The three buckets
| Bucket | Examples | Typically |
|---|---|---|
| Your company's know-how | SOPs, playbooks, templates, internal wiki, training decks, QA rubrics | Licensable |
| Operational records with personal info | Tickets, CRM notes, emails, Slack | Licensable after PII removal, if contracts allow |
| Data held for clients or regulated | Client files, patient records, payment card data, data under client NDAs | Usually excluded |
Special case: agencies, BPOs and service firms
If your team works inside clients' systems, as a BPO, agency, outsourced accounting firm or property manager does, the records often belong to the client. Your own training materials, QA frameworks and internal SOPs usually don't. License those.
Quick check before you include a dataset
- Did we create it, or did a client give it to us?
- Does any contract or NDA restrict sharing it?
- Does it contain personal data, and can it be removed?
- Is it regulated (health, financial account, children's data)?
- Would the people in it reasonably expect this use?
Next: prepare your data.
Frequently asked questions
Can I sell my clients' data to AI companies?
Usually not without permission. Data you hold for clients is typically governed by your contracts with them. License your own process documentation instead.
Can a BPO or agency license data to AI?
Yes, its own SOPs, training materials and QA frameworks. Records created inside client systems usually belong to the client.