I am often sold the return on investment of an AI project in a single line: so many hours saved, therefore so many euros saved. It looks good, it fits on a slide, and it is almost always wrong, because that sum forgets half the equation and suits the person presenting it a little too well.
The real gain is not the time saved
Let us start with the benefits column, because that is where the first illusion hides. An hour saved is only a saving if it is reinvested in something useful. If it dissolves into extra breaks or meetings that run longer, the gain exists on paper, not in the accounts.
The real benefit of automation, then, is not the time freed up, it is what your teams do with it. A salesperson who gains two hours a week and spends them prospecting, that is revenue. The same one who spends them on something else, that is zero, bar a little comfort. This nuance changes the calculation completely, and it forces the one question that matters: what are we going to do with the time this tool frees up?
Costs are not just the subscription
On to the costs column, systematically underestimated. The price of the tool is the visible part, and the smallest.
There is setup time, almost always longer than expected. There is the time to train the teams, because a tool no one knows how to use returns nothing. There is checking time, especially at the start, when you still verify everything. And there is the cost of errors that go uncaught, the most insidious of all, the one you do not see until it has blown up, and which can wipe out months of accumulated gains in one go.
An honest calculation puts the two columns face to face, over time, not over the euphoric first month. The net gain once the tool is running smoothly, against the total cost, setup, training and vigilance included. It is less flattering than the original pitch, and it is the only figure you can decide on.
The practical rule I apply
Here is the simple rule I give to leaders. A good AI project pays for itself on a precise, measurable use, not on a vague promise of transformation. If you cannot say in one sentence what it earns you, for whom, and what it really costs you, the project is not ready. It is just fashionable, and fashion has never filled a till.
This demand for precision has a powerful sorting effect. Many attractive projects do not survive the question how much, concretely. And that is fine, because they would have cost a lot for a fuzzy return. Conversely, the projects that pass this test are often modest, targeted, almost boring to present, and they are the ones that pay.
It is in fact a paradox I see often. The best return on investment is almost never the big, ambitious undertaking that was meant to change everything. It is the small project you control, that you knew how to measure, and that you extended once its profitability was proven. The spectacular impresses in meetings. The measurable ends up in the accounts. For a leader, the difference is not a detail, it is the whole gap between a cost and an investment.