How to Test a Target's AI Claims before the IC Vote

Two different assessments both get called AI readiness diligence. One tests whether the target's AI claim holds. The other tests what AI could earn once you own the business.

Key takeaways

• Demand for AI and machine learning advisory grew 238% across the 90 days, against the same period in 2025, according to the BluWave Activity Index.
• Two different assessments both get called AI readiness diligence. One tests whether the target's AI claim holds. The other tests what AI could earn once you own the business.
• Only the first one answers the question in front of the investment committee, and it has to land in three to five weeks.
• Most AI advisory capacity was built for post-close roadmap work. Deal-clock capability is a separate qualification and it is rarely advertised.
• Deferred AI work at a target is a priced item, not a reason to pass, and not something to leave for the next owner.

When a target claims AI capability, the pre-close question is whether that capability is owned, transferable, and durable. Test what the company built against what it rents from a model vendor, who holds the rights to the data underneath it, and how much of it sits with two or three people. Then price what you find.

AI diligence is a durability question, not an inventory

Deal teams rarely want a catalog of what a target has deployed. They want to know whether the thing they are paying a multiple for is still there in three years.

One partner put the brief plainly on a recent call: how strong are the moats, how quickly could a competitor or a frontier model release reproduce this, and does the business we are underwriting survive that. He tied the current pace of exits directly to that uncertainty, particularly in software.

The volume behind that question is moving fast. Demand for AI and machine learning advisory grew 238% across the 90 days, against the same period in 2025, according to the BluWave Activity Index. Most of it is post-close work, which is precisely why the pre-close version needs sorting out.

Context helps calibrate what a target's claim is worth. The U.S. Census Bureau's Business Trends and Outlook Survey put the national AI use rate at 19.8% as of May 3, 2026, with 37% among firms of at least 250 employees. Real, in-production AI at middle-market scale is still the exception. A target that says it has AI is making a claim, not describing an industry baseline.

The threat test and the headroom test are not the same engagement

Buyers usually describe two workstreams and expect them from one scope.

The threat test asks what in this operating model is exposed. Which revenue lines get compressed if a competitor ships the same capability, which functions are replaceable, and whether the target's own AI is defensible or decorative.

The headroom test asks the opposite. Where does AI create margin once you own the asset, what would you fund on day one, and what does the data environment have to look like first.

Both are legitimate. They need different evidence and different specialists. One deal team described it as two forks: what is at risk of being obsolete in three to five years, and where the quick wins are that should be teed up before close so they can start the week after. Only the first fork changes the bid.

What the pre-close assessment has to return

An IC-ready read comes back with positions, not observations. Six lines carry most of the risk.

• Owned versus wrapped. What did the company actually build, and what is a thin layer around a third-party model that any competitor can license tomorrow.
• Data rights and lineage. Who owns the training data, where it came from, and whether customer contracts permit its use. This is where deal risk concentrates and where sellers are least prepared.
• Model and vendor dependency. Which providers the capability rests on, what happens at renewal, and what a pricing or deprecation change does to unit economics.
• Key-person concentration. How many people could rebuild this. If the answer is two, the capability is a retention question before it is a technology question.
• Governance and incident history. What controls exist, what has gone wrong, and what was disclosed.
• Evaluation evidence. The measurements behind the performance claims. A demo is not evidence, and neither is a customer logo.

What a deal-clock scope looks like

Three to five weeks ahead of IC review is the working window. That constraint shapes everything about the engagement.

A deal-clock scope names its deliverables at the start: a findings memo the deal team can put in front of the committee, a risk register with severity attached, and a position on what the findings do to the model. It works from the data room and management sessions rather than an open-ended discovery period.

An open-ended readiness study looks similar on a website and behaves nothing like it. It runs on a maturity model, produces a roadmap, and lands after the vote.

The reason firms go outside for this is usually stated in the same breath. On a recent intro call, a PE firm evaluating two software portfolio companies in the same vertical said it had no engineering capability in house to judge the work, and that the companies' own technology leadership could not be objective about assessing itself. That is the requirement: an outside read across people, processes, and systems, delivered before the decision.

How to tell a deal-clock specialist from a value creation advisor

Most AI advisory capacity was built for hold-period work, because that is where the demand has been. Many firms in that category are excellent and will still miss on a live deal. Four questions sort them quickly.

• Ask what evidence they will request in week one. A pre-close specialist answers with artifacts: model documentation, data processing agreements, evaluation results, vendor contracts. A roadmap firm answers with stakeholders to interview.
• Ask whether they will take a position on price impact. Diligence work resolves to a number or a term. Advisory work resolves to a recommendation.
• Ask for references from engagements that ran inside a deal timeline, not from implementations.
• Ask what happens if the answer is that the AI is not real. A specialist who has written that memo before will say so directly.

