Your company's data is worth money to AI labs.
Labs pay $100,000 to $2 million USD for operational data like tickets, chat and CRM history. We find the buyer. Personal information and anything you mark confidential are removed on terms agreed in the contract. The whole process takes 3 weeks to 60 days, depending on complexity.
A support thread as your team wrote it, and as the buyer receives it
My background is in data. I worked with America's Fortune 500 companies, including a global enterprise software company and a national defense contractor, as a data analyst, data engineer and software engineer, and trained in data science. I'm also a Canadian military veteran: a commissioned officer who specialized in data analysis and reporting. I understand data from both sides of the table, so I can speak your language, whether you're selling or buying.
We're your trusted advisors, from first call to close. If you're selling, we work to get you the best price and handle your data with security and privacy in mind. If you're buying, we find quality data that fits your AI lab's requirements and see it through to delivery, so your team can focus on advancing artificial and superintelligence research.
- VC-backed software startup~250 staff at peak. Full archive, ~3.7 TB over 3+ years: email, Drive and Slack, 65 private repos (6.4M lines of code), 13.7K tickets and ~300M product and telemetry records.
- Creator marketplace55 staff, 2021 to 2026. ~2.1 TB: 96 repos, 313K CRM contacts, ~83K sales-call recordings, 57K support tickets, ~200K Drive files, 120K+ Slack messages and 58M analytics events.
- Engineering RL environments12K+ containerised, auto-graded tasks: 3,400 CAD and PCB design, 9,400 chip design and 1,500 Lean proofs.
- Industrial parts manufacturer10 staff, 20+ years in business. ~500 GB: 11 years of CRM, accounting, inventory and orders, CAD drawings of client orders and contract email.
- Scientific-research RL environmentAgents write hypotheses, experiment designs and NIH-style grant sections, then answer peer review, scored on rubrics. 400+ eval tasks across 7 tracks.
- Freight brokerage70 staff, 9 years on one TMS. ~800 GB: dispatch chat, rate quotes, load records, invoices and claims email, with outcomes recorded.
- Accounting firm40 staff, 12 years. ~400 GB: client email threads, review notes, reconciliation workpapers and a full ticket history.
- SaaS support team5 years on one helpdesk. ~1.2M resolved tickets with agent replies, escalations and QA scores.
I represent my company. We have data to sell or license.
Owner, executive or authorized representative of a company with 30 or more people.
- Fit
- 30+ people, 3+ years on the same tools, data in English
- You send
- Counts and years per system. No files
- You get
- A straight yes or no, then a quote with the amount
- Your data
- Stays put until you agree. Then it is anonymized and personal information is removed
I want to sell how my company works.
Documented processes, recorded task runs or AI-assisted tasks, at a company with 30 or more people.
- Fit
- 30+ people, work that repeats every week and leaves a written record
- You send
- Answers about your workflows and departments. No data
- You get
- A straight yes or no, then a quote per workflow, by department if you prefer
- Capture
- Agreed case by case, and only after you agree. Personal information is removed and people review it
I am buying data for a lab or model team.
LLMs, agents, evaluations, non-LLM models and superintelligence research.
- Sellers
- Many, across industries, company sizes and departments, ready to license
- Data
- Every type of enterprise data, plus workflows, to train LLM and non-LLM models
- Rights
- Checked first. Provenance memo with every package, anonymized and reviewed by people
- Terms
- Exclusive or not, one-off or ongoing
How a deal runs
Two sides, two paths. Sellers license their company's data out; buyers license it in.
If you are selling or licensing your company's data
Tell us what you have, then a short discovery call
Ten-minute form, then thirty minutes on how your data is used.
Rights and fit check
Client-held material out. Contracts read. Matched to this month's demand.
Labs quote, then you negotiate
We send your summary, never your data, to several AI labs and bring back the best quote and conditions for your needs.
De-identification
Names become employee 632. Automated tooling plus human review. Who does it, and raw-file handling, agreed case by case.
