Seller handbook

A practical route from archive to licensing conversation

An end-to-end workflow for preparing, evaluating and licensing data to AI companies.

Sources and guidance · Reviewed October 2026

Start with a data inventory

List each system, the owner responsible for it, the type of material, its date range, and whether the files can still be exported. Measure usable records rather than storage size. An old shared drive full of copies is not the same asset as a coherent history of decisions.

Do not send an archive to a prospective buyer just to find out whether it is relevant. Begin with a high-level description that excludes personal details, customers and confidential project names.

Resolve the right to offer it

Trace ownership to employment agreements, contractor assignments, content licences and customer contracts. Review privacy notices and confidentiality obligations separately from copyright. Export access establishes technical possession, not permission to license.

Record unresolved questions in a rights register. Exclude problematic material until counsel or the rights holder resolves them.

Make the package understandable

Prepare a datasheet with coverage, formats, limitations, transformations and proposed permitted uses. Offer a small reviewed sample under appropriate evaluation terms. Keep a versioned manifest of the files that the sample represents.

Compare the whole agreement

Ask each counterparty to state its use, term, exclusivity, onward-sharing rights, security obligations, acceptance process and payment timing. Compare net proceeds after preparation, advice, fees and any retention obligations.

Deliver against a written acceptance plan

Agree the format, secure transfer route, checksums, test window and rejection grounds before delivery. Keep the accepted manifest and contract together. Record what must happen to rejected copies and how payment will be evidenced.

Use this as a preparation framework. It is not legal advice, a valuation or a promise that a buyer will accept your data.