Start with the work a buyer needs to understand.
Organizations accumulate examples of expert work: decisions, corrections, complex cases, and the context that makes a good outcome distinguishable from a plausible one. Some of that material may help train or evaluate AI systems. The important word is may. Volume alone does not establish usefulness or demand.
OpenAI’s data partnerships program describes interest in public and private datasets. Its application asks about data type, size, rights, and preparation needs. Those are useful initial diligence questions; they are not a promise that a particular archive has a market.
Determine what you can actually share.
Ownership of an operating system does not automatically establish rights to commercialize everything inside it. Contracts, personal information, customer obligations, confidentiality, and third-party contributions may constrain the material. An evaluation should begin with the organization’s legal and privacy advisors before samples leave its control.
A useful inventory identifies the source, permission basis, sensitive fields, preparation required, and approved uses. Removing names alone may not resolve the risk: context can reveal identities or expose confidential decisions.
Test usefulness before building a large supply operation.
Ask what a prospective buyer would do with the material. Does it represent a hard task? Does it include the inputs, the expert decision, and a way to judge quality? Can the work be reproduced or updated? A small, appropriately cleared sample can establish more than a broad claim about access to millions of records.
Any test needs agreed limits on access and use. Commercial terms should address scope, exclusivity, permitted derivatives, security, payment, and what happens when the relationship ends. The right structure depends on the material and parties; there is no universal price per record.
Separate opportunity assessment from a commitment.
Transformative works on the company’s side to assess the opportunity and help develop a commercial path. The company’s legal and privacy advisors provide legal advice and review agreements. Exploration should leave a clear next decision: proceed with a defined pilot, improve the material, or stop.
The ambition is to make valuable expertise usable without losing control of the obligations attached to it. That is both a product question and an operating question.