AI News Today
← All episodes
Episode 1 · March 15, 2026 · 18:42

Microsoft Copilot Cowork: New Autonomous AI Agent

Microsoft & Anthropic Partner: The New Era of Autonomous AI


Microsoft has shocked the tech industry by partnering with Anthropic to launch Copilot Co-work, an autonomous AI agent built for Microsoft 365. This shift signals a massive move toward model diversity and agents that can execute complex workflows across your entire enterprise data.


00:00 - The Microsoft & Anthropic Shock

02:29 - How Copilot Co-work Actually Works

03:58 - Real-World Use Cases & Automation

05:00 - Microsoft's New AI Strategy

06:56 - Pricing & The ROI of AI Agents

08:17 - The Power of Enterprise Data Context

10:50 - Google & Samsung's Agent Revolution

13:22 - The Future: 1.3 Billion AI Agents

Full transcript

Microsoft Co-Pilot Co-Work, new autonomous AI agents. So Microsoft just announced a brand new partnership with Anthropic, the company behind Cloud AI, and together they have built something called Co-Pilot Co-Work, and it's about to change the way hundreds of millions of people do their jobs inside Microsoft 365. That's the story today, and it's a big one. For years, the partnership between Microsoft and OpenAI seemed like an unbreakable alliance.

With billions of dollars invested, Microsoft's AI strategy was essentially an OpenAI-only story, but things can change quickly, and Microsoft has now partnered with Anthropic, Claude's parent company, to launch Co-Pilot Co-Work. Let that sink in for a second. Microsoft, the company that put $13 billion into OpenAI, just went to OpenAI's biggest rival and said, we need what you've built. Why?

Because what Anthropic built with Claude Co-Work is something that Microsoft couldn't get from OpenAI, and Microsoft was watching their own stock fall off a cliff because of it. Microsoft shares have fallen more than 14% since Anthropic debuted its Claude Co-Work product in mid-January. At the heart of Microsoft's announcement is Co-Pilot Co-Work, a new feature built in close collaboration with Anthropic. 14, that's billions of dollars of market value gone because investors looked at what Anthropic was doing with Claude Co-Work, and started asking a very serious question.

If AI can now do the work that enterprise software can do before project management, writing, data analysis, workflow automation, why do companies need to pay for all those expensive software subscriptions? Anthropic's initial Claude Co-Work applications released in the first two months of 2026 helped trigger a $285 billion sell-off in enterprise software stocks as investors repriced companies, whose core functionality, project management, writing, data analysis, workflow automation, overlapped with what Anthropic's AI could do. $285 billion, not from one company, from the entire enterprise software sector in one wave of selling, because investors could suddenly see where this was heading. And Microsoft saw it too.

So instead of fighting it, they partnered with Anthropic and brought the same technology into the Microsoft 365 ecosystem that already sits on the desktops of over a billion workers worldwide. That's a move, that's a plan, it's a smart one. So let's break down what Co-Pilot Co-Work is, why it matters, and what it means for anyone using Microsoft 365 for their job, which is, by the way, most of the working world. Co-Pilot is built to help a co-pilot take action, right?

So you've got Co-Pilot Co-Work, and you've got Co-Pilot, and this basically turns it into an autonomous AI agent, not just a chat. You describe the outcome you want, and Co-Work automatically grounds work in your emails, meetings, messages, files, and data. When you hand off a task to Co-Work, it turns your request into a plan. The plan continues in the background with clear checkpoints, so you can confirm progress, make changes, or pause execution at any time.

Now let me explain that in plain English. Old Co-Pilot, you ask it something, it gives you an answer, you go off and do the thing. The new Co-Pilot Co-Work, you tell it what you need to get done, it looks at your whole Microsoft 365 account, your emails, your calendar, your team chats, your Excel files, and your Word docs. It builds a plan, it starts executing that plan in the background, it checks in with you at key moments, and it gets the work done.

So you're not the one doing it anymore, the AI is doing it, and you're supervising. That's a fundamentally different relationship between a human and a computer. It's not a smarter search engine, it's a digital co-worker. Co-Pilot Co-Work can handle tasks such as resolving calendar conflicts, preparing meeting briefs, compiling research memos with citations from web and Workspace sources, and building product launch plans with competitive analysis and pitch decks.

Think about how much time those things take in a normal week. Calendar management alone, how many hours a week do people spend scheduling, rescheduling, figuring out what needs to be done in what meeting, declining things that are a waste of time, and blocking focus time. Co-Work just does that. You tell it what you're trying to prioritize this week, it looks at your whole calendar, it proposes changes, you approve of them, done.

