The board asked an AI question the business did not have

Meta description: AI can produce detailed plans for the wrong business objective when leaders omit identity, purpose, context, and intended outcomes.

I keep encountering board packs that are technically excellent and decisionally empty. The AI section has accuracy figures, adoption curves, cost trends. The board asks informed questions. Everyone leaves feeling current. And nothing in the meeting would have made any decision different the following Monday. The questions being asked are good questions — about the model. They are not the company’s questions, and the precision of the model’s answers is exactly what makes the substitution hard to see.

A board can ask an AI system how to acquire twenty percent more customers and receive an elegant, detailed growth plan. The recommendations might be reasonable. But imagine that the company’s larger problem is customer loss: new customers arrive, then leave because the experience does not meet expectations. The model has answered the acquisition question. It has not established that acquisition was the question the company needed to ask.

What does a wrong question look like when it is well presented?

A wrong question, well presented, looks like diligence. It asks about model performance, data coverage, vendor risk. These are real topics and they are also, in the cases I watch, a way of talking about AI without talking about the business. The tell I have learned is simple: if the answer could not change any decision the leadership team owns, the question was about the model, not the company. Rooms full of smart people can spend a year on model questions and call it AI governance.

Why do precise model answers make the error harder to see?

Miklos Roth

My foundation is straightforward: Who are we? What do we want? Why do we want it? How do we want to achieve it? I use that sequence as a way to make the purpose visible before asking for external assistance, whether the assistance comes from another person or from an AI system.

The first question establishes context. A property advisory business and a subscription software company can both ask for growth advice, but their business realities are not interchangeable. The second names the desired change. The third forces the organization to explain the reason; it may reveal that the apparent target is only a proxy for something more important. The fourth provides constraints on implementation and on what a usable answer should look like.

Precision is seductive in a board setting. A number with a decimal point closes a discussion; a vague strategic discomfort opens one. I have watched committees gravitate toward the questions the model can answer precisely, because those meetings end on time and produce clean minutes. The company’s actual questions — the ones about what the business should now do differently — arrive with no decimal points attached, and they keep getting scheduled for a later agenda.

What replaces a decision in these meetings?

The false explanation is that the committee needs better dashboards. I have watched dashboards improve and the meeting stay exactly as empty. What replaces the decision, in the pattern I keep seeing, is the request for more information. Each meeting ends with an action item to bring additional data, and the additional data arrives, and it answers the question that was asked, which was never the decision that was needed. The information request is a decision-shaped object that lets everyone leave without deciding.

What pattern separates a real executive question from a model question?

A real executive question has a decision attached: if the answer is X, we do this; if Y, we do that. A model question has a report attached. Across the steering committees I have watched, the drift is consistent — the agenda fills with report-questions and the decision-questions migrate to hallway conversations, where they are discussed by two people and then dropped. The committee is functioning. It is functioning on the wrong class of question.

I see this as a form of insufficient self-knowledge, not merely a technical prompting mistake. Many people begin by asking a question before they have clarified what they want to accomplish. The same tendency appears in companies. Leaders may assume that their objective is obvious because it has been repeated in meetings. To a model, the unstated reasoning is not there.

A prompt can be beautifully written and still lack business substance. I would rather provide a model with relevant history, operating limits, an explicit goal, and the reason for pursuing it than rely on a clever phrase without context.

I sometimes describe an LLM metaphorically as a micro-universe built to assist us — a system that can connect and generate information across a vast range of topics. The metaphor should not be confused with a scientific claim that the model understands reality as a person does. A model can be broadly knowledgeable and still fail to infer the company’s unstated situation.

What is the question under the question?

In every one of these stories, there is a moment — usually after the meeting, in a corridor — when someone says the actual thing. The committee had spent an hour on model metrics, and in the corridor a person says what they are really uncertain about, and it is a different question entirely. That question, the corridor question, is the one the agenda was built to avoid. It has a shape I recognize now.

The question I would leave in a leadership meeting is not “What can AI recommend for our target?” It is “Have we explained clearly enough who we are, what we are trying to change, why that change matters, and what would count as a useful answer?”

PR excerpt (80–120 words, Miklos’s own words): A board can ask an AI system how to acquire twenty percent more customers and receive an elegant, detailed growth plan. But if the company’s larger problem is customer loss, the model has answered the acquisition question without establishing that acquisition was the question to ask. My foundation is straightforward: Who are we? What do we want? Why do we want it? How do we want to achieve it? A detailed answer can disguise how much the prompt left unresolved.

By Miklos Roth — AI marketing consultant, Budapest. More: Miklos Roth — AI marketing consultant.