Like many executives I know, AI is very good at sounding sure. That does not mean that either actually is sure. That is why my favorite question to ask AI (and executives) is one of the simplest: Are you sure?
The comparison is not entirely a criticism. Executives are expected to create clarity in conditions that are rarely clear. They are rewarded for decisiveness, asked to make predictions with incomplete information and expected to project confidence so that other people can move. AI has been designed to offer much the same experience: a direct, coherent answer without making us sit through all the uncertainty behind it.
The problem is that fluency and certainty are easy to confuse with truth.
Ask AI whether it is sure and it is striking how often a confident answer gives way to doubt. The response is frequently some version of, “Actually, that’s not true. Good catch.” The reversal can feel reassuring, as though the system has returned to the evidence and corrected itself. Sometimes that is exactly what happened. But sometimes the first answer was right.
Researchers have found that when an AI model is challenged, it may abandon a correct answer for an incorrect one. It is not necessarily discovering that the user knows something it does not. It may simply be responding to a new prompt: the user appears unconvinced. This tendency is known as sycophancy (https://arxiv.org/abs/2310.13548). The question can prompt better reasoning, but it can also prompt agreement.
Executives operate within their own version of this dynamic. A leader enters a room with a point of view, and people start to discern, is this a proposal we are meant to evaluate or a decision we are meant to affirm? Does “What do you think?” indicate genuine curiosity, or is there already a right answer? People learn quickly whether challenge improves the conversation or merely makes their position in it less comfortable.
The leader is reading the room as well. Sometimes new information causes an executive to reconsider. Other times pressure causes a leader to defend a position more strongly, even when the evidence underneath it has weakened. Still other times, leaders change direction because the loudest person in the room disagreed.
In other words, neither humans nor AI respond to evidence alone. We also respond to incentives, authority and the desire to be accepted.
Until recently, producing a polished recommendation, a persuasive rationale or a well-structured analysis required some combination of time, expertise and effort. Now, AI can generate all that in seconds, with or without the same depth behind it. The question is was your preference hidden in the prompt?
And that is why “Are you sure?” helps when you’re working with AI. Let’s be clear that it is not a magic prompt that forces either a machine or a leader to tell the truth, but it does ask them to examine what the confidence is made of.
What do we know? What are we assuming? Which part of this conclusion is evidence, and which part is interpretation? What information would cause us to change our minds? These are useful questions for AI, but they are even more important for the executives deciding what to do with its answers.
In a culture of inquiry, that question is not an accusation. It is a form of stewardship. A leader can explain the basis of a decision, acknowledge what remains unknown and identify what would justify changing course. Employees can test an assumption without being labeled negative, and executives can revise their thinking without appearing weak.
You can call that calibrated confidence: being decisive about what must happen next without pretending to know more than we do. The leaders I like working with are not threatened by being challenged. They understand that confidence is not weakened by examination; it becomes more trustworthy.
AI can shorten the distance to an answer, but it cannot determine whether that answer deserves authority. Neither can a title. That is the work of judgment, and judgment gets better when people are free to challenge it.
So ask AI. Ask the executive. Are you sure? The quality of the culture will be revealed not by whether the answer is yes, but by what happens after the question is asked.
Elsewhere In Culture
My Imaginary Board of Directors with Megan Miller
What if your board of directors included Bob Iger, Sara Blakely, and other leaders you’ve never actually met? Megan Miller joined me to explain how she built an imaginary advisory board using AI and now consults it before major decisions. We talked about why the best use of AI may be less about getting answers and more about gaining perspectives you did not have before, plus what a story about Bob Iger and Steve Jobs can teach us about the power of getting out from behind the PowerPoint. Apple: https://podcasts.apple.com/us/podcast/ceo-daily-brief-with-dr-jessica-kriegel/id1725350421?i=1000783780485 Spotify: https://open.spotify.com/episode/33lr0vx0p88Uf0BTKWTgNb?si=5fcd2d8df561444e
Invisible Value with John Frehse
John Frehse and I debated the value of the things in business that are hardest to measure: trust, context, relationships, listening, and all the information flowing beneath the surface of a decision. We spend so much time trying to quantify efficiency that we can miss the invisible factors actually helping people make better decisions and drive results. Apple: https://podcasts.apple.com/us/podcast/ceo-daily-brief-with-dr-jessica-kriegel/id1725350421?i=1000784029047 Spotify: https://open.spotify.com/episode/3aV4fSfcByZ8XfxE8HxxXg?si=417ac6bf89fa4840
The Job You Think You Have
There is a fundamental disconnect between how employees and leaders think about a job. Employees are usually handed a list of responsibilities, while leaders are hiring them because they need a result. I shared a simple exercise we use to help people stop defining their jobs by the tasks they perform and start thinking about the result they are accountable for achieving. Apple: https://podcasts.apple.com/us/podcast/ceo-daily-brief-with-dr-jessica-kriegel/id1725350421?i=1000784274977 Spotify: https://open.spotify.com/episode/1nRgflOLfylGPihw3su9Yq?si=a383994e0f544582
And coming later this week…
Solving Burnout Is Your Problem
I am over burnout. I have been burned out myself, and one of the most important lessons I learned was that waiting for my workload, my company, or the people around me to change was not going to solve it. I shared what happened when I started taking accountability for my own boundaries and stopped fighting the reality of the situation I was already in.
Narcissists with John Frehse
John and I started with research suggesting narcissistic CEOs may be more likely to pursue acquisitions and ended up debating how casually we use the word narcissist in the first place. It is easy to look at a highly visible, successful CEO and diagnose their personality from afar, but people are a lot more complicated than the labels we give them. Sometimes calling someone a narcissist says more about the person doing the labeling than the person being labeled.