Evidence-readiness criteria
Mark each criterion as verified, partial, missing, or not assessed. Verify against a live demo, export, contract, or documentation—not sales copy alone.
Methodology
The score rewards verifiable reporting controls. “Verified” earns the full criterion weight, “Partial” earns half, and “Missing” earns zero. Unassessed criteria earn zero and keep the result explicitly incomplete.
1 · Define the use case
Specify engines, locations, prompt families, users, reporting cadence, and decisions the tool must support.
2 · Request evidence
Inspect a real answer record, export, data dictionary, version history, and governance documentation.
3 · Test controls
Run a small prompt set twice and verify whether context, changes, citations, and errors remain traceable.
4 · Record exceptions
Export the result, attach supporting evidence, and assign an owner and deadline to unresolved gaps.
| Score band | Interpretation | Responsible next step |
|---|---|---|
| 0–39 | Evidence controls are weak or the assessment is substantially incomplete. | Do not use summary scores for consequential decisions; request evidence and close required gaps. |
| 40–69 | The tool may be usable with documented manual controls and reporting caveats. | Run a bounded pilot, preserve independent snapshots, and price the extra QA work. |
| 70–100 | The assessed workflow has comparatively strong evidence readiness. | Validate contract terms, data retention, accuracy, and fit for the actual team before purchase. |