Results and sensitivity

Review ranking, close calls, influential criteria, low confidence areas, and lightweight sensitivity scenarios.

Workbench guidance

Keep the model simple enough for non-experts while preserving enough structure for a reviewer to understand the decision. The assistant may suggest alternatives, criteria, weights, scores, and report language, but user or group confirmation is required before treating the result as final.

Use warnings as facilitation prompts. A warning does not always mean the model is wrong; it means the decision rationale needs review before publication.