Meeting prep. So let's say you've got a big client meeting in two hours. Old way, you would scramble, you pull emails, you check notes, you try to remember what was discussed last time, you maybe open a few docs if you have time. The new way is that you tell Co-Work to prepare for the meeting.

It reads your email history of that client, it reads your team messages, it pulls the relevant files, it writes your briefing document, and it's ready before you even think to ask. During a demonstration, an agent powered by Anthropx reasoning analyzed a quarter's worth of meetings, cross-referenced them with email priorities, and declined unnecessary invites all without human supervision. A quarter's worth of meetings. That's three months of calendar data.

Analyzed, cross-referenced with email priorities, and acted on. In the time it takes you to make a cup of coffee. Now, I want to give you the full pitch here, because there's something really interesting happening in this story that goes beyond just a product. Microsoft's decision to embrace Anthropx, whilst remaining OpenAI's largest investor, is a key move.

The AI industry is currently fractured by growing feuds between startups and tensions with national security agencies. The Redmond giant is making sure its customers aren't locked into one vendor. So Microsoft CEO for AI at Work, Jared Spataro, told Fortune, and I love this quote, every 60 days, there's a new king in the hill, right? Every 60 days, a new king of the hill.

That's the most honest description of the AI industry right now that I've ever heard from a senior executive at one of the biggest companies in the world. They're not betting on one model. They are not betting on one company. They are building a platform that can run whatever the best model is whenever it changes.

Microsoft 365 Copilot is model diverse by design. Rather than betting on a single model, they build a system that makes every model useful at work, customers get the choice, performance, flexibility in an open, hetero, generous environment. And that's a massive strategic shift where Microsoft was two years ago, right? So they've really moved on.

Two years ago, it was all open AI all the time. Now they're running Claude from Anthropic, open source models from DeepSeek and Quen, and they just announced Fireworks AI integration for even more model options. They are essentially building the operating system for enterprise AI. Not the AI itself, the layer that sits on top of all of the different AI models that makes them useful in the context of the actual work.

And the numbers are showing it's working. Microsoft recently delivered its strongest quarter yet with Copilot, with paid seats growing more than 160% year over year and daily active usage up 10 times. The number of customers deploying Copilot at significant scale, more than 35 seats, tripled year over year, and 90% of the Fortune 500 now use Copilot. 160% seat growth, daily usage up 10 times, and 90% of the Fortune 500.

That's not product people are testing anymore. That's a product that is embedded into how the biggest companies in the world operate. And now Cowork is being layered on top of all of that. Now, let me talk about the pricing because this is where it really gets interesting for anyone making business decisions.

Copilot Cowork requires a Microsoft 365 Copilot license at $30 per user per month on top of existing enterprise subscriptions. For organizations seeking a comprehensive package, Microsoft introduces Microsoft 365 E7, which is priced at $99 per user per month. And that's available from May the 1st. The E7 tier bundles Copilot, the Microsoft Entry Suite, and Agent 365, $30 per user per month for Cowork access, or $99 per user per month for the full bundle, including security tools and Agent 365.

Now, some of you are going to say that's expensive. And for a small business, yes, it can be. But here's a math I want you to think about. If Cowork saves one person on your team just four hours a week, which is extremely conservative based on everything we're seeing, that's roughly 16 hours a month.

Even a modest salary, that's easily worth more than $30. So you're getting leverage on the salary you're already paying. And for bigger teams, the math gets better fast, right? So 10 people each saving four hours a week is 40 hours a month of recovered productivity.

That's essentially a full-time employee's work output for $300 a month. That's the economic argument. And it's why enterprises are paying for this. Now, here's the thing I need you to understand about why this Microsoft Anthropic Partnership is so significant.

This is something that a lot of the tech coverage is really missing. So Copilot Cowork operates in the cloud, inside Microsoft's 365 infrastructure, and draws on something called Cloud Cowork that simply cannot access, right? The full graph of a user's enterprise work data. That means Outlook email threads, team conversations, calendar history, SharePoint files, Excel workbooks, and the relationships between them.

When Cloud Cowork runs on your desktop, it can see your local files, your downloads folder, your desktop, whatever you point at it. When Copilot Cowork runs in Microsoft 365, it can see everything, right? So every email you sent and received, every team's message, every calendar event, every shared document, every relationship between all those things. And that's a different level of context.