Buyers are candid about the failure mode here. One deal partner said he did not need anyone to come in and pontificate about AI, he needed someone to answer a question. That is the bar. For the broader vetting questions that apply to any AI engagement, see how to vet an AI advisory firm before you sign.

What the findings should change

A finding that changes nothing was the wrong scope. Pre-close AI work resolves into one of three places.

Price. If the capability is rented rather than owned, the multiple was built on someone else's technology.

Terms. Data rights gaps and key-person concentration are what reps, escrow, and retention packages exist for. Findings that arrive before signing become terms. The same findings after close become surprises.

The day-one plan. The headroom work belongs here. What gets funded in the first 100 days, what the data environment needs before any of it ships, and who owns the sequence.

Where AI readiness sits in the diligence sequence

This question comes up constantly and it has a clean answer. AI readiness does not replace technology diligence. It sits inside or immediately after it.

Technology diligence establishes whether the stack can carry the business, what modernization costs, and where the debt is. That is the foundation, and pre-close technology work and post-close digital transformation are different engagements with different specialists. AI readiness then tests a specific claim on top of that foundation. Cybersecurity and talent assessments run alongside both, not after them.

The order matters because an AI finding is hard to interpret without the stack read underneath it. A model that performs well on infrastructure that cannot scale is a different asset than the same model on a stack that can.

One firm ran exactly this sequence on a software target, bringing in specialists to assess code quality, transportability, and scalability before the deal closed. The technology diligence engagement produced both an evaluation and a forward plan, and exact-fit options were in front of the deal team within 24 hours of the scoping call.

Do not leave it for the next owner

There is a pattern worth naming. Sponsors approaching an exit increasingly describe AI and data cleanup as work for the next owner rather than something to fund in the final stretch. The logic is understandable. The arithmetic does not hold.

Deferred AI work does not stay invisible. It shows up in the next buyer's diligence, and it shows up as a discount rather than a line item. A target that arrives with unresolved data rights, a rented capability presented as owned, and a two-person dependency is not neutral in a process. It is cheaper.

The same logic runs in reverse for buyers. A target with real, documented, owned capability is worth paying for, and a target without it is worth buying at the right number and fixing on your clock. Neither conclusion is available without testing the claim first.

The work is a hold-period asset either way. The firms treating AI readiness as something to establish now, on assets they already own, are the ones who will not be answering these questions under time pressure in a sale process.

Where BluWave fits

BluWave, the private equity market network and enablement platform, connects private equity firms and their portfolio companies with BluWave Vetted™ AI diligence specialists, technology diligence groups, and AI advisors matched to the question in front of you. Deal-clock engagements are scoped to the IC timeline, not the hold period. Share your need and exact-fit options arrive within 24 hours, with no cost to connect.

Frequently asked questions

What is AI readiness diligence?

AI readiness diligence is a pre-close assessment of whether a target's AI capability is owned, transferable, and durable. It covers what the company built versus what it licenses, data rights and lineage, model and vendor dependencies, key-person concentration, governance, and the evidence behind performance claims. The output is a findings memo the deal team can take to the investment committee.

How is AI diligence different from technology or IT due diligence?

Technology diligence assesses the stack as a whole: architecture, code quality, scalability, security posture, and modernization cost. AI diligence tests a specific claim sitting on top of that stack. The two are complementary, and AI diligence is difficult to interpret without the underlying technology read. Most deal teams run them together or in sequence, not as alternatives.

How long does AI readiness diligence take?

Typical engagements run three to five weeks ahead of IC review. That window is the constraint that separates diligence-capable specialists from advisory firms scoped for hold-period work. If a provider quotes a timeline in months or leads with a maturity model, they are describing a value creation engagement rather than a pre-close assessment.

What evidence should a target be asked to produce?

Model and system documentation, data processing agreements and customer contract language covering data use, vendor agreements for any third-party models, evaluation results behind stated performance, incident and governance records, and a clear map of who built and maintains the capability. Sellers are frequently least prepared on data rights, which is also where deal risk concentrates.

What does an AI readiness finding change about price or terms?

Rented capability presented as owned goes to price. Data rights gaps and key-person concentration go to terms, through reps, escrow, or retention packages. Headroom findings go to the day-one plan. A finding that changes none of the three usually means the engagement was scoped as general AI advisory rather than diligence.

Who runs AI diligence when the firm has no technical partner in house?

An outside specialist, for the reason most deal teams state directly: the firm lacks the engineering depth to judge the claim, and the target's own technology leadership cannot assess itself objectively. The deal team still owns the questions and the decision. The specialist supplies the read, and through BluWave that match arrives within 24 hours.