What is the smallest useful first step for an AI licensing project?
Write a one-page decision brief for one archive, use and deadline, then identify the missing fact most likely to change the decision. This keeps exploratory work proportionate before an archive-wide programme is funded.
Should we prepare the archive before approaching a potential buyer?
Prepare enough metadata and approved examples to describe the opportunity and test a specific requirement. A full programme should follow documented needs, rights readiness, acceptance criteria and a budget decision.
Can an NDA authorise every requested sample use?
Do not treat it as doing so in this process; ask counsel to review the actual permitted purpose and restrictions. The release decision should identify the sample, recipient, retention, downstream access and controls explicitly.
Should a successful pilot trigger an exclusive mandate?
No automatic trigger is recommended. Assess exclusivity separately against defined uses, term, performance commitments, release conditions and realistic alternatives.
How long should the process take?
No reliable market-wide duration is established by this research. Agree stage dates from the volume of rights work, procurement access, technical scope and decision deadline, and revise them when dependencies change.
Can an adviser guarantee that we recover the cost of diligence?
No; a useful outcome may be a decision to stop or defer a transaction. Contract for clear questions, evidence standards, deliverables and cost limits rather than a promised financial return.