AI training data
Your company's work data is worth money to AI labs.
AI labs are buying real records of how work gets done: team chat, code, tickets, documents and AI agent sessions. We find out what your data is worth, remove personal and sensitive information, document that you have the right to sell it, and take it to buyers. You approve every buyer and every file.
Sound familiar?
“We have years of Slack, Jira and GitHub history and no idea whether it's worth anything.”
“We're winding down a product and want to recover value from what we built.”
“A data buyer contacted us, and we don't know if the offer is fair or what we'd be signing.”
What we do
Data inventory and valuation
We list what you hold, which parts buyers want, and a realistic price range based on current deals, before you commit to anything.
Rights and consent check
With your counsel, we confirm what you're allowed to sell. Customer data, HR records and anything your contracts restrict stay out.
Personal-data removal
Names, emails, phone numbers, secrets and identifying details are replaced or removed in messages, code and tool outputs, and the result is tested for re-identification.
Documentation buyers require
A datasheet, a de-identification report and a record of where every file came from, so buyers can check what they are licensing.
Competitive sale
We take your data to several buyers, including AI labs, data marketplaces and lab data vendors, and negotiate the licence terms and price.
Custom datasets
For labs: analytics and data-engineering tasks in English, Spanish and Portuguese, each with a seeded warehouse and an automatic check of the answer.
How we work
Data assessment
Free
30-minute call
Tell us what systems you use and roughly how much history you have. We tell you which parts buyers want and whether a sale is worth pursuing.
Preparation and sale
Fixed fee + success fee
Typically 4–8 weeks
Inventory, rights check, personal-data removal, documentation and a competitive sale. The fixed fee is quoted after the assessment; the success fee is a share of the sale price.
Custom datasets for AI labs
Quoted per project
Pilot in 4–6 weeks
Verifiable analytics-engineering tasks (SQL, dbt, semantic layers and data agents) built to a lab's specification, with pass-rate calibration and full documentation.
Why us
We are data engineers first. Our founder delivered analytics platforms for Fortune 500 companies at Datatonic and has trained 16,000+ engineers through his Udemy courses. Cleaning, documenting and moving large volumes of company data safely is the work we already do.
About Dream Hatch Labs →Frequently asked questions
What kinds of company data do AI labs buy?
Mostly records of real work that isn't on the public web: team chat such as Slack or Teams, code repositories, issue trackers such as Jira, shared documents, email, design files and logs of AI agents doing tasks. Labs use it to train and test models on realistic workplace tasks.
How much is company data worth?
It depends on the volume, the field, how unusual the data is and whether the licence is exclusive. Publicly reported deals range from tens of thousands of dollars for a small archive to several million for a large company's data. The free assessment gives you a realistic range for yours.
How do you remove personal information?
We replace names, emails, phone numbers and account identifiers with consistent placeholders, generalise details such as exact dates and small locations, scan code for secrets, and leave out HR and customer records. We then test whether anyone could be re-identified and document the process in a de-identification report for the buyer.
Is it legal to sell our company's data?
Often, but it depends on your contracts, your privacy notices and the countries your data comes from. We check this with your legal counsel before anything leaves your systems, and we never include data you don't have the right to license. This page is not legal advice.
What about our employees?
Employee messages are the most sensitive part of any sale. We recommend telling employees in advance, removing personal and HR content, and excluding private channels and direct messages unless counsel advises otherwise.
Who are the buyers, and do we get a say?
Buyers are AI labs, the data vendors that supply them, and data marketplaces. You approve every buyer and the licence terms, including what the data may be used for and when it must be deleted.
Do you keep a copy of our data?
No. Where possible we process the data inside your own cloud account. Anything we handle is deleted when the sale closes, and we don't resell it.
We're shutting the company down. Is it too late?
No. Companies winding down are some of the most common sellers. Contact us before systems are switched off or accounts are closed, because exports are much harder afterwards.