#payment

Design AI agent → search → offer → reprice → confirmation → payment → booking → servicing with travel-specific authorization, idempotency and audit boundaries.

Prevent duplicate booking and payment caused by agent loops, retries and tool invocation using idempotency, side-effect guards and UNKNOWN-aware recovery.

Model payment and supplier booking as a distributed transaction using authorization, capture, compensation, UNKNOWN outcomes and reconciliation.

Diagnose payment-booking divergence when payment succeeds while booking remains unknown or fails, then choose safe recovery and compensation.

Diagnose divergence between booking cancellation and payment refund state across amount, currency, partial refunds and provider evidence.

Design Saga and compensation patterns for travel booking, payment, cancellation and refund workflows where rollback is impossