built into Edge.

AI isn’t just changing how people work — it’s changing what their work is. In a conversation on Microsoft’s WorkLab podcast, Katy George, a vice president for workforce transformation, says companies falter when they treat AI as a simple rollout. Real change, she says, comes from redesigning work itself — unlocking new capabilities, reshaping roles and requiring workers to continuously adapt as their roles evolve.

AI is modernizing the world’s most complex industries, from factory floors to building materials to geotechnical data, by capturing, scaling and democratizing expert knowledge. Examples from Japan, Mexico, New Zealand and Saudi Arabia illustrate the creative ways AI is transforming high-skill workflows in manufacturing and industrial companies.

When two Stanford students saw a social media post about a student innovation competition, they almost scrolled past, assuming they’d have no chance. They decided there was no downside to trying and applied anyway.

Daniel Kim and Arjun Oberoi were soon pitching their project, an AI-powered device for people with low vision, onstage at the Red Bull Basement’s global finals and went on to win last year’s Imagine Cup — two contests supported by Microsoft. They say the experience gave them skills and contacts they’ll carry throughout their careers, especially at a time when AI is making development the easiest part of creating new technology.

A new group of students from around the world is getting ready to compete in this year’s contests in early June.

Microsoft Research just built and released a large, open dataset that models the U.S. power grid, aiming to give researchers realistic network layouts without relying on proprietary or restricted infrastructure information. The models span 48 states and support detailed, physics-based analysis — like identifying where transmission is constrained, where new demand can be added and how upgrades could improve the system.

Early examples show how the models can reveal practical tradeoffs, like how placing a large datacenter in one location can overload nearby transmission lines while a similar site just 50 miles away can be absorbed without issues. The broader goal is to make it easier to study real-world energy challenges like rising demand and grid congestion while enabling both traditional engineering analysis and newer, AI-driven approaches.

Xbox and Discord are expanding their partnership. Discord Nitro members will now get a starter edition of Xbox Game Pass, giving eligible members a new way to discover and play games, while Discord rewards will be added to Xbox Game Pass subscriptions. Discord has also rolled out updates that make it easier to discover and jump into games included with Game Pass. When users see a friend playing something they want to try, whether in their game stream or Discord game activity, they can click “Play” and select “Xbox Game Pass” to check it out.

OpenAI’s GPT-5.5 Instant is now available in Microsoft 365 Copilot and Copilot Studio. GPT-5.5 Instant improves the quality of everyday work tasks and is designed to deliver more accurate and concise responses. It also features improved performance on image analysis and STEM-related tasks and reduces unnecessary follow-up questions.

As the way we work changes amid the AI era, one industry is ahead of the curve. Software engineering is experiencing a broad shift: AI isn’t just boosting productivity, it’s reorganizing how work gets done. Four human‑agent patterns have emerged as capabilities grew and responsibilities shifted— writing, reviewing, delegating and orchestrating. Those patterns are now spreading across knowledge work in general. But as other industries shift, the key will be matching the work to the right patterns.

It’s World Passkey Day! Yes, that’s evidently a thing, and yes, we were just as surprised as you are. But it’s actually a chance to reflect on how we can reduce our reliance on passwords and other cyberattack-prone methods by adopting passkeys. Unlike traditional sign-on options, passkeys are device‑based cryptographic keys, so there’s no shared secret to steal or reuse — even in a breach. They’re built to be phishing‑resistant and tied to user’s devices, making common attacks like fake logins and credential theft largely ineffective. Now, Microsoft is expanding passkey adoption across the digital ecosystem.

Microsoft released its latest Global AI Diffusion Report today. The global adoption of artificial intelligence continued to rise in the first quarter of 2026. During the quarter, AI usage increased from 16.3% to 17.8% of the world’s working age population. At the top of Microsoft’s National AI Leaderboard, the UAE continued to lead global AI diffusion at 70.1%. The United States finally started to move up the national rankings, though only from 24th to 21st based on a 31.3% usage rate by the working age population. Dive into the full report to see how adoption varies by region and sector.

A new Microsoft-powered workflow analyzes crash images in seconds to propose likely damaged parts — so teams can start repairs sooner. With telemetry streaming every three seconds into Microsoft Fabric, engineers can also spot abnormal behavior early and intervene while cars are still on track. That means faster damage assessment and smarter interventions, for the same goal: safer racing and a better experience for drivers, sponsors and fans.

You know that stressful moment before meeting with your manager — the one where you’re scrolling through emails, flipping through Teams threads and trying to remember just what on earth you’ve been working on this week? Now, Microsoft 365 Copilot can handle the scramble for you and get all your updates in order before your next one-on-one with this all-encompassing prompt.

