A founder I spoke with last year was quietly pleased with his AI position going into a sale process. The team was using generative AI across sales, support, and reporting. Tools had been bought, a few things had been automated, and he assumed all of it counted in his favour. He had taken visible adoption as evidence of a modern, valuable business.
The diligence team read it differently. By the time they had finished, the AI he was proud of had become a line item working against the price, not for it.
That gap, between what a founder assumes acquirers reward and what they actually reward, is now wide enough to move a headline number by tens of percent. Here is how the two sides of the ledger work, and why the side you land on is mostly decided long before anyone opens a data room.
What do acquirers actually assess on AI now?
By 2026 acquirers assess AI as a distinct line in due diligence rather than folding it into general technology checks. They score it across five weighted dimensions, infrastructure readiness at around 30 percent, data governance at 25 percent, model lifecycle at 20 percent, risk controls at 15 percent, and strategic alignment at 10 percent. The presence of AI is no longer the question. Its depth and ownership are.
The shift matters because each dimension is testable. Infrastructure readiness asks whether your core systems can carry AI rather than just host it. Data governance asks whether you can show where training and operational data came from and that you had the right to use it. Model lifecycle asks whether models are validated, version-controlled, and monitored rather than left to drift. Risk controls ask what happens when a model is wrong. Strategic alignment asks whether any of this connects to the business in a way you can evidence.
A diligence team does not take your word on these. They request model documentation, ask for data lineage, and in higher-value deals they stress-test the systems that drive value. What they find against each dimension is what becomes the AI premium or the AI discount.
What earns a premium?
The premium goes to AI embedded in core processes with a clear, evidenced business case. Acquirers pay for implementations that contribute to a specific revenue gain or cost reduction you can document, that sit inside how the business genuinely runs, and that are backed by validation, monitoring, and clear ownership. Businesses with strong AI governance have achieved revenue multiples well above peers with weak maturity.
What this looks like in practice is unglamorous. It is a model inventory that lists each system, its purpose, its owner, and its validation status. It is data lineage you can produce on request. It is evidence that a named person is accountable for each implementation rather than a system everyone uses and nobody owns. It is a business case that says this AI does this job, saves this much, and here is the historical data that proves it.
None of that requires the most advanced tools. It requires the work to have been done deliberately, with the diligence question in mind, rather than bolted together for show. That is the difference an acquirer pays for.
What earns a markdown?
The discounts cluster around three patterns, all common where AI was delegated without strategic oversight. Superficial adoption with no measurable impact, often called AI theatre, has drawn documented discounts of 18 to 25 percent. Unowned, unmaintained systems, the so-called AI orphans, draw a further 15 to 25 percent. AI bolted onto legacy infrastructure adds technical debt the buyer prices in as remediation cost.
The mechanism behind each discount is the same. The buyer cannot trust what they cannot verify, so they price in the cost of fixing it. AI theatre gets discounted because diligence reveals the AI claimed to drive value has no business impact behind it, which undermines the growth story the price was built on. AI orphans get discounted because unmaintained systems develop reliability problems, and a buyer has to fund stabilisation before they can rely on the outputs.
The painful part for a founder is that visible adoption can sit on the wrong side of this line. Tools in use across the team, with no ownership, no validation, and no documented business case, read to a diligence team as risk rather than value. The AI you were proudest of can be the AI that costs you the most.
Why does strategic alignment depend on you?
Strategic alignment carries the smallest weighting at around 10 percent, yet it is the one dimension only the founder can supply. It asks whether each implementation has a clear business case, whether its metrics tie to strategy, and whether there is evidence of executive oversight rather than an isolated experiment. That evidence cannot come from the operator running the tools. It has to come from the person who owns the strategy.
This is where delegated programmes tend to fall down. When a founder hands the AI mandate to a capable operator without the strategic context behind it, the operator optimises for what they were briefed on, usually functionality and adoption. They build things that work. What they cannot build is the connection between those things and the business case a buyer is paying for, because that connection was never theirs to make. Diligence teams treat last-minute executive involvement as superficial, so a founder who turns up to the data room having never engaged with the AI strategy cannot retrofit the alignment dimension.
The fix is a behaviour rather than a tool. It means the founder staying close enough to the AI programme to give it strategic direction, document the business case, and show genuine oversight over time.
What should you ask for before the brief goes out?
Before you hand the AI mandate to anyone, decide that the programme will aim at the rewarded behaviours from the start. The work that earns a premium, data governance documentation, model validation, clear ownership, and an evidenced business case, takes 6 to 18 months to do credibly. That timeline makes the decision yours now, not the operator’s later, because retrofitting it under deadline reads as theatre and gets discounted.
Ask your operator for three things in the brief. First, a clear business case for every significant AI implementation, with the metric it moves and the evidence behind it. Second, named ownership for each system, so nothing becomes an orphan when someone leaves. Third, documentation as you go, model purpose, data sources, validation, and monitoring, built into the work rather than compiled at the end.
If an exit is anywhere on your horizon, this is a founder decision dressed as an operational one. The AI programme either compounds into value a buyer will pay for or accumulates debt a buyer will discount, and which one it becomes is set by the brief you give today. If you want to think through what that brief should contain for your business, book a conversation.
