A buyer never calculates with the result shown on the profit and loss statement. He calculates with the profit that remains once all extraordinary items, one-off expenses and owner benefits have been removed. That process is called normalization. A company car that is actually private use, a management fee below market price, a claim that occurred only once this year: those kinds of items are corrected to show what the company structurally earns, independent of the person currently running it.
AI deployment adds a new category of corrections to that, and this category is less straightforward than a credit card statement. If part of the work is no longer done by people but by a task that AI takes over, the cost structure changes. The question a buyer then asks is not whether profit has become higher, but whether that higher profit continues to exist without the current owner, the current team, and the current contract with a supplier.
The shift at issue here falls into three categories. AI can take over a task entirely, AI can take over a task partially with a person approving or rejecting with reason, or the work remains human work. Each category normalizes differently.
A task that has been fully taken over and anchored in the process counts as a structural saving: costs go down and stay down, regardless of who leads the company going forward. A task that has been partially taken over with human oversight yields a saving that partially persists, because the oversight still costs people. And a saving that rests entirely on a single subscription with a single supplier is something different from a saving embedded in the work processes themselves — upon cancellation, a price increase, or an acquisition of that supplier, the saving disappears just as quickly as it appeared. Exactly what shifts from human work to AI work, and how firmly that is anchored, is answered per task with the work scan that determines per task what AI already takes over in the company.
Some companies already have this distinction sharply defined: they can indicate per process which step has been automated, what that step used to cost, and which contract lies behind it. Other companies only know that personnel costs have gone down and call that profit. The difference is not in how advanced the AI application is, but in how well what happened before and after the change has been recorded, and how that change is substantiated.
A buyer who receives no substantiation calculates with the worst-case scenario: he strips out the saving, or he prices in a risk. The substantiation needed to make an AI saving convincing to a buyer addresses exactly that difference between a claim and proof.
Normalized profit is an estimate, not an established fact. The method can make it plausible that a saving is structural, but cannot guarantee that it remains so. A supplier can raise its price, an AI application can be restricted by regulation, and oversight that is currently light can become heavier once the margin of error increases. Every estimate depends on how the work is organized, how dependent the company is on a single party, and how recently the change was implemented — a saving of three months says less than a saving that has held up for two years.
No personnel advice is given here either, nor any substantiation for dismissal decisions; what an employer does with its personnel is up to the employer, and separate legal requirements apply to that. What normalization does do is gather facts about which work ends up where, so that a buyer can judge for himself what those facts are worth.
An outcome, moreover, says nothing if the underlying dependency has not been taken into account. A saving that rests for the most part on the owner's knowledge, or on a single contract renewed without negotiation, weighs differently than a saving anchored in multiple layers of the company. Which part of the profit weighs down the price because that dependency is too great can be read via the dependency on an AI supplier that affects the sale price.
Normalized profit is not separate from the rest of the company. A saving that has been laid down in a contract with a customer or supplier has a different value than a saving agreed verbally that could change tomorrow; the contracts that structurally increase the value of a company show what difference that makes in the final price. And a profit increase that largely comes from a single customer is weighed by a buyer differently than profit spread across many customers — something the explanation of customer concentration and the weight a buyer gives it elaborates on further.
Normalization of profit is one of eight drivers a buyer weighs. The shift caused by AI affects not only this driver but also owner dependency — if the owner is the only one who knows which task has been automated and why, that weighs down the price more heavily than the saving itself raises it. Which driver currently weighs most heavily differs per company and is difficult to determine without substantiation.
The underlying question — which work in this company can genuinely be taken over by AI, and what of that has already been recorded — can be answered per task with the work scan from FTE TO AI. For those who first want to know where the price is currently most under pressure, there is the free value check: eight short questions, one per value driver, providing a picture of which driver currently counts most heavily. The full value scan, with maturity scores per driver and a two-year calendar toward the exit moment, is under development.