Copilot Cowork is now available across desktop and mobile devices. Microsoft announced today that workers can use Microsoft 365 on their computer, iOS or Android devices to describe the outcome they want and Copilot Cowork will create a plan, reason across their tools and files and carry the work forward. With the new mobile connections, Cowork can operate across business systems and data, even when users aren’t at their desk.

Microsoft is partnering with government-backed groups in the U.S. and U.K. to test the company’s most advanced AI systems, check their safety measures and better understand potential risks to national security and public safety. The goal is to work with government experts to catch problems early — like cyberattacks or misuse — and build more trust in AI by making sure the systems work as intended and are as safe as possible.

Every business leader knows the world is changing, but far fewer know what to do next. The real work now is to rethink how organizations operate around four human-agent collaboration patterns: author, editor, director and orchestrator. As that shift happens, people will spend less time on step-by-step tasks and more on setting direction, defining standards and evaluating outcomes. Learn more in the Microsoft 2026 Work Trend Index, a report on the state of AI at work.

Microsoft is introducing new security features for Microsoft Agent 365 and expanding Microsoft Defender’s integration with GitHub as AI systems grow more autonomous. The updates are meant to give organizations clearer oversight of what their AI agents are doing while adding real-time protections, linking code to live environments and helping teams spot and address risks more quickly.

Email threats didn’t slow down in early 2026 — they got smarter. New Microsoft research shows attackers leaning heavily into link-based phishing, with QR codes more than doubling in just three months and fake CAPTCHA pages surging in the first quarter of the year. Taken together, the trends offer a clear picture of how email attacks are evolving — and what defenders should prioritize next.

AI is already doing work inside the software companies rely on, quietly handling tasks in the background and changing what the software tools are built for. Instead of people navigating the software, the AI does the work. That means the focus now must shift to making sure systems are set up properly and that humans stay in charge of setting direction and judging the results.

Xbox is rolling out more options to customize the gaming experience. On the console, users can now pin more Groups to Home, customize colors and disable Quick Resume Settings on a per-game basis. There’s also a new play history tab, more resolution options and a network quality indictor. On PC, users can now manually add their favorite games to their Xbox library and reposition where notifications show up. Plus, the Gamepad cursor turns a player’s gamepad into a mouse, giving them a controller driven pointer to interact with apps that aren’t optimized for controller input.

As organizations move toward becoming Frontier Firms, AI is shifting from being a productivity tool to a driver of creative reinvention. Microsoft’s approach centers on intelligence and trust: Microsoft IQ adds context and precision across data and workflows, while Agent 365 provides governance, security and observability across agents — supporting open, heterogeneous environments.

AI isn’t meant to stay in pilots. Accenture is expanding Microsoft 365 Copilot to nearly its full global workforce of 743,000 across Accenture and Avanade, making it part of everyday work. Company data from the first 200,000 employees using Copilot in 2025 shows it helped speed up routine tasks by up to 15x and deliver a 53% improvement in productivity and efficiency.

Azure Local now scales to support deployments of up to thousands of servers within a single sovereign environment, allowing organizations to run much larger workloads locally across large-footprint datacenters, industrial environments and edge locations while maintaining control within their sovereign boundary.

AI’s promise is real. But the challenge to realizing its value isn’t technology — it’s people. AI scientist, entrepreneur, and investor Rana el Kaliouby joins podcast host Molly Wood to explain why adoption breaks down when AI doesn’t fit how people actually work and how to close the gap between experimenting and making AI actually useful.

“It’s not just about like getting the work done,” el Kaliouby says. “It’s actually how do you … dissect an everyday workflow and inject AI in it in a way that’s sustainable, that’s repeatable, that’s trustworthy?” She says it’s easy to try tools, but much harder to build them into your day‑to‑day in a way that people trust and keep using.

For the first time, researchers have mapped agricultural field boundaries across the entire planet — creating a new foundation for tracking food systems, water use and agricultural change. The map is powered by AI models that have been trained to identify field boundaries from satellite imagery and built through a global collaboration between Microsoft, Taylor Geospatial and other partners. Explore the interactive map to see how the world’s fields take shape in unprecedented detail.

Large language models (LLMs) are designed to have conversations. That means they can help not just when a task is clearly laid out, but also when users are still figuring out what they need by talking it over. Now, new research shows that LLMs often do worse when a task is spread across multiple turns instead of just being given all at once. Across many tests, performance dropped sharply — about 39% — mainly because models became more unreliable with subsequent iterations. Researchers found the models tend to guess early, jump to answers too soon and then stick with those wrong assumptions instead of adjusting as new information comes in. It’s definitely something to keep in mind when assigning a task to an LLM.