That's the difference between a new assistant who just started and read your files versus an assistant who's been sitting in every meeting, reading every email, and watching every decision your company has made for the last three years. That depth of context is what makes enterprise AI actually useful for complex work. Not just finding information, understanding the full picture. And that's why I keep saying to anyone who will listen, the people who learn how to work with these tools right now, who understand how to delegate AI effectively, who can build workflows around these gigantic systems, those people are going to be extraordinarily valuable in the next three to five years, because here's the shift that's happening in real time.

We're moving away from chatbots and simply talking towards agents that do, right? So chatbots just talk and agents are autonomous. They can do stuff. The first wave of AI was information.

You asked it, it answered. The second wave, the one we're in right now, with Copilot, Cowork, and ClawCowork, and OpenClaw, and Google Gemini agents, and all the rest, is implementation. You delegate acts. That changes everything about how work gets organized.

It changes what skills matter. It changes what you actually need people for. And before I go further, I want to take a moment to tell you about something that I think is genuinely important if you want to stay ahead of this stuff. If you're watching this and thinking, okay, I understand the big picture, but I don't know how to actually use these tools in my own business.

I don't know how to delegate effectively to AI. I don't know how to set up workflows to save me time instead of create more work. That's exactly why I built the AI Profit Boardroom. It's a community where I share the real stuff, not just the headlines, natural workflows, the tools that work, the ways that real businesses like myself are using AI right now to save hours every single week and build competitive advantages before the competitors even figure out what's happening.

Feel free to join us, link in the comments description, or go to the AIProfitBoardroom.com. Now let's keep going because the CoPilot CoWorks story is the biggest piece of news this week. It's not the only piece. And when you put it all together, a very clear picture starts to emerge.

So Google just dropped their March Pixel update this week and buried inside it is something that tells you the exact same story. Gemini Automation is expanding on the Pixel 10 series. It can now book rides, order food, or purchase groceries by launching apps in a secure window and tapping, scrolling, and typing on your behalf. You can watch the process live or let it run in the background with notifications.

Final confirmation is always required before completing a transaction. Your phone is now doing the same thing with CoPilot CoWork, and it's doing that inside of Microsoft 365. It's not just answering your questions, it's taking actions on your behalf. It's going into other apps, navigating them, completing tasks, and reporting back.

One year ago, that was like a demo on a stage somewhere that nobody believed we could actually ship. Now it's just shipped to millions of people as a free software upgrade on their existing phones. Samsung launched Galaxy S26 with agentic AI that anticipates what you need before you ask. The phone includes Now Nudge for automatic suggestions, Photo Assist for editing photos by describing changes, and multiple AI agents, including Gemini and Perplexity working on your behalf.

Samsung and Google are putting multiple AI agents on the same phone, working together on your behalf. That's not three apps anymore, that's three employees that never sleep, never ask for a raise, never call in sick, and then there's a browser story on top of all of this. So Microsoft Edge now has a full-powered CoPilot mode, and it's fully AI-powered, and it lets you compare tabs, summarize pages, draft content, all inside the same browser. Now from May 2026, Edge will automatically be open on the CoPilot side pane when users click links from Outlook emails, surface contextual insights, and gives you actionable suggestions based on email destination content.

So when you look at all of this, right, and we paint the picture of what a normal workday starts to look like, with all of this running together, it's pretty courtesy. So for example, you wake up, your phone has already looked at your calendar, checked your email, and prepared a briefing on what matters today. You get to your desk, you open Outlook, you click a link in an email. Before the page loads, CoPilot's already summarized it and told you what to do.

You tell CoWork to prep for your 11 a.m. meeting. It goes into your email history, your team's chats, your files, and builds your briefing doc. You approve a few calendar changes CoWork suggests.

And then you tell your phone to book a restaurant for the client dinner tonight. Gemini opens the app, finds options, books a reservation, and confirms it all before noon without you manually doing any of it. That's the direction, that's what's being built right now across Microsoft, Google, Apple, Samsung, and every other major platform simultaneously. Now, let me give you the economic context that makes all of this make sense.

The AI agent-agent market, right, reached $7.6 billion in 2025, and is expected and projected to grow 49.6% annually through 2033. IDC predicts 1.3 billion agents in circulation by 2028, and the Fortune 500 are already using Microsoft agents. 1.3 billion AI agents by 2028, that's only two years from now. And that's not a niche technology trend, that's infrastructure, that's the plumbing of how work gets done.

