Singin’ the Agentic Windows blues

Once upon a time, when you ran Windows on your desktop, it was your desktop. Oh, the IT department might have called the shots on how much you could do with it, but you could write what you needed to, and it was all kept nicely on your PC or your choice of network drive. 

Those days are long gone. 

First, Microsoft started replacing standalone applications with cloud-based Software-as-a-Service (SaaS) programs such as Microsoft 365. Today, you have little choice but to run Microsoft 365. In addition, unless you turn it off, every file you make or save ends up in OneDrive.  Microsoft has also been pushing companies to say goodbye to Windows on the desktop entirely and move to running Windows in the cloud. That move has been less successful. 

Now, there’s a new twist. Microsoft’s been hinting at it for a while, but on Nov. 10, Pavan Davuluri, Microsoft’s president of Windows, tweeted: “Windows is evolving into an agentic OS, connecting devices, cloud, and AI to unlock intelligent productivity and secure work anywhere.”

What does that even mean? 

It means — as near as I can decipher — that with Microsoft’s Agent Workspace and Copilot Actions notes in this latest and greatest version of Windows 11 (OK, that may be an oxymoron, but bear with me), you’ll run AI agents in isolated, secure workspaces. These agents will have their own user accounts, Agent ID, which are separate from the primary user (you). Mind you, to work, these agents must have access to your account’s permissions via the Windows On-Device Registry (ODR) to manage your files, automate routine tasks, adjust settings, and work with your system tools. 

These tools, Microsoft argues, will have excellent security, privacy, and transparency features. Each agent’s actions will be logged and easily auditable. Agentic Windows 11 will include features such as Model Context Protocol (MCP) and agent connectors for apps like File Explorer and System Settings.

Microsoft also likes to talk about how it can do a lot of its work using on-device AI processing. (If you have an AI-chip-equipped PC, of course.) However, to really get the most of AI, you’ll need access to cloud-based large language models (LLMs) 

If you drink Microsoft’s Kool-Aid, you’ll see Agentic Windows as the next frontier in desktop computing. The company frames it as a helpmate that will safely automate your repetitive or complex tasks, and lay the groundwork for new, exciting applications by both individuals and enterprises.

Yeah, right.

First, for all that 2025 has been the year of AI Agent hype, as Marina Danilevsky, an IBM Senior Research Scientist, noted: “We haven’t even yet figured out ROI (return on investment) on LLM technology.” Besides, she wrote, “[Agents] tend to be very ineffective because humans are very bad communicators. We still can’t get chat agents to interpret what you want correctly all the time.”  

Mind you, IBM has its own dog in this fight, Watson AIOps, so it wants agents to take off. It’s just more realistic about them. 

AI agents are still largely hype. As PricewaterhouseCoopers (PwC) pointed out in a recent study, “Reports of full [agentic] adoption often reflect excitement about what agentic capabilities could enable — not evidence of widespread transformation.” 

Ya think!?

Mind you, three-quarters of these same executives agreed or strongly agreed that “AI agents will reshape the workplace more than the internet did.” Oh please, I was there when the internet changed everything. You have no clue what you’re talking about. And, then, as now, there was a hype bubble (that ended in the dotcom crash). The NASDAQ then collapsed by 78% and took 15 years to recover.  

Sure, some AI Agents can do useful work, but is that any reason to embed them in Windows? Or any other operating system? If AI-enabled web browsers, such as ChatGPT Atlas and Perplexity Comet, are too unsafe to be used, why would you think it safer to have them even deeper in your computer??

Even Microsoft admits, “AI models still face functional limitations in terms of how they behave and occasionally may hallucinate and produce unexpected outputs. Additionally, agentic AI applications introduce novel security risks, such as cross-prompt injection (XPIA), where malicious content embedded in UI elements or documents can override agent instructions, leading to unintended actions like data exfiltration or malware installation.” 

This sounds like a barrel of laughs to me.

Let’s get real. This is another Microsoft attempt to monetize AI. When you’ve invested something like $80 billion this year alone in AI, you want to see a return on your investment. I get that. What I don’t get, despite all the CEOs suffering from AI agent FOMO, is why anyone else wants it.

I mean, when I want to use AI — and I do use it — I go through Chrome to Perplexity. There’s no fuss, no muss, and a minimum of security worries.  Indeed, as the top post on X responding to Davuluri’s note said, “Stop this nonsense. No one wants this.“

Another person added to this thread, “Nobody wanted ads in their start menu either.  Or nonstop telemetry.  Or disabled local login.” No one did, and yet here we are. 

So, what can you do? I assure you, Microsoft will not be backing down on this. They see it as a cash cow, since to make the most from it, you’ll need to subscribe to their AI services. 

One last tweet provides the answer: Windows is “evolving into a product that’s driving people to Mac and Linux.“

I’ve been telling you to switch to Linux for decades now, and I’ve even had nice things to say about macOS at times. Really!

Seriously, though, most users have been locked into the Microsoft tech ecosystem for ages now. But do you really want built-in AI security holes moving forward? To have an AI Big Brother watching your every move? 

If you want any kind of control over your desktop, if your company wants control over your desktops rather than Microsoft, it’s now time to start migrating away from Windows to Macs or Linux.