Key takeaways
- AI labs pay for ordinary operating records: tickets, threads, CRM history, SOPs.
- Connected records that show work from start to finish set the price. Volume matters least.
- Published deals land around $110K–$450K. Plan around that, not the headlines.
- It's a license, not a sale. Check your rights with a lawyer before anything leaves the building.
01
Why AI labs are buying company data
AI labs have already trained on most of the public internet. What they're short on is a record of how work actually gets done inside a business: the ticket and the fix, the thread and the decision, the deal from first call to close. That record only exists inside companies, and it's what we help owners license.
Labs put it to work in two ways. Some becomes seed data for simulated workplaces where AI agents practice a job until they're reliable. The rest becomes worked examples of a task done well, start to finish.
The budgets are real. micro1 has committed $1B to buying enterprise data over the next 12 months, and more than 50 companies now sell data and training environments to labs.
"I have still never seen a data contract get turned down by a top lab if it's good quality data for budget reasons."
A data-industry insider, on what labs spend
02
What buyers want
The request we hear from buyers is consistent: the everyday records of a working business. The industry matters less than owners expect. HVAC firms, SaaS companies and manufacturers all qualify. What counts is that the work happened in software, over years, and left a trail.
- Support
- Tickets and how they were resolved
- Sales
- CRM history from lead to closed deal
- Engineering
- Docs, bug trackers, product planning
- Finance
- Approvals, reconciliations, month-end
- Legal
- Contract workflows and reviews
- Operations
- Dispatch, fulfillment and the SOPs behind them
It usually lives in tools you already pay for: Slack or Teams, email, Google Drive, Jira, Asana, ClickUp, Monday or Zendesk.
Where we start
List the apps your team has used and for how many years. That's the first thing we ask for, and it's enough for a buyer to put a number on it. Nothing confidential changes hands until you sign.
See how a data partnership works →03
What determines the price
The first question owners ask us is how much data they need. Volume is actually the factor that matters least. Buyers pay for records they can follow from start to finish: a ticket that links to its fix is worth more than a folder of files nobody can trace.
Connected records
Work you can follow start to finish
Scarcity
Niche industries, unusual tools
Clean rights
Proof you're allowed to license it
Exclusivity
Reportedly worth several times a shared deal
Years of history
Several years across several systems
Volume
Matters least
What raises a dataset's price, most to least.
Outcomes add value too. Did the deal close? Was the ticket solved? A record that shows how the story ended teaches a model more than one that stops halfway.
04
What deals actually pay
micro1 publishes three tiers for what it pays a company to license its data:
Qualified companies licensing their operating records
Large datasets, or ongoing work across several teams
Highly unique data with real value to frontier AI
We tell owners to plan around the published deals, which have landed between about $110K and $450K. Online estimates for similar companies can differ by 20×, so treat any single number with care.
Your archive
One payment
A snapshot of the records you already have, licensed once.
Your workflows
Paid every month
Your team works with frontier AI tools and each finished task pays, around $900 a task.
Our advice
Get the offer in writing and don't let a 48-hour deadline rush you. Exploding offers are common in this market, and a serious buyer will still be there after your lawyer reads the contract.
05
What to check before you sign
Before anyone outside your company sees the data, personal details are swapped for consistent stand-ins, so conversations still make sense. You don't do the scrubbing yourself. Toggle the example to see it:
Anonymizing is the easy part. These are the four things we have every owner check with a lawyer:
- Your right to license itYour CRM holds customers' personal details, and email holds other companies' confidential material.
- App termsSlack, HubSpot and Zendesk forbid AI training through their APIs. Native admin exports are the usual route.
- SamplesA sample is a disclosure. Don't send one before legal review.
- Deletion termsDeletion covers the dataset, not what a model already learned. Get the terms in writing anyway.
You keep ownership. An operating company licenses its data rather than selling it, and you can leave out categories like HR or health records, though excluding more can lower the price.
06
How the process works
From first form to payment usually takes a few weeks. Until there's a signed contract, you only describe what you have.
Apply 2 minutes
A short form about your company. Nothing confidential.
Intro call & data manifest
Which apps, how many years, roughly how much. No files change hands.
Priced offer 1–2 days
A specific number based on quality, uniqueness and value.
Contract and transfer 1–7 days
A licensing agreement, then an encrypted transfer you can watch.
Payment within 30 days
Paid in full after the data is reviewed and accepted.
Bottom line
Prices are high now, and won't stay that way
We expect prices to level off as more supply comes online, possibly within a year. If your company has years of connected records, now is the time to find out what they're worth, and our data partners will walk you through it and take care of everything. If you know owners who fit, you can earn $50K or more by introducing them. Just click Become a referral partner.
Not legal, tax or financial advice. Figures come from micro1's published pages and public reporting as of October 2026.