A buyer likes to hear that AI has taken work out of the company. The conversation that follows concerns a different question: who arranges for this to keep working when the current vendor stops, raises prices, or changes the product. Savings that hang on a single external party weigh differently in a valuation than savings anchored in the company's own process. Not because AI is unreliable, but because a buyer pays for what keeps running without depending on chance.
Dependency on an AI vendor does not sit in one thing. It sits in at least three layers, and each layer counts differently in a buyer's eyes.
The first layer is the contract: a subscription that can be cancelled monthly weighs differently than a multi-year licence with specific conditions. The second layer is the data: is the system's output portable, or is the knowledge locked into an environment that does not simply move to another vendor. The third layer is the person: is there someone in the company who understands how the system is set up and why, or is that knowledge something that exists only with the vendor itself.
A company where AI partially takes over a task, with an employee who assesses the outcome and can adjust it, keeps that third layer intact. A company where the task has been fully outsourced to a system that no one internally still understands has lost that layer. That difference is exactly what why a buyer asks who maintains the AI is about.
The shift that this all revolves around is not something for the future. In some companies, customer service already largely runs through an AI system with an employee who only steps in for exceptions. In another part of the same sector, the same work is still done entirely by hand, or is fully automated without any human control. That difference does not come from ambition or from budget. It comes from how a company has set up the takeover of work: is there oversight with a reason for approval or rejection, or is that oversight missing entirely.
Companies with oversight can show a buyer why a result is correct and who checks it. Companies without oversight can only show that the result exists, not why. That is an information gap, and information gaps are exactly what a buyer factors into the price.
An assessment of vendor dependency is an estimate, not a measurement with decimal points. It relies on what is demonstrable: contract terms, presence of internal knowledge, portability of data and the degree of human oversight of the output. Where these four things are not documented or cannot be substantiated with evidence, the estimate remains broad and is expressed as a range, not as a precise figure.
The outcome also says nothing about whether it is wise to deploy staff differently. That touches on decisions subject to their own legal requirements, and that assessment lies outside this method. What the method does do: map which part of the realised benefit depends on an external party, and which part is anchored within the company itself. That split is what counts in the earnings quality a buyer assesses, as also shown in what working capital is and why it counts in a sale, where one-off or external benefits are weighed differently from structural ones.
Beneath the question of vendor dependency lies a broader question: which work in this specific company can truly be taken over by AI, which part only with oversight, and which part remains human work. That question is answered task by task in the FTE TO AI work scan, and only with that answer in hand does it become clear how heavily a vendor dependency really weighs on the price.
This also touches on who carries the work within the company. If the knowledge about the AI system sits with one person, that counts towards how you prevent your value from sitting in a handful of heads, and towards what succession readiness means for the price. A vendor that can be replaced is one risk; an employee who cannot be replaced is another risk, and they are often assessed in the same breath.
A buyer accepts that AI plays a role in the result. What matters is the story behind it: which task has been taken over, with what oversight, and what happens if the vendor falls away. That story, and how it is convincingly presented, is the subject of what story you tell a buyer about the work AI does. Separate from that, but equally relevant in a sale, are the contracts and terms themselves: which vendor agreements are legally transferable comes back in which legal matters need to be in order for a sale.
Whether an AI vendor pushes down your price, and how hard, cannot be answered with a hunch. It requires evidence: contracts, data transfer, internal knowledge, degree of oversight. The free value check offers a first indication with eight short questions, one per value driver, resulting in a picture of which driver is pressing hardest on your price today. The full value scan, with maturity scores per driver and a two-year calendar towards the exit moment, is under construction.