Data licensing guide
Data licensing guide

What data do AI companies buy from businesses?

By Ade Molajo · Updated · How we check figures

Quick answer

Records that show how real work gets done. micro1 asks for SOPs, knowledge bases, project histories, ticket workflows, CRM records and code. Mode buys email, documents, chat, call transcripts, CRM and billing records. Mercor lists what it wants per business type: closed legal matters, tax workpapers, project drawings, shop-floor records, private code and more. Public marketing material and templates on their own aren't what they want.

How work gets done
is what every buyer describes
3
buyers publishing their lists
24
business types in Mercor's list
Not wanted
templates or marketing on their own

The public internet taught AI models how people write. It didn't show them how businesses actually work: how a ticket gets resolved, how month-end is closed, how a project goes from brief to sign-off. That's what the programmes are paying for.

How licensing works: records from your own systems become an agreed, de-identified copy, which the programme licenses for a fee.01Your systemsSOPs, wikis,tickets, CRM02Agreed scopeWhat goes in, whatstays out03De-identifiedcopyPersonal detailsremoved04Licence feePaid by theprogramme
You license an agreed copy. Your originals stay with you.

01What each buyer says it wants

BuyerIn its own words
micro1SOPs and process documentation, internal knowledge bases and wikis, project histories and ticket workflows, CRM records and sales playbooks, templates and operational frameworks, code
ModeEmail, documents and files, chat and collaboration, calls and transcripts, CRM and sales, billing and operations, from 1,000+ named systems
MercorA list per business type: complete closed-matter files (law), returns and workpapers (accounting), drawings, RFIs and site imagery (construction), production and inspection records (manufacturing), private code with history (software), and more

Each programme's own pages, checked 30–31 October 2026.

02By data type

  • Process documents: SOPs, playbooks, checklists, wikis. Named by micro1 first. See Notion and Confluence.
  • Work records: tickets, project histories, matters, jobs. See Jira and ticketing.
  • Communication: email, chat, call transcripts. Named by Mode. See email and Slack.
  • Customer and sales records: CRM pipelines and support history, people removed. See CRM.
  • Code and technical files: private repos, CAD, test results. See software.
  • Sector files: closed matters, workpapers, clinical notes, project drawings. See the industry pages, such as law firms, healthcare and construction.

03What they don't want

Process documents your team wrote are usually licensable; customer records, chat history, NDA material and personal data need care or stay out.Usually straightforwardSOPs, playbooks and checklistsInternal wikis and knowledge basesProject histories and ticket workflowsQA processes and review notesTraining and onboarding materialNeeds care, or stays outCRM and customer recordsEmail, Slack or Teams historyAnything covered by client NDAsPersonal data about staff or customers
The lowest-risk material is documentation you wrote about how you work.

Mercor's listings are the clearest on this: templates or marketing material on their own, active matters, logs without context, copies of open-source code, AI-written code, and anything client contracts restrict. The common thread is that buyers want the real work, complete and in context.

04Check whether you're likely to match

Our free fit check asks about your size, location, records and tools, then shows which programmes you're likely to match. The programme decides.

05Questions people ask

What kind of data do AI companies want from companies?

Operational records: process documents, tickets, project histories, CRM, email and chat, and sector-specific files like closed legal matters or project drawings.

Do they want customer data?

Mostly the work, not the people. Programmes de-identify data, and customer personal data usually has to be removed.

Sources

Check which programmes you’re likely to matchA few short questions, about a minute. Free, no sign-up. The programme decides who qualifies.trackimpact.io/data →