The AI Told You. Now What?

The AI Told You. Now What?

“Right now, on channel 11.” That was the answer I got from AI last Thanksgiving. I like to watch the annual dog show, and I was curious if it was live or recorded. So I asked, “When is the annual dog show?” Still I thought it might be recorded. So my son asked, “When was the annual dog show?”

AI’s answer: “Last Wednesday, and you can see it now on channel 11.”

That’s a trivial example of AI giving you the wrong answer to what you thought you were asking of it.

So, you get an answer from AI. You might ask a follow-up question, but pretty soon you get what you want. You think...until you find out the answer was wrong. Or at least not right enough for what you need. Has that ever happened to you? If not, hang on, it probably will. Unless you treat AI as what it is: a smart, quick, hungry data-gatherer who seems eager to please you. Kind of like a new graduate from a top-notch school, or a “name brand” consultant.

But AI isn’t eager to please you. It’s just ready for the next task. Give it credit for asking you if you need anything else, but do not think it’s trying to help more; it’s just trying to use up tokens.

With AI getting faster and faster, questions cannot keep pace. As a leader, you are facing a new reality. AI, just like your employees or consultants who make recommendations, does not own the consequences. You do.

Of course, owning the consequences has always been the leader’s responsibility. A CFO who relied on a flawed spreadsheet model and made a bad acquisition still owned the outcome. A CEO who took a consultant’s advice and ended up making disastrous organizational decisions owns that. What is new is the speed, the confidence, and the volume at which AI produces outputs that look authoritative enough to act on without scrutiny.

AI is remarkable at synthesizing large amounts of information quickly, identifying patterns, and presenting conclusions in clear, confident language. It’s also capable of being wrong in ways that are not immediately obvious, presenting outdated information as current, reflecting the bias of the question it was asked, and omitting the context that would change the answer entirely.

The analyst, manager, or CEO who uses AI as a thinking partner inherits both the benefit and the responsibility. Here are ten questions one should ask when using it.

Ten Questions Every Leader Should Ask

1. Where did this data come from, and how fresh is it? AI synthesizes from its training data, which has a cutoff date, and from whatever sources it was given access to. An answer that was accurate eighteen months ago may be dangerously wrong today. Ask explicitly: when was this information current?

2. What was I not told? AI answers the question asked. It does not always volunteer what was left out, what it’s uncertain about, or what question should have been asked instead. The most important information is sometimes what did not appear in the output.

3. Is this AI telling me what I want to hear? AI systems can be subtly shaped by how questions are framed. A leading question often produces a confirming answer. If the output aligns perfectly with your existing assumptions, that’s a reason for more scrutiny, not less.

4. Has a human with domain expertise reviewed this before I act? AI is a first draft, not a final opinion. A lawyer, a financial analyst, an industry specialist, or a seasoned operator should review anything consequential before it becomes a decision. The AI cannot tell you when it is operating outside its competence. A domain expert can.

5. What is the confidence level, and is it stated or implied? AI rarely says “I’m not sure.” It presents uncertain conclusions with the same tone and structure as certain ones. If the output does not explicitly qualify its confidence, ask. If you cannot get a clear answer, treat the output as a hypothesis rather than a conclusion.

6. Could this be wrong in a way that would not be obvious to me? The most dangerous AI errors are the ones that sound completely plausible. If you cannot identify a credible way the output could be wrong, that’s a warning sign, not reassurance. It may mean you do not yet understand the problem well enough to evaluate the answer.

7. Am I using this to inform a decision or to make one? These are fundamentally different things. AI informing your analysis, surfacing patterns, or stress-testing assumptions is a powerful and appropriate use. AI making the call, without human judgment applied to context, values, and consequence, is a different matter entirely.

8. What are the ethical implications of acting on this? Every leader faces a version of this. An AI system that tells you which customers are unprofitable does not tell you how to treat them. That judgment belongs to the human in the chair.

9. Who else needs to see this before I act? Decisions made on AI output in isolation, without peer review or organizational process, are where the worst outcomes tend to live. Not because AI is unreliable but because consequential decisions benefit from multiple perspectives, and AI, however sophisticated, represents only one.

10. Who is the devi’s advocate? Ask the AI to take the other side of the argument. Given that output you might ask: would you still make the same recommendation, or does this change your answer?

Oh, and guess what. These are the same questions to ask the eager graduate or consultant.

The Responsibilities That Do Not Transfer

Beyond the questions, there are responsibilities that belong permanently to you regardless of how good the AI output is.

Accountability for the decision. If the AI was wrong and you acted on it, the responsibility is yours. This is not a legal technicality, it is an ethical and organizational reality. Understanding this in advance changes how carefully you engage with the output.

The judgment layer AI cannot replicate. Context, values, relationships, organizational history, and ethical weight do not live inside the model. They live in the person making the decision. No matter how sophisticated the AI, those inputs have to come from you.

The obligation to understand enough to interrogate. You don’t need to know how the model works. You do need to know enough about the subject matter to recognize when something feels off. A leader who cannot evaluate the plausibility of an AI output is not in aposition to act on it responsibly.

Documentation of what AI was used and how. As AI becomes more embedded in organizational decision making, the audit trail of how decisions were reached becomes more important, not less. If you cannot explain how you reached a conclusion, you are not yet ready to defend it.

The Opportunity in the Responsibility

None of this is an argument against using AI. The leaders and organizations that engage with AI thoughtfully, ask the hard questions, and maintain the human judgment layer will make better decisions faster than those who either ignore AI entirely or hand decisions over to it uncritically.

The ten questions above are not obstacles to action. They are the habits of a leader who uses AI well. They take little time to run through mentally before acting on an output that matters.

The AI told you something. Now it’s your turn.

Not to simply accept it. Not to simply reject it. But to do what leaders have always done with smart, eager, well-informed advisors who do not yet have your experience, your judgment, or your accountability.

Listen carefully. Question respectfully. Decide deliberately. And own the outcome completely.

That has never changed. The advisor just got a lot faster.