I was watching with deep sadness the thousands of layoffs that Microsoft has announced in the past days, and the backlash when they removed Claude licenses for their own workers instead of using Copilot, wondering how much things have changed in the past 3 years.
I remember when in 2023 we started using Bing, non-ironically, because it had GPT-4 integrated into the AI search. Its market share was rising, and Microsoft, thanks to its partnership with OpenAI, was really riding the AI wave.
But it all comes down to friction, and that is what today’s article is about. “Obvious 🤓”, you may say, but it doesn’t look that obvious to a 2.8 Trillion USD company. Essentially, even if you have the most powerful model (GPT-4 back in 2023, Fable/Mythos now, and more to come), if the experience with it creates friction, it’s not going to end well. As Microsoft Copilot products are clearly showing us.
I distinctly remember my excitement when we had exclusive preview access to the Copilot tools in our Office (disclaimer, I don’t use Office anymore). Following the wave of initial excitement, I immediately realized I didn’t need them (plus, they were extremely expensive!). It was much easier to just copy the entire context into my company ChatGPT tab than using the slow, clunky, and hallucination-prone small box in my Word or PowerPoint.
the infamous Copilot button (credits: Microsoft)
an example of an annoying interaction with Copilot in PowerPoint (credits: @No-Wave2924)
For most of my personal experience, Copilot products are slow, often they just hallucinate things or don’t understand me, and they lack context. They provide more annoyance than profit.
And while things slowly improved, from what I am seeing around, Microsoft clearly didn’t focus its attention on making the user experience better. They just slammed AIs and Copilots everywhere, from NOTEPAD to a manual button on PCs, creating a lot of confusion. Also, the drama with OpenAI didn’t help keep things calm, but I wouldn’t enter into that territory.
so many Copilot products released… (credits: teybannerman)
It’s the clear example of “let’s add AI here, users will figure it out.” It works for immediate bonuses and increasing stats, but then reality kicks in.
And now better products like Claude Cowork are eating their market share, including usage from Microsoft’s own employees. It’s unbelievable how such a massive company with at least a 1-year head start lost terrain by just not focusing on user experience.
By trying to understand their point, I assume Microsoft’s objective is to sell these licenses in bulk, so they can be compliant and “safe” for the big enterprises.
But surprise surprise, it looks like enterprises also want productivity, and if the “compliant” tool doesn’t help that much, they will go to OTHER enterprise services that are more helpful. Microsoft initially won by inertia, but in such a fast-paced world, things change quickly.
The more powerful model is not the point. Let’s focus on making it work frictionless with your users, gaining trust, and showing that AI can really provide help.