By Dorsey Standish, Mastermind Founder & CEO

I have a confession to make. I ordered three different front racks for my new bike, and none of them fit.

I bought the bike to take my son to school. I needed a way to carry his backpack and our things, so I asked ChatGPT for help finding the right rack. Each time something didn’t fit, I went back to AI, explained the problem, and tried its next recommendation.

It almost got me a fourth time before I finally stopped and measured everything myself.

Looking back, I kept asking the same tool to correct itself without checking its answers against what was right in front of me. I wanted help getting something off my list, and its recommendations sounded convincing.

I think a lot of us are doing some version of this.

When AI makes us more confident than accurate

Good judgment with AI includes recognizing what we don’t know, checking an answer, and seeking someone else’s experience. A convincing response can make those steps feel unnecessary.

In a preliminary study of AI advice and uncertainty, researchers used difficult movie-trivia questions that the AI they tested usually answered incorrectly. In one experiment, participants said “I don’t know” or withheld an answer 44% of the time without AI. With AI available, that fell to 3%. Related experiments found that people using AI were less accurate but more confident in their answers.

The paper is a preprint, not yet peer reviewed. These were deliberately difficult trivia tasks, not workplace decisions, so the results don’t tell us how often AI causes mistakes at work. They do give us a reason to question whether feeling more certain means we understand more.

That misplaced confidence can follow us into meetings and decisions. If we think we already know, why ask someone else?

The high-performing teams I work with regularly name the same commitments to each other to operate at a high level: ask questions, stay curious, speak up. Those habits help us learn from each other and catch mistakes. They require us to admit when we don’t understand something, even when we could quickly generate an “answer.”

Thoughtful uncertainty is becoming a key leadership skill and a strategic advantage in the age of AI. Knowing when to say “I don’t know yet,” what to question, and whose experience to seek helps us make better use of both technology and each other.

Three ways to practice judgment while using AI

I’ve been thinking about how we can keep practicing these skills while using AI. A few ways to cross-train:

• Do some thinking without it. Sketch out a problem on paper, outline a presentation from memory, or talk through a decision with a colleague. Giving an idea time to develop is part of practicing cognitive patience.
• Switch up where AI helps. Write the first draft yourself and ask for feedback. Next time, let AI brainstorm, then develop an idea yourself. Keep practicing different parts of the work so you keep using those skills.
• Check beyond the conversation. Read the original source. Take the measurement. Ask someone with firsthand experience. Another prompt isn’t always the best next step.

Wisdom includes knowing where our understanding ends. We need to stay willing to question, ask for help, and take responsibility for what we do next.

To help your team practice focus, emotional regulation, and connection at work, explore Mastermind’s leadership training and brain health programs.