Due diligence on an applied-AI product studio

Once you are under a letter of intent to buy an applied-AI product studio, diligence exists to answer, project by project, the question the studio’s marketing page was never designed to answer: who actually owns what was built, and is that ownership documented well enough to survive a change of control? Because the studio’s most valuable assets — retained IP, equity stakes, an internal tooling stack — are intangible and were often created across dozens of separate client engagements over several years, the diligence here is less about a single master contract and more about reconciling a stack of individual agreements against each other.

The documents to request first

• The underlying contract for every shipped product the studio claims a retained interest in, showing exactly what was assigned to the client and what, if anything, stayed with the studio.
• Signed IP assignment agreements from every contractor who worked on a client deliverable or the studio’s internal tooling stack.
• Cap table, option agreements or shareholder agreements documenting any retained equity stakes, including vesting and transfer restrictions.
• Written documentation of the studio’s evaluation and testing process for shipped AI features.
• A current list of which client products still run on API keys or cloud accounts held under the studio’s own name.

Registry searches worth running

A federal trademark, patent and copyright search confirms who is actually recorded as holding any registered IP the studio claims to own, which is not always the same as who the studio’s internal summary says holds it — registrations lag behind reality often enough that this is worth checking directly rather than assuming. A Personal Property Security Act search shows whether a lender or another party has registered a security interest against the studio’s assets, which for an IP-heavy business can functionally include the tooling stack and any retained product interests, and any such registration needs to be identified and discharged before you close. A corporate search on the studio itself, and on any client company where the studio holds equity, confirms both entities are in good standing and reveals whether anyone besides who you are dealing with has a registered interest.

Checking the studio’s own vendor and platform agreements

Beyond client-facing contracts, review the studio’s own agreements with the foundation-model vendors, cloud platforms and any third-party tooling its internal stack depends on, since a vendor relationship held on non-standard or personally negotiated terms may not transfer to a new corporate owner in the way a standard commercial agreement would. Confirm whether any vendor agreement includes usage commitments, minimum spend, or exclusivity terms that would bind you as the buyer, and check whether the studio’s standing with any vendor partner program — the kind that can generate referral leads — depends on a specific named founder rather than the company itself.

Confirming the vendor-account handover is actually workable

For every client product still running on a foundation-model API key or cloud account under the studio’s name, confirm directly with the vendor or platform what is required to transfer or re-issue access without breaking the client’s live product — some platforms make this straightforward, others require a formal ownership-change process that can take longer than either side expects. Ask the studio for usage and billing history on each shared account, since a sudden spike in another client’s usage on a shared account you are about to inherit responsibility for is a cost exposure worth knowing about before close, not after the first invoice arrives under your ownership. Where a client’s product depends entirely on infrastructure the studio controls personally rather than infrastructure the client owns, treat that as a standing operational risk to manage post-closing, not a one-time item to check off.

Findings that actually kill these deals

The finding that ends the most deals is ambiguity over whether the studio or the client owns the IP in a specific shipped product, because it means the studio may be selling you an asset it does not actually have clear title to — and untangling that after signing usually requires reopening a conversation with a former client who has no particular incentive to help. A close second is a contractor-built component of the internal tooling stack or a shipped product with no assignment on file, which creates the same ownership problem one level down. A retained equity stake that turns out non-transferable, diluted, or simply never properly issued is a serious finding but usually not fatal on its own — it just means that piece of the purchase price was never really there, and the deal needs to be repriced rather than abandoned.

What a finding actually means once you have it

An unassigned contribution from a contractor who is reachable, cooperative and has no competing interest is usually a fixable condition to close — get the signature before you fund. The same gap involving a contractor who has since gone to work for a competitor, or is simply unreachable, is a materially different problem, because you cannot manufacture consent from someone who has no reason to give it. Treat every IP or ownership finding by asking how easy it would be to cure if the other side stopped cooperating tomorrow, not by how serious it looks on the page today — that question is what actually separates a closing condition from a reason to walk away.