Guide6 min read
Which kinds of company data are worth the most?
A look at the six categories of company data AI labs license, from codebases and tickets to documents and ledgers, and what makes each one valuable.
Not all company data is equal in the eyes of an AI lab. Here is how the six categories we license compare, and what pushes each one toward the top of the price range.
Codebases
Codebases are our highest-demand category. Repositories with full history, including commits, review comments, failed attempts and the issues that triggered the work, are some of the best available material for training coding agents. Longer history and active review culture both raise value.
Tasks
Tickets and projects captured end to end, from first report to resolution, are the most direct training signal for agents. They show a goal, the steps taken and whether it worked. Systems like Zendesk, Asana, ServiceNow and Linear tend to hold this in clean, structured form.
Documents
Contracts, reports, SOPs and PDFs, especially drafts and redlines, carry domain-specific writing that never touched the public web. Documents from regulated industries such as healthcare, finance, legal and insurance earn a premium.
Messages
Coordination threads in Slack, Teams or email capture the reasoning behind the work in the words of the people doing it. They are priced per de-identified thread, and threads that link to tickets, code or documents are worth more than standalone chatter.
Raw data
Records, logs, invoices and ledgers from tools like Salesforce, HubSpot, QuickBooks and Xero are priced per normalized record. Their value is real-world messiness: the exceptions, corrections and edge cases that synthetic data misses.
Workspaces
A workspace combines several of the above into one connected trail of how work gets done across tools. Because the links between sources are preserved, a complete workspace can be worth up to six figures.
The common thread
Across every category, the same things raise the price: completeness, history, domain depth and links between records. If you are not sure where your data lands, the estimator on our home page gives a quick range, and a free evaluation gives you a real number.