When a company has to be cutting edge in its field, it'll often make mistakes. It'll place a lot of its chips on an emergent piece of tech in the belief that it will take off, and when it doesn't, it'll quietly cover it up and pretend it never happened.

Microsoft was in such a position with Copilot, and has spent the better part of the last few years trying to make its AI assistant work with its user base. However, as its users revolt and the focus has shifted to tackling Windows 11's biggest pain points, the company has now found itself working to make its previous pride and joy disappear.

Microsoft had an excellent headstart to the AI race...

The cards seemed to fall perfectly in place for the tech giant

After OpenAI introduced LLMs to the greater market, Microsoft spent no time twiddling its thumbs. We saw it perform rapid progress with its AI offerings, starting off as a humble addition to Bing in early 2023, before introducing Copilot onto Windows 11 in September of the same year.

The company whipped itself into a furor over the AI assistant, and honestly, who wouldn't? LLMs were a super exciting piece of tech at the time. We got our first glimpse into an assistant that could code, write, and draw, and people immediately began speculating as to how this new technology could benefit us and our workflows.

Microsoft likely believed that, if it acted too slow, someone else would swoop in and take all of its users. So, it began rolling out Copilot as far as it could, and it did so very quickly. We began to see Copilot pop up in Microsoft 365, Notepad, and Paint. We saw the company show off new features where Copilot could recommend ways to use your PC. We saw the introduction of Copilot+ devices that leverage the AI locally. And when Microsoft ran out of places to cram Copilot on Windows, it began adding it to things like smart TVs.

I don't think Microsoft ever felt that it was doing something wrong. From what we heard from insider sources and posts on X from the higher-ups, Microsoft was enamoured with the tech. It saw a future with Copilot that would revolutionise how we use our computers, and it likely believed all the naysayers would eventually die down when they realised how good Copilot would be.

...but its focus on quantity ended up backfiring

Copilot fell behind

There was just one problem with Microsoft's approach. The company was determined to plant the AI flag on as many frontiers as possible. If it could grab the hearts of its users before anyone else could stake a claim, it would be very hard for competitors to shift people off Copilot. People would see a new service appear, say "Oh, no thanks, I'm already all set up with Copilot," and continue using Microsoft's product.

However, there was a huge flaw in Microsoft's plan, one that people didn't really see coming. In the early days of LLMs, we saw them as general assistants that could do anything. Sure, they weren't great at doing everything, but they could do everything. As the AI race matured, we began to see LLMs less as a generalist and more like a specialist.

Eventually, people had their favorite LLMs for different things. Some people like using Claude for coding projects, and Gemini for lifestyle questions. People may also gravitate toward Gemini if they want to tap into Nano Banana 2 for video generation, while others are giving Seedance 2 a try. You no longer used one LLM for all of your tasks; you used different LLMs depending on how good you felt it was at each one.

This left Copilot in a weird spot. Microsoft had been pushing it to become a jack of all trades; it could write emails, arrange your documents, search your PC, draw pictures, and generate code. But that just meant it would be destined to take, at best, second place for performing certain tasks, and in a world where you can use a top-rated LLM for free, second place just doesn't cut it. Copilot was being left behind.

The company had a chance to keep up with the competition, but yet another problem arose. Much like how OpenAI's ChatGPT introduced LLM to the masses, OpenClaw brought agentic tools into the spotlight. Now, having a chatbot is so passe; AI companies now need an agentic service that allows the assistant to do the job for you. No longer do you need to say "Can you please write code that does this for my game?". Now you pop open Claude Code, tell it to code an entire game for you, have it automate the testing and debugging process, and have a functional title within the hour without lifting a finger.

The agentic era of AI had begun, and all Microsoft had to show for it was a simple LLM system that wasn't winning any trophies. So, the company pivoted.

Now the company is trying to fix its 'Microslop' title

Windows K2 is putting attention back to the core

Due to Microsoft's push to get Copilot onto everything, it earned a degoratory title from its user base around the start of 2026: "Microslop." It summed up Windows user's distaste with Microsoft's focus on using AI to generate everything and signaled that users would much prefer a functional, working operating system over a Copilot button in Notepad.

So, the company listened. It launched what it internally calls "Windows K2," which sounds like a new version of the operating system, but is actually a reshuffling of engineers away from Copilot and onto the major pain points people have had with Windows 11. Given how Microsoft was full steam ahead on Copilot since 2023, the company has three years of grievances to address.

Part of the Windows K2 focus is winding back the very same features it spent a lot of time pushing. We've already seen Microsoft scale back the Notepad and Paint implementations and let people move the floating Copilot button in Microsoft 365 to the ribbon to get it out of people's way. The company is waking up from a three-year-long Copilot-driven stupor, and now it has to pick up the pieces.

Microsoft now needs to pick up the pieces

After three years of going all-in on Copilot, the company now has to spend time making its implementations go away. And while the company is likely unhappy with undoing all its hard work, its users will be grateful in the long run.