Nonprofit AI field kit
Browser-safe methodPilot evidence log
A bounded experiment needs a visible record of quality, burden, risk, and decisions.
A reviewable record that separates observations from conclusions and keeps failures useful.
- 01
Log the case ID, workflow version, reviewer, date, result, time, corrections, and escalation.
- 02
Record unsupported claims, omissions, harmful framing, inaccessible output, sensitive content, and near misses.
- 03
Note staff preparation, review, correction, documentation, and support burden.
- 04
At the gate date, summarize evidence for stop, revise, continue, or scale and name the owner.