[Enhancement] mobieusAI: No feedback signal on AI answers (no thumbs up/down) — quality can't be measured or improved

Area: mobieusAI (audit p11) · Surface: mobieusAI Q&A drawer + helpdesk/forum AI outputs · Dimension: existing-improve · Severity: minor

Intercom (CSAT on Fin answers), Zendesk, and Help Scout all capture a helpful/not-helpful signal on every AI reply, which feeds answer quality dashboards and tuning. mobieusAI captures spend and error rate but zero quality signal, so a tenant admin cannot tell whether the AI is actually helping members or quietly giving bad answers. Without this, the 'KB gap detection' and config-recommender features are flying blind.

Evidence

platform/templates/partials/mobieus-ai-drawer.php:198-225 — the only per-answer control rendered is a Copy button (addCopyButton). There is no rate/helpful control. A grep for `thumbs|helpful|rating` against src/Services/AI and src/Models returns only unrelated PostReaction/TradeFeedback hits — no AI-answer feedback table or endpoint exists. ai_feature_usage records cost/status (AdminAIController.php:337-343) but no human quality signal.

Suggested fix. Add a thumbs-up/down (and optional 'report bad answer') control to the drawer and to forum/helpdesk AI outputs, persist to an ai_answer_feedback table keyed by conversation/message, and surface a helpfulness rate on /admin/ai alongside the existing spend chart.

Filed by the automated tenant-app audit and adversarially evidence-verified. Status: verified. Open — not yet actioned.