Data that qualifies
Ordinary records, already sitting in your software
Email, drives, chat, help desk and more. Choose system by system, starting with just one if you like.
See my estimateThe eight record types
Inboxes in Microsoft 365 or Google Workspace, with years of back and forth.
Files and drives
Procedures, quotes, plans and spreadsheets saved in shared folders.
Chat
Conversations in Slack or Teams that end in an answer or a decision.
CRM
Notes on accounts, moves through the pipeline and follow ups.
Help desk
Support tickets traced from opening report to resolution.
Project tools
Real project work captured as tasks, boards and status changes.
ERP and finance
Workflows for orders, purchasing and invoices.
Code
For software teams, repositories and their code reviews.
Why everyday records
AI labs license real business records to train and test models. From public text, a model learns how people write. Business records teach how people work: who asks whom, what is checked, what order steps take and how a problem gets closed.
So the plain, untidy, day to day records are exactly what matters. A thread where a customer complains, a colleague digs into the order and a manager signs off on a credit is exactly the kind of sequence labs cannot find in public.
What adds weight to your records
- More history. Your years in business is one of two answers that shift the estimate.
- More people. Team size is the second answer. A larger team in shared systems leaves a fuller record.
- Continuity. Records that stayed in one system for years, rather than being scattered or deleted.
- Variety. A few linked systems, for example email with a CRM and a help desk.
Leaving records out
Anything. A department, a mailbox, a date range, a shared drive or a whole system can be left out during scoping. Sensitive data in the records you do include is excluded or de-identified before anything is shared.
Law, accounting and healthcare practices keep client and patient records out, with scope decided together with you and your counsel.
Not sure what you have?
That is common. Starting does not require any inventory. There are only three questions in the estimate. After that, a specialist from License Your Data calls within minutes to book a 15-minute call, under NDA, where your systems get mapped. Nobody asks for a password or a file, and you decide afterwards what goes into the de-identified copy.