Morgan Stanley predicts that transformative AI will become a powerful deflationary force as AI tools replicant human work at a fraction of the cost. The bank says executives are already executing large-scale workforce reductions because of AI efficiencies. OpenAI CEO, Sam Altman, has envisioned entirely new companies built by just one to five people that can out-compete large incumbents. One to five people out-competing large incumbents, right?

That's not a future prediction, that's already happening right now. Now, Cursor, you've got, you know, running with $300 million a year, you've got Mid-Journey with 40 people and making $200 million a year, right? A lot of these companies are growing because they're AI-first and they can just do with one person what it normally took 10 or 50 or 100 people. And this is what the co-pilot co-work story is really about at its core.

It's not about Microsoft catching up to Anthropic, it's not about which company has the best model, it's about the fact that these tools now exist to let small teams do the work of large ones, and large teams do the work they never had time for before. The question isn't whether it's happening, it's already happening, the question is whether you're going to be on the side of people learning how to use it or the people hoping it goes away. Because if I'm really direct with you about what I think you should do based on everything you've shared, I've shared today, let me talk you through it, okay? So if you're in a job that involves a lot of information management, you know, emails, reports, documents, start learning how to delegate to AI right now.

Not eventually this week, pick one thing you do repeatedly. That could be research, meeting prep, email drafting, calendar management, whatever it is, and find out if Copilot, Cowork, or a similar tool can do it. Spend a week figuring out how to make it work for your specific situation, automate it, and then move on to the next thing. If you run a team or manage a business, the strategic question you need to be asking yourself right now is not how do I use AI, the question is how do I redesign my workflows so that AI does the commodity work and my people do the high judgment work.

That's the organizational challenge over the next three years, and the companies that figure it out first will pull so far ahead that the ones who figure it out last may never catch up. And if you're a creator, a freelancer, a marketer, or someone who builds content or runs campaigns or manages clients, the tools being built right now are designed specifically for you, right? The research, the drafts, the analysis, the scheduling, the admin, all of it can increasingly be delegated to AI. Your competitive advantage is not going to be how fast you can write an email or how quickly you can pull data from a sheet, it's going to be your judgment, your relationships, your taste, and your ability to direct AI to produce the right outputs.

OpenAI has surpassed $25 billion in annualized revenue and is reportedly taking early steps towards a public listing. Rival Anthropic is approaching $19 billion in annualized revenue. The figures signal that the market for advanced AI models has rapidly become one of the fastest growing sectors in the technology industry. $25 billion, $19 billion for AI companies that barely existed a few years ago.

The market has spoken, the customers have spoken, and the growth is not slowing down. The people who are building skills and workflows and competitive advantages in this space right now, they're going to have an enormous head start over everyone else who waits. I talk about this exactly inside the AI Profit Boardroom every day. Real workflows, real tools, real examples.

So if you want to join us, link in the comments description or come to the AIProfitBoardroom.com and just see what it's about, see what people are building and connect with us. Because here's what I keep coming back to as I watch all of this unfurl. The Microsoft and Anthropic partnership is not just a business deal, it's a signal. It's the biggest company for software in the world looking at what Anthropic built and saying, we need that.

We can't build our own way this fast, right? And we need to partner with someone who figured it out already. That's how fast this is moving. Even Microsoft with all its resources, its engineers, its billions of dollars in AI and invested into OpenAI, even Microsoft have to go outside and find something better for this problem.

Now, imagine you're a small business owner or a freelancer or someone trying to build something. You don't have Microsoft's resources, but you have access to the same tools Microsoft just paid a fortune to integrate. They're available to you right now. This is the moment we are in.

And whether it's co-pilot co-work or whether it's Claude co-work or Gemini agents ordering groceries on your phone or Claude handling your research and writing, all of the above running together across every platform you use, the direction is the same everywhere you look. AI is not something you go looking for anymore. It's arriving into your browser, your phone, your email, your calendar, your spreadsheets, your documents. It's being embedded by default into every tool that every working person uses every single day.

You can push back on that. You can debate whether it's good or bad. That's a legitimate conversation, but the most important thing you can do right now, in my view, is make sure that when all of these tools land in your workflow, they're landing right now, by the way, whether you invite them or not, you have to become the person who knows how to use them well. Not the person who ignores them, not the person who's intimidated by them, but the person who looked at this moment early, took it seriously, put it in the work to understand it, and came out the other side with a genuine competitive advantage.

That's what this is all about. That's why I cover it every single day. Stay ahead of it. Keep learning.

I'll see you in the next one. Thanks. Bye-bye.

More episodes

Browse all episodes →