Probative LabsIs your AI's result typical?
One company's result cannot say whether it is normal, and AI vendors do not publish how their products perform across customers. Benchmarks answer it without anyone sharing data.
Measure
Run an evidence pack or a resolution audit, with Probative Labs Cloud or the open-source engine.
Contribute, if you choose
Share an anonymous summary of the result. You see exactly what it contains before anything leaves.
Compare
Once five organizations have contributed for the same AI product, contributors see their result beside the peer median and middle range.
A contribution contains
- The AI product and the ticket platform
- The measured change and its range, with the evidence grade
- Headline and mix-adjusted resolution time
- Guardrail results: pass, watch or breach
- Share of claimed resolutions that did not hold, with its range
- Size bands, such as "20–49 teams", and the month
It never contains
- Company, program, team or person names
- Ticket text or identifiers
- Costs, prices or ROI in money
- Exact dates or exact volumes
Contributions are never sold. AI vendors cannot buy, preview or suppress benchmark figures. See the independence rules.
Open for contributions
No figures are published yet: no AI product has five contributing organizations. Design partners who contribute see them first.