One payback number hides the argument

Somebody puts a business case in front of an investment committee and it says the build pays back in fourteen months. One number, in bold, at the bottom of the page. It is the most confident thing in the document and it is the least useful.

The problem is not that the number is wrong. The problem is that a single figure buries the only thing worth discussing, which is what the number is made of and how much of it you would still believe on a bad day.

So we state payback three times, at three levels of evidence. Cash only counts the benefits you could prove with an invoice: the licence you stop paying, the overtime you stop authorising. Cash plus protected adds the margin or revenue you stop leaking, which is modelled but traceable to a report somebody already runs. All valued adds capacity released, where the hours have a named redeployment and a named owner. Same build, same assumptions, three answers, and the spread between them is the argument.

I ran this recently on a build where the three numbers came out at fifty-six months, twenty-one months, and fifteen. That spread tells you something no single figure could. On cash alone it does not pay back in three years and it is not close. The case rests entirely on the margin protection, which is modelled. So the conversation stops being about whether to fund it and becomes about whether that one modelled number is credible, which is exactly the conversation the committee should be having.

That is the point. You are not trying to make the number look good. You are trying to put the actual decision in front of the people who have to make it.

And it makes the case stronger, not weaker. When you say plainly that the build does not pay back on cash alone, and then make the argument for the modelled benefit rather than quietly moving it into the cash column where it would produce a flattering figure, you have done two things. You have shown you understand your own case better than anyone in the room. And you have removed the thing they would otherwise have spent the meeting hunting for.

Then name the single number that breaks it. In that example, the whole case turned on a sixty percent reduction in below-floor pricing. At thirty-five percent the payback stretched from twenty-one months to twenty-eight. So we wrote that down, in the document, above the signature block: here is the figure that decides this, and here is what happens if we are wrong about it. Ask about that one first.

Nobody does this, and I have never understood why. Putting your weakest assumption in front of the committee is the single fastest way to be trusted about everything else. It is also just accurate. Every business case has one number holding it up. Pretending otherwise does not make it less true, it just means somebody else finds it later, in a worse meeting.

Two more things belong on that page and are usually missing. The first is a realisation curve. Almost no benefit lands on day one, and a model that assumes full value from month one is the most common reason a case is later found to have been wrong. Forty percent in the first six months while people are still learning it, most of it by month twelve, is exactly the right level of precision. The second is the counterfactual: what else could move these numbers in the same period, and how will you tell the difference? If a pricing policy change lands the same quarter as the build, you will not be able to separate them unless you decided in advance how.

None of this is financial sophistication. It is the same operating discipline that makes anything else work. State the outcome. Say who owns it. Say when you will check. Say what would prove you wrong. The reason it matters more with AI builds is that the technology is unfamiliar enough that people stop applying the standards they would apply to anything else, and a fourteen-month payback with nothing behind it sails through on novelty.

Give them three numbers and the assumption that separates them. Define it, measure it, own it, close it. A committee that can see the argument will make a better decision than one handed a conclusion, even when the decision goes against you.