How to get paid for your company's data

This is about company records, not personal data. If you're looking to sell your own phone data or browsing data, this isn't that.

01Why AI companies now pay for business data

AI models learned from the public internet. That taught them to write. It didn't teach them how a business actually runs, because that was never published: how a support team resolves a ticket, how a finance team closes the month, what operations does when a shipment goes wrong. That knowledge lives inside companies.

So AI data companies have started licensing it:

• In October 2026, micro1 announced a $1bn, 12-month commitment to license company data.
• In August, Google won a bankruptcy auction for the internal business data of Spirit Airlines with a $10M bid: about 100 million emails and 500 million Teams messages. The sale still awaits court approval.

A company that's trading can do this on its own terms, with its own scope.

02What kind of company data qualifies?

The programmes want the working record of how your business operates. Usually straightforward:

• SOPs, playbooks and checklists;
• internal wikis and knowledge bases;
• project histories and ticket workflows;
• QA processes and review notes;
• training and onboarding material.

Needs more care (often included, but with de-identification and legal review):

• CRM and customer records;
• email, Slack or Teams history;
• anything covered by client NDAs;
• personal data about staff or customers (UK GDPR, CCPA).

It doesn't have to be glamorous. A plumbing firm's call-triage SOP is exactly the kind of record AI hasn't seen.

03How much can a company get?

No one can tell you your number until a programme reviews your data. These are the published figures:

All three say price depends on factors like data volume, years of history, quality and uniqueness. These are ranges, not offers. Our programmes comparison sets out the rest of each programme's terms side by side.

04Who qualifies?

Each programme sets its own rules. In broad terms they look for:

• established companies, typically 30+ staff. Mode accepts law firms from 6 and accounting, finance and medical firms from 10, and Mercor lists some business types from 5;
• several years of operational records;
• documentation mainly in English;
• the US first, with the UK and Canada also in demand.

05How the process works

1. Check fit. See which programmes you're likely to match.
2. Apply directly to the programme. It's free to apply.
3. Review. The programme assesses your systems and records, and may hold a call or an interview.
4. Terms. Agree scope, de-identification, retention and use. Have your own lawyer review the licence.
5. Transfer the agreed data.
6. Approval and payment. For example, micro1 says it approves data within 30 days and pays within 30 days of invoicing.

06Risks to check first

• Rights: do client contracts restrict how you can use the data?
• Personal data: what needs removing or anonymising under UK GDPR or CCPA?
• Confidentiality: exclude privileged, regulated or client-confidential material.
• Employment terms: is work product owned by the company?
• Exit plans: a licence may need disclosing in due diligence.

See our guide: Is it legal to sell company data for AI?

07How to choose a programme

Compare the published ranges, the eligibility rules, what each does with the data (ownership, de-identification, retention), and how payment works. Many companies can apply to more than one. See micro1 vs Mode.