Why 'We Added AI' Always Fails

These days, every app is adding "AI features."

Let me be honest: almost none of them get used.

Why they don't get used

The answer is simple. "Adding AI" became the goal.

The right order is: first, find where users are struggling. Then think about how to solve it. If AI is the best solution, use it. That's it.

But most products go backwards. They decide "let's add AI" first, then force-fit it somewhere.

What successful AI features have in common

Used AI features share one thing: you don't feel like you're using AI.

Think about email reply suggestions, or traffic predictions in map apps. These use AI, but users don't think about it. Things just got a bit easier.

That's the right way. AI shouldn't be visible. It should just be there, making life a little simpler.

The right question to ask

Not "what can AI do?" — but "where do our users feel the most friction right now?"

If AI can remove that friction, use it. If not, don't.

When nanakazoku builds AI features, we don't write a single line of code for the first two weeks. We only listen to users. That's the only way to succeed.