The market for company training data
AI labs and the vendors that supply them pay companies for real examples of work: tickets, SOPs, project history, CRM activity. We publish research on how this market works and, if your company looks like a fit, introduce you to a buyer. The buyer's process and your counsel run the deal. We never receive your data.
Market news
Recent deals in the market, as reported in the press.
Gizmodo, reporting on Forbes April 17, 2026
SimpleClosure reported payouts of $10,000 to $100,000 per company.
Read the reportTechCrunch, reporting on Wired January 10, 2026
An IP lawyer warned that it left individual contractors to decide what was confidential.
Read the reportCBS News February 23, 2024
Announced in February 2024.
Read the reportVariety May 2024
Announced in May 2024.
Read the reportHow deals go well
Buyers pay for records that show work being done, cleaned under a written process. The deals that go badly usually skip one of those steps.
What tends to work
What tends to go wrong
How it works
We qualify and introduce. The buyer brings the NDA, the data process, the license and the payment. Your counsel reviews their paper.
Answer about ten questions about your company and the records you keep.
Your partTen questions
If you look like a fit, we tell you which buyer and how we're paid. Nothing happens until you say yes.
Your partA yes or no
One introduction by email. Your company's name goes to that buyer only.
Your partOne email
The buyer sends its own NDA, data requirements and license. Your counsel reviews them.
Your partYour counsel
What buyers pay for
The most valuable records follow a task from start to finish, with the decisions along the way. Chat and email help, but on their own they're harder to place.
What buyers usually require
What buyers look forOur role
Hawthorn Data is a small introducer. We publish research and make introductions. Buyers pay us if a deal closes.
We do
We don't
Whoever you license to, these terms protect your customers, your staff and you. The full checklist
01
A schedule listing exactly which systems, record types and date ranges are licensed.
02
Personal and confidential details removed under a stated standard, with samples you approve.
03
A ban on trying to identify any person or company from the data.
04
Your records go to the buyer named in the license and nobody else.
05
Raw and cleaned copies deleted or returned, confirmed in writing.
06
Warranties limited to your right to license, and a cap tied to the price.
A small introducer. We publish research on this market and introduce companies that qualify to buyers. We are not a law firm, broker-dealer or data processor, and we never receive your data.
No. We only need your answers to the assessment. If you go ahead with a buyer, any data moves between you and the buyer under their contract with you. We never receive files, exports or credentials.
The buyer pays us a referral fee if a deal closes. You don't pay us anything. Before any introduction, we tell you which buyer it is and how we're paid.
Usually, if the records are yours and your contracts allow it. The things to check are customer contracts and data processing agreements, what you have told employees about how work data is used, and privacy laws like GDPR, CCPA and HIPAA. Your own counsel should review the buyer's terms before anything is shared.
Often, yes. Private repositories with their full history of commits, issues and code reviews are among the records buyers ask for, as long as the code is yours to license and secrets and third-party code are dealt with first.
The buyer makes the offer, and it depends on how much you have, how complete it is and how many other companies could supply the same thing. Publicly reported deals for closed startups have run from about $10,000 to hundreds of thousands of dollars.
Closing companies are some of the most active sellers right now, because the records have no other use. The same rules apply: personal information comes out, customer contracts still bind you, and the terms should be in writing.
Guides
The whole process, from taking stock of your records to signing a license and handing over a cleaned dataset.
Read 02The four kinds of buyers, what each one wants, and which one fits a company like yours.
Read 03Which records are in demand, which are hard to place, and what makes one dataset worth more than another.
Read 04Why AI buyers want private codebases with their history, what makes a repository valuable, and what to strip out before anyone sees it.
Read 05The factors that set the price, with the figures that have been reported publicly.
Read 06The three standard methods (cost, market and income), how each applies to a dataset you might license, and where each one misleads.
ReadAll 11 guides, including contracts, dataset licenses and de-identification
Assessment
Answer about ten questions. If you look like a fit, we'll offer an introduction to a buyer, and you decide. No files, no credentials.