Data type
License your project management records for AI training
Project management records can often be licensed for AI training because they describe work, not customers. Outrich Data is a direct buyer and pays for a licensed copy of task histories, estimates, dependencies and status changes. Assignee names are replaced with roles, client projects are excluded unless the client consents, and you choose which workspaces are in scope.
What counts as project management data
Project management data is the plan and the diary of a piece of work. A project is broken into epics, tasks and subtasks. Each has a description, an owner, an estimate, a due date, dependencies and a status that changes as the work moves. Comments, attachments, sprint assignments and time logs fill in the rest.
The history is as important as the current state. A task that was estimated at two days, blocked for a week, reopened after review and closed a month late tells a truer story than the finished board. Projects run for a client inside your tool, or inside the client's own workspace, are generally the client's confidential work and are excluded.
Why AI labs want it
Planning is a skill models are asked to perform and rarely get to observe. A project archive shows how experienced people split a goal into tasks, what they thought each would take and what happened instead. AI labs use plan and outcome pairs to train models to decompose work and to estimate with some realism.
Long-running agents face the same problems a project team does: dependencies, blocked steps, changing scope and rework. Task histories with status transitions and comments show how humans noticed a blocker and replanned around it. Labs use them to build evaluations where a model's plan is compared with the path that was really taken.
Removed before licensing
What comes out first.
- Assignee, reporter and commenter names and @mentions, replaced with role labels
- Client names, project code names and customer references in titles and descriptions
- Attachments, pasted screenshots and linked files
- Credentials, tokens and internal URLs pasted into comments
- Per-person time logs and velocity figures that could single out an individual
- Security vulnerability tickets for systems still in use
- HR, legal, finance and acquisition projects, excluded in full
Data title search
What gets checked.
- Client work. Statements of work and consulting agreements usually treat project details as client confidential information and assign deliverables to the client.
- Employee data. Task histories show who did what and how fast. Check employee notices and, for European staff, GDPR and works council requirements with counsel.
- External collaborators. Guests such as contractors, vendors and customers who commented in your workspace did not accept your internal policies.
- Linked content. A ticket may point to code, designs or documents that carry their own restrictions. Linked items are assessed under their own type.
- Government and regulated projects. Public sector and defense work can carry contract clauses and export rules that bar disclosure.
Who signsThe COO, head of engineering or head of the project office sponsors it. The general counsel approves the license and the workspace administrator selects projects and runs the export.
Value
What moves the price of project management data.
| Factor | Effect on value |
|---|---|
| Full change history | Workspaces that retain every status, estimate and assignee change are generally worth more than a snapshot of closed tasks. |
| Estimate and actual pairs | Tasks with an original estimate and recorded effort give a direct target for estimation models. |
| Descriptive tasks | Clear descriptions and acceptance criteria teach more than one-line titles. |
| Dependency structure | Linked and blocked tasks show sequencing that flat lists miss. |
| Type of work | Engineering, construction, implementation and compliance projects each bring structure that general to-do lists lack. |
Industries where this is common
Questions
Licensing project management data: what companies ask.
Our boards are messy and half the tasks were never closed. Does that matter?
We deliver projects for clients. Are those boards ours?
Could this expose how productive individual employees are?
Is this the same as licensing our source code?
Find out if your data qualifies.
Eleven questions, about five minutes, no call required. You get an answer on screen and a written follow-up.