AI Economics Playbook

The Board Isn't Testing You. Sometimes They're Right.

Hard questions ask for evidence of discipline. Show the bets you stopped, the trajectory you track and the cost of standing still.

Kiran SusarlaOriginally published October 23, 2026 · 7 min read
Two people pause at a blue footbridge while one holds a lantern over a missing plank in the walkway.
Conceptual illustration: a lantern reveals a missing bridge plank before anyone crosses. Illustrative, not a measured assessment of any AI program.

Executive summary

Board scrutiny can reveal a real gap in an AI program, rather than simply a need for a better-rehearsed answer. The article connects difficult questions about spending, payoff and waiting to concrete evidence: stopped initiatives, portfolio trajectories, explicit stop conditions and the cost of inaction. Defensiveness is a signal to examine what has not been checked.

The Board Isn't Testing You. Sometimes They're Right.

Years ago I worked alongside a quality engineer who had a habit that annoyed almost everyone in the room. He would wait until a review was nearly over, until the deck was closed and people were reaching for their laptops, and then he would ask one more question. It was never hard on its face, rarely took five minutes to ask, and rarely sounded clever. But often enough to be unsettling, it turned out to be the one question that mattered, the small loose thread that, once pulled, unravelled something we had all quietly agreed to stop looking at.

I disliked those questions at the time (I will admit that), mostly because of the timing and the habit of it. I mean, come on, you know... why not earlier, or somewhere in the early or middle of the conversation, and not just as the room is trying to conclude! They felt like someone slowing down a train that was finally moving. I recognized all the while that these were not obstacles but sort of warnings, and the discomfort I felt was not really about him at all. It was the sound of something the team had not checked, could have checked, or perhaps did not want to check.

I think about him often lately, because the conversations I am part of about AI, whether informal ones, the work of helping clients reason through their AI initiatives, or the rooms at the conferences I have been going to, keep circling the same scene. A finance leader walks into a board meeting. The spending has climbed past the number everyone approved a few quarters ago. The returns are not yet obvious. And the board has questions, asked in that particular tone that is polite, pointed, and just slightly impatient. It is not always a formal board. Sometimes it is the CEO, sometimes the executive committee, sometimes whoever ultimately holds the purse. Wherever it happens the meeting has the same temperature, and the questions rhyme, so read "board" here as whoever you answer to for the money.

Evidence instead of a script

The instinct is to treat the meeting as a test to be passed, an exam with answers rehearsed and ready. A small industry exists to hand finance leaders exactly those scripts: say this about cost, say that about timelines. But a script can only answer the question it anticipated. The board member who asks the follow-up nobody rehearsed, the quiet one who waits until the deck is closed, walks straight past it and into the thing it was covering for.

There is a different way to walk into that room, and it is less about having answers than about knowing how to listen. Underneath almost every hard question a board asks about AI is the same question wearing a different costume. Is this program disciplined, or is it not? A big number, a missing figure, an unflattering comparison, an absent payoff date, these all look like separate ambushes, but each is really asking you to point at the evidence of discipline. So the move, when a question stings, is not to reach for a better sentence. It is to reach for the thing the question asks you to point to, and notice, honestly, whether it is actually there.

Show what you stopped

Take the one that usually comes first, asked with an edge: why is this so expensive? The weak reflex is to explain that AI is hard and talent is scarce, all of which may be true and none of which helps, because it sounds like a list of apologies for a number that got away from you. What the question is really asking you to point at is the things you stopped. A well-run program is not one large bet that grew expensive; it is a portfolio of smaller bets, most of them expected to fail, funded precisely because the few that work pay for the rest. If you can say, plainly, here are the initiatives we killed quickly and on purpose, that is why some of this spending produced nothing, then the large number stops being evidence of waste and becomes evidence of a portfolio doing what a portfolio does. But that answer is only available to you if you actually ran it that way. If you have spent heavily and killed nothing, you do not have a portfolio, you have a collection of projects no one had the nerve to end. And the board has not been difficult. They have been right.

Show the trajectory

Or the pointed one, really asking whether to keep funding you at all: when does this pay off? The tempting answers are both bad. A confident single date is a fiction you will be held to later; "these things take time" is true and fatal, because it hands the board nothing to hold. What the question is asking you to point at is a trajectory. Not a date, a shape: here is where we are, here are the initiatives already returning value, here is the point where the whole portfolio crosses into net positive, and here are the specific signals that would make us stop. Stating the stop conditions feels like inviting doubt, but it is the single most reassuring thing you can say, because it turns you from an advocate defending a budget into a steward managing one. And again the test cuts both ways. If you genuinely cannot draw the trajectory, cannot say where the crossover is or what would make you walk away, then the uncomfortable truth is that you do not have a plan, you have a hope, and the board asking for the shape has done you a favour by revealing it.

Price standing still

Or the one that turns the whole conversation around: what happens if we just do nothing, wait a year, let the prices and the technology settle? The defensive reflex is to wave at the fear of being left behind, which sounds like fashion rather than analysis, exactly the impression a skeptical board is primed to distrust. What the question is asking you to point at is the price of standing still, because doing nothing is not free, it is a choice with its own cost. The ground you lose, the capability that takes years to build and cannot be bought in a hurry later, the thing you only learn by doing. Lay those out as numbers and the decision becomes an honest comparison of two costed paths rather than a suspicious interrogation of the expensive one. But if you find you cannot price inaction, if your whole case for acting rests on everyone else doing it, then you may be funding AI because it is expected rather than because it earns its place, and the board member who asked has quietly done the work you skipped.

You will have noticed the pattern by now, because it is the same every time. The question might come across like an attack. The honest signal is whether you can point at the thing it asks for. And if you can, you say so with total confidence, and the room relaxes. If you cannot, the board has caught something real, and no better-worded answer will fix it, because the problem was never the answer.

Listen to the flinch

Which leaves one more tell, the most useful of all, because you can read it in real time, before the question even finishes. It is your own flinch. That hot little urge to get defensive, to reach for the complexity of the technology or, worst of all, your own credentials, that urge is diagnostic. When your program is genuinely disciplined, hard questions do not frighten you, they give you something to show. The defensiveness rises in exact proportion to the gap between the discipline you are claiming and the discipline you actually have. So when you feel it, do not push it down and reach for a smoother line. Treat it the way I learned to treat that quality engineer's one-more-question. Not as an obstacle to the train. As the sound of something you had not wanted to check.

The board meeting, in the end, is not a threat to survive. It is the discipline you have (or have not) been practising all along, applied out loud, under pressure, in front of the people you answer to. If you have done the quieter work, found the hidden spending, valued the bets honestly, stopped the ones that failed, the meeting is simply where you get to show it. And if you have not, the meeting is where you find out, which is not a pleasant gift but is a gift all the same, because it arrives while there is still time to act on it.

I am not pretending this makes the room comfortable. It rarely is. But I have come to think the hardest question in it is usually the early warning you would have been wiser to hear sooner. I would be curious whether that matches what you have seen from the other side of the table.

#CFO #FinOps #EnterpriseAI #AIGovernance #AIStrategy

Editorial note: The original article wording and sequence are preserved. The executive summary, section headings, takeaways and cover illustration are editorial additions.

Continue reading

Browse Insights