This appendix is the report a sponsor receives at the end of an OG-CAIE evaluation, written for one reader: the executive who runs the sponsoring organization, a public-health office or a non-profit, who knows the domain and little about AI. It answers the question that executive asked, in one sentence, and says who is accountable for the words. Under the answer: what was tested and whom it serves, how it did against each requirement (open a row for the criteria beneath it), what the report rests on, and what to do next. When the evaluation is not complete, because a criterion was not tested or the report is not final, the headline says so.
The case is synthetic (ruling R-23). The people named recognise the roles of real people, but no judgment or signature in the record was made by them, and the chatbot, its vendor and the county are placeholders. The report is what a real engagement would produce; the content is not a real evaluation.
The sample report for the synthetic case. Open it full screen for a larger canvas.
Every line on the page is read from the evaluation record in the repository, and the share of criteria tested is recomputed from it by the coverage query, never copied from the stored report. The page is regenerated by the gate and diffed byte for byte. The full record and every check behind the report, the parties, the plan, the sessions, the shapes that checked the record and the rulings behind the vocabulary, are in Appendix B, which opens the whole model as a knowledge graph.