Some Clients are banning AI. And They’re banning the wrong thing.

Last week I read a client brief that asked for ideas no AI tool could come up with. Using ChatGPT or similar tools for the work was not allowed, and leaning on AI too much was treated as a kind of dishonesty. And the client obviously wanted to find people who can do what AI can’t.

I understand where this comes from. Anyone who hires has seen the same polished, empty proposals arrive in a pile, and a lot of generic AI output looks like more of the same. Banning the tool feels like a quick way to filter for real thinking.

But I kept reading, and the brief got more interesting.


The brief asked for two things that can’t both be true

The product had to feel smooth, obvious and effortless, with no guesswork for the user. In the same document the team was told not to use conventional layouts, components or UX patterns just because they are industry standards, and to add several custom components the user has never seen before.

Here is the problem. A product feels obvious mostly because the user has met that pattern somewhere else. Jakob Nielsen put it as a law: users spend most of their time on other products, so they expect yours to work the same way. A search field at the top, a cart in the corner, a back arrow on the left: nobody has to think about these, and that is the whole point. Every new pattern you invent is something the user has to learn.

So the real job in that brief was a decision: where is novelty worth the learning cost, and where should the product stay boring on purpose?

That decision is the hard part of product work. A tool ban doesn’t test whether someone can make it.


Most clients now ask for the opposite

Since 31 October I’ve been writing down what each job post I review says about AI. It is a small sample, about 35 posts in a week, mostly product manager and product owner roles. Fifteen of them mention AI use at all:

  • 11 expect or require it. Several want to see something you have built or automated yourself in the last few months, and one said AI fluency would count more than anything else in the decision.
  • 1 limits it to the application: please don’t send AI-generated cover letters with made-up information.
  • 1 bans it from the work, the brief above.
  • 2 sit in the middle. One of them put it in the way I liked most: they wanted someone who uses AI to work faster but keeps the reasoning and judgment their own.

That last line is close to how I work, and I think it describes the line clients actually care about.


What AI does in my work, and what it doesn’t

I use AI every day. In my job search it reads a vacancy, checks it against my profile and my rules, and drafts a fit score and a first version of a cover letter. That saves me hours.

It also gets things wrong. Last week it read “Georgia” in a job filter as the US state, and for a while I was looking at remote jobs in Atlanta instead of jobs open to someone in Tbilisi. Another time a CV draft said I had six years of product experience in one place and eight in another. I caught both because I read what it gives me.

The decisions stay with me. Which roles I apply to, which rate I accept, which claim I won’t make because I can’t back it up in an interview. I make those calls and I live with the results. AI makes the work faster, and I’m the one who signs it.

That’s what doing what AI can’t means to me. Originality is part of it, but the bigger part is judgment: choosing the problem, cutting the ideas that don’t serve it, and owning the trade-off when it goes wrong.

A model can produce twenty concepts in a minute. It has no stake in which one ships.


A brief that would test the right thing

If I were hiring for “AI-proof thinking”, I would skip the ban and ask for evidence of judgment instead:

  • Show one idea you rejected and why.
  • Name one trade-off you made and what it cost the user.
  • Tell us where you used AI and what you checked yourself.

The last question does more than a ban. People who use AI carelessly struggle to answer it, and people who use it well answer it in one sentence. Banned or not, most people will use the tools anyway, so asking for honesty about it gives the client more information than asking them to pretend.

This is an early observation from one week of reading job posts, and I’ll keep counting.

If you hire product people, I’d like to know: what do you ask candidates about AI?



Related reading

More on the same topic.