Skip to content
Outrich Data

Data type

License your workflow logs for AI training

Workflow logs record how work moves through your systems, step by step. They can often be licensed for AI training because they contain little personal content once IDs are scrubbed. Outrich Data reviews vendor and customer terms, replaces user and customer identifiers, and pays you for a licensed copy that you approve in writing.

What counts as workflow logs

Workflow logs are the event records your business systems write as people and software complete tasks. An order is entered, checked, approved, picked, shipped and invoiced, and each step leaves a timestamped row naming the action, the status and the role that performed it. The same applies to claims, purchase orders, service jobs, loan files and IT changes.

They take several forms: audit trails and status histories in an ERP or CRM, run logs from automation tools, clickstream and UI event data from internal applications, and screen recordings of staff completing a process. Logs generated inside a system you operate for a client belong to that client and are excluded.

Why AI labs want it

AI labs are building models that carry out multi-step tasks in software, not only answer questions. To learn that, a model needs examples of real sequences: which screen came next, which field was changed, where an approval was required and what happened when something failed. That information is almost absent from public data because it only exists inside company systems.

Logs also record exceptions. The order that was put on hold, the invoice that failed matching and the ticket that bounced between three teams show how a process behaves under stress. Labs use complete traces with a known end state to train agents and to build evaluations where success can be checked against what really happened.

Removed before licensing

What comes out first.

  • Employee user IDs, usernames and workstation names, replaced with role labels
  • Customer, patient and supplier identifiers, replaced with consistent placeholders
  • Free-text note and comment fields that contain names or personal details
  • Exact prices, margins and supplier terms where they are commercially sensitive
  • IP addresses, hostnames, session tokens and credentials captured in logs
  • Personal data visible on screen in recordings, by masking fields or cutting segments
  • Per-employee productivity measures that could single out an individual

Data title search

What gets checked.

  • Software vendor terms. Some SaaS and ERP agreements restrict exporting or disclosing screens, schemas or usage data. Check the license and any API terms.
  • Customer contracts. Logs tied to customer orders or cases can fall under DPAs and confidentiality clauses, even when names are removed.
  • Employee monitoring. Activity logs and screen recordings are employee data. Notice requirements and, in some countries, works council rules apply. Review with counsel.
  • Privacy law. Pseudonymized logs are often still personal data under GDPR. CCPA and sector rules such as HIPAA can apply to the underlying records.
  • Trade secrets. Logs reveal how your operation runs. Decide which processes are sensitive enough to exclude or generalize before approving a package.

Who signsThe COO or head of operations sponsors it, since the logs describe core processes. The head of IT confirms vendor terms and runs the export, and the general counsel approves the license.

Value

What moves the price of workflow logs.

FactorEffect on value
End-to-end tracesCases that can be followed from first event to final outcome across systems are generally worth more than fragments from one tool.
Action-level detailLogs that record each screen, field and click are more useful for training software agents than status changes alone.
Exceptions and reworkA healthy share of failed, held and corrected cases makes the data more instructive than a record of only clean runs.
DocumentationA data dictionary, process map or SOP that explains what each event means raises interest and shortens review.
Process complexityLong, branching processes with approvals and handoffs between teams are scarcer than simple linear ones.

Questions

Licensing workflow logs: what companies ask.

Our logs are just rows of status codes. Is that useful to anyone?
It can be. A status history shows the order, timing and branching of a real process, which is what labs building task-completing models are short of. Value rises when the codes come with a data dictionary and the cases can be traced to an outcome.
Do these logs count as personal data?
Often yes, at least before processing. User IDs tie actions to employees and case IDs tie them to customers. Replacing those with placeholders reduces the risk, but pseudonymized data can still be personal data under GDPR. The method is documented for your counsel to assess.
Could a buyer work out our pricing or margins from the logs?
Commercially sensitive values are a must-remove category. Prices, margins and supplier terms can be dropped, bucketed or scaled so the process sequence stays intact while the numbers do not. You approve what is kept in each package.
Does our ERP or SaaS vendor have a say?
Possibly. The event data about your own operations is generally yours, but some vendor agreements restrict sharing screenshots, schemas or exported usage data. The relevant license terms should be checked for each system before export.

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.

esc