Are we overbuilding AI features that users don’t actually need?

I’ve been thinking about this a lot lately… Every product roadmap today seems to have “AI” somewhere in it — but how much of it is actually solving a real user problem? In one of the recent features I worked on, we added an AI-powered recommendation layer. Technically, it was solid. The model worked well. But adoption was… underwhelming. What we realized later: Users didn’t want “smart suggestions” — they wanted faster control and clarity. It made me question something: 👉 Are we sometimes building AI features because we can, not because users truly need them? A few things I’m curious about: How are you deciding where AI actually adds value vs just hype? Have you ever removed or rolled back an AI feature? What’s one AI feature that genuinely improved user experience in your product? Would love to hear real experiences — wins and failures.