The AI Confidence Trap

Last month, I asked Claude to thought partner on some strategic recommendations I’d written for a client.

What came back was better than what I’d written.

I sat there for a moment, staring at the screen, thinking, am I even adding anything here?

So I decided to run an experiment.

For one month, how much of my work could I do without AI?

If you are anything like me, you have used AI heavily over the past year - to write that email that needed to land the message, or to draft a project proposal.

So when I started cutting it out, it felt unnatural. Why was I making my brain work this hard? Wouldn’t AI do it better anyway?

But something interesting happened after a few days.

I started getting clearer on what I actually wanted to say, what was at the core of my message. And I felt way more confident about the things I wrote, because I had fully thought them through.

Two weeks in, I was on a call with my client’s MD. They asked me, completely out of the blue, for a quick update on where the project was heading.

Normally I’d have been mentally buying myself time to write them a concise update later (and maybe used AI to refine it). Instead, my brain was switched on, and I gave them a clear and confident answer on the spot. They nodded, and gave me the green light for the next phase of the project.

That moment made me curious about what AI was actually doing to my confidence.

I went and did some research. What I found was fascinating.

AI affects your confidence in two opposite ways. Which one shows up depends on how well you already know the topic.

When you don’t know much about a topic

When you ask AI a question, it explains things in a clean, logical way. You can ask follow-up questions and the answers keep making sense.

So you start to feel like you’ve got this, like you understand the topic.

For example, I once asked AI how to lead a culture change with a client. It gave me a detailed plan. By the time I’d read it twice, I felt like a culture change expert.

But here is what I missed. AI didn’t tell me about the messy human pitfalls, that an experienced people leader would know about.

So my actual competence was still quite low. But I didn’t know what I didn’t know - and I felt more confident than I should have been.

That is exactly what the Dunning-Kruger effect describes. People who know less about a subject often feel more confident than those who know more.

AI amplifies this effect, it makes you feel like you know even more than you do.
 

When you do know a topic well

Let’s flip it.

Say you’ve written a stakeholder update for the Board. You’ve done these updates a hundred times. Now you decide to run it through AI just to check.

What comes back is sharper than what you wrote.

You sit there thinking, well, actually, AI does this better than me.

Research published by the American Psychological Association shows that when we hand off a lot of our thinking to AI, we stop trusting our own reasoning. We stop feeling like our ideas belong to us. (Link at the bottom.)

For us women in male dominated environments, this hits even harder. AI adds another layer to the confidence gap we might already feel. Now there is a machine on your laptop doing your job in a smoother, slicker way than you can.

This is how AI also amplifies imposter syndrome.
 

Why this matters for your career
 

Neither effect is good for you.

Dunning-Kruger means you walk into a meeting with a half-baked answer, sound very confident about something you don’t really understand, and look foolish in front of the executives.

Imposter means you don’t apply for the next level up. Because if AI can already do your job, how could you possibly be ready for a bigger one?

I am not saying you should cut AI out. What I did for a month is not the rule I want you to follow. AI is genuinely useful for research, writing, drafting, and a hundred other things. I have gone back to using it.

What changed for me is how I use it.

The important point: keep your own thinking switched on while you use AI.
 

The 30-second prediction
 

Here is one easy thing you can do, starting tomorrow.

When you ask AI a question, it often takes 20-30 seconds to think and give you the answer. Most people just sit there and wait. Instead, use that time.

Ask yourself:

• What kind of answer am I expecting?

• What 3 points should be in there?

• What does a good response look like?

For example, say you are about to run your stakeholder update through AI. While it loads, predict it. Actually think: What would you say is the conclusion? What main points would you select?

When the answer comes back, you have something real to compare it against. You are not just sitting there absorbing whatever it gives you.

This matters a lot. Critical thinking disappears the moment you stop being active and start being a consumer of answers. The 30-second prediction keeps your brain in the room.

Two more ways to keep your brain engaged
 

If you want to take this further, here are two more techniques you can layer on top.

1. Find one thing that could be better. Whatever AI gives you, look for an error, a missing angle, or something you would add. This forces your brain to engage rather than just accept.

2. Imagine solving it without AI. Ask yourself, if I couldn’t use this AI suggestion, how would I approach this from scratch? You don’t have to ignore the AI version. You just give yourself another anchor point to think from.
   
   

The bottom line

AI can make you feel like an expert in things you barely understand. And it can make you doubt yourself in things you have spent your whole career mastering.

The thing that protects you from both: Your own thinking. Don’t let AI do your thinking for you.

P.S. The research I mentioned is linked here.