AI for CEOs: walk into the MT with numbers you can defend

You know roughly what the bank balance is. What you do not know, at 21:00 on a Sunday, is what happens if Q3 collections slip two weeks and the September payroll lands before the receivables do. Your controller can answer it, on Wednesday, in percentages. You decide in euros, and the moment that matters is not the month you run short, it is the week you should have called the bank. The file that holds the answer, 13-Week_Cash_Forecast.xlsx, is usually one version behind and buried in a mail attachment.

Monday the MT presents the forecast. Sales talks pipeline, operations talks the order book, finance talks budget, and the three do not reconcile. You sign the board pack anyway, because there is no version of the afternoon in which you personally cross-read Budget_vs_Actuals_2026.xlsx against the order list. Then a shareholder or a bank asks the one question underneath the number, and you are defending a figure you did not build. That exposure, not the gap itself, is the part that costs you sleep.

Revenue is up and the bottom line is flat. Nothing in the management report says anything is wrong, because nothing is wrong at the level the report is written: the erosion sits inside individual accounts, in maintenance contracts that were never indexed, a few points at a time. Nobody produces that view without being asked, and you only think to ask in month six. It is money already given away, and the reason it hurts is that it was avoidable housekeeping, not a market shift.

Ask how many months of liquidity you really have if collections slip, and you get the base case and the downside together, the specific week the pinch lands, the amount you would be short, and what to do about it while you still have room to do it. Months and percentages are paired with an amount, because that is how the decision is actually made. Every figure names the file and version it came from, so you can check it in seconds rather than trusting it blind.

Ask whether you are going to hit budget, and you get the gap located rather than just reported: how much of it is underlying and how much is two named orders sliding into the next quarter, with the customer and order number attached. The distinction between a timing problem and a revenue problem decides whether you intervene, so it is stated plainly, along with how firm the conclusion is and what the file cannot tell you.

Ask where margin is going and you get the accounts it concentrates in, how many points each has lost and since when, the annual amount involved, and when each contract comes up for renewal. That last detail turns a finding into a Monday action: which contracts to index before they roll over. It is a signal to act on, presented as one, and never dressed up as a diagnosed cause when the file only shows the pattern.