You get paid
Buyer accepts, then pays on the written terms. Never perpetual.
If you want to sell how your company works
Tell us about your workflows
A short form on the workflows, departments and tools. No data.
Fit call
Which workflows suit AI labs, and in what form: documented, recorded or AI-assisted. Agreed case by case.
Labs quote, then you negotiate
We send your summary, never your work, to several AI labs and bring back the best quote per workflow.
Capture and anonymize
Only after you agree. Automated tooling plus human review remove personal information, and you review a sample.
You get paid
Buyer accepts, then pays on the written terms. Never perpetual.
If you are buying data for a lab or model team
Tell us what you need
Data types, industries, volume, exclusivity. No cost to ask.
We source and clear rights
Matched to sellers. Ownership and customer contracts checked first.
Review and send a quote
One-page summary and de-identified samples. We take the best quote to the seller.
Sign, then receive the package
Terms in writing. The package comes with a provenance memo; raw-file handling is set in the contract.
Accept, then pay
Written acceptance window first. Fees are discussed with you up front.
Latest from the blog
All articles- For sellers3 min read
Can you license your support tickets to AI labs?
Learn if you can sell support tickets, what requirements apply, and how to start the licensing process.
- For buyers3 min read
How to evaluate a company dataset before buying
Learn the three things you must check, relevance, rights, and context, to decide if a corporate data set is worth licensing.
- For buyers3 min read
How to source licensed operational data for AI
Learn what operational data looks like, why labs need it, and how to tell us your needs so we can match you with the right seller.
- For buyers4 min read
How to secure the right operational data for AI labs
Learn the steps to get de-identified operational data licensed for AI training, from discovery to delivery.
Latest videos
All videosWhat Are RL Environments? Explained for Business Owners
Questions sellers ask
What company data do AI labs buy?
Operational data from real work: support tickets with resolutions, chat threads where a decision got made, CRM and sales activity, SOPs and internal documentation, knowledge bases, QA records, project histories, accounting exports. Labs pay for the chain from request to outcome. An email archive on its own is worth less than most owners expect.
How much is it worth?
Buyers publish ranges, not averages. Most approved deals land between $100,000 and $2 million USD, and buyers who mention a typical deal put it around $200,000 to $300,000 USD. What moves the number: how much data you generate, the type of data, years of continuous history on the same tools, visible reasoning, a rare industry, clean rights.
Can we sell how we work, not just data?
In some cases, yes. Documented processes, recorded task runs and AI-assisted tasks can be licensed too, priced per workflow and by department if you prefer. See the workflows page.
Are we selling our data?
No. You license it for a fee and a fixed term. Your company keeps ownership and keeps using its data. Never perpetual.
Is it legal?
Licensing your own operating data is an ordinary commercial deal. Three checks decide a specific dataset: you own it, your customer contracts allow de-identified use, and personally identifiable information (PII) can be removed. All three are done before a buyer sees anything, and every license goes through counsel.
Who sees the raw data?
That is agreed case by case in the contract, depending on the negotiation. All personally identifiable information (names, emails, phones, accounts) is replaced with stable pseudonyms like employee 632 using automated tooling, and trained human reviewers check the result. Anything your own data policy marks confidential is removed too, and how the raw files are handled and deleted is written into the contract. You review a sample before anything is accepted. The broker sees summaries and samples only.
What does it cost?
Nothing up front. The buyer pays a finder's fee when a deal closes. Preparing, presenting and negotiating on your behalf is an advisory fee, owed by you only when the deal closes and agreed in writing first. No deal, no fee.
Exclusive or perpetual?
Never perpetual. The usual buyer ask is 24 months of exclusivity. A shorter term is possible at a lower price. If you plan to sell the company, the term is timed to your exit.
Buying data? Questions buyers ask
Find out what your company's data is worth
Asking costs nothing and commits you to nothing. Nothing moves until you sign.
Buying data instead? Tell us what you need