Who is doing the extra work?

I might be giving my own AI tool too much credit.

I was working on a newsletter draft and it had pulled apart ideas that belonged together. I could see the connection, so I explained what needed to change.

For me, that was a normal edit. But I already knew what the argument was supposed to be. Someone using the tool to help figure out their argument might not know what to ask it to change.

I can imagine doing the same thing with a tool that puts together a launch plan. If I know the subject, I can question a suggestion that doesn't fit. The person buying it may be looking for help making that call in the first place.

So I'd want to sit with someone while they try it on a job they actually need done. When the answer comes back, let them explain what they'd do with it and why. I'd be looking for whether that gets them closer to finishing the job, not just whether the answer sounds good.

And when I need to step in, I'd pay attention to what I'm adding. Sometimes it's just showing them where to click. Giving them information they didn't have is different, and helping them decide which advice to trust is different again.

I wouldn't redesign a product around every hesitation. But if the same missing help keeps people from doing what I said the tool would do, I'd test a way to provide it or make a smaller promise.

Some tools are for people who already know the subject. That's fine. I just need to find out whether those are the people I'm building for, instead of assuming they know what I know.