Skip to main content

EP-MEL-11 — AI-Assisted Operations

Companion: Backlog README · EPICS.md · canonical: 07-epics-and-user-stories.md §13

Summary​

WaveR2
PriorityP1
Primary ownerai-orchestrator-service
Participating servicespricing-service, reservation-service, housekeeping-service, maintenance-service, notification-service, billing-service
Journeys realisedJ-11 (AI-suggested operations), J-21 (HITL acceptance)
WorkflowsWF-10
Frontend surfacesElectron Desktop · Mobile · Control Plane
Story count8

Outcome​

AI is a guarded copilot, never an autopilot: dynamic pricing suggestions appear with provenance and confidence; demand forecasting drives availability views; anomaly detection raises alerts on bookings/payments/locks; multi-language guest messages are drafted but human-approved; upsell recommendations surface contextually; tenant-level cost guardrails cap spend; HITL gate is enforced everywhere; edge and cloud paths produce equivalent outputs for offered-both-ways capabilities.

Cross-cutting AC for this epic​

  • All AI requests routed through ai-orchestrator-service; no direct LLM calls from other services.
  • Every AI output carries AIProvenance (model, version, prompt-id, cost, latency, locale).
  • HITL approval mandatory for irreversible actions (price publish, message send, lock action).
  • Per-tenant budget caps enforced in real-time; over-budget calls fail-soft with degraded UX.

Stories​

IDTitle
US-MEL-0091Dynamic pricing suggestions (HITL)
US-MEL-0092Demand forecasting
US-MEL-0093Anomaly detection (bookings, payments, locks)
US-MEL-0094AI-drafted multi-language guest messages
US-MEL-0095Upsell recommendation surfacing
US-MEL-0096AI cost guardrails per tenant
US-MEL-0097HITL gate enforcement
US-MEL-0098Edge-vs-cloud equivalence for offered-both-ways capabilities

Full AC in ../07-epics-and-user-stories.md §13.

Cross-references​