Licensing company records without giving them away
Ownership stays with you. A plain guide to what you can license, what never leaves your company and the three ways to structure a deal.
Companies sit on years of records that show how skilled work actually gets done. AI teams want that record because they cannot scrape it. Licensing it does not mean handing it over. Here is how it works with us.
What you can license
Records of how your teams work: ledgers and journal entries, close workpapers and reconciliations, tickets and approval queues, and the email threads where exceptions get resolved. We start with finance and accounting.
What never leaves
- Identifiable personal data. It is removed or replaced with labels before release.
- Client-owned material, unless the client agrees in writing.
- Anything you leave out. You choose the workflows, systems, teams and time window.
Three ways to license
- Non-exclusive term. One dataset licensed to more than one named buyer for a fixed term. The most common start.
- Exclusive term. One buyer holds the dataset alone for the term. For records nobody else has.
- Ongoing feed. New records from a live workflow on a schedule, with your approval on each release.
In every model you keep ownership, you can exclude named buyers and competitors, you have audit rights over how buyers use the data, and buyers delete it at the end of the term.
How a deal runs
Six steps: tell us what you hold, sign a mutual NDA, we run the rights check, you approve a small de-identified pilot sample, your counsel agrees the license, and you are paid as the license says. You can stop at any step before the license is signed.
How price is set
Per dataset, after the rights check and the pilot. It depends on how rare and useful the records are, the volume and whether the license is exclusive. We do not publish ranges until we have closed deals to base them on.
See the data owners page to check whether your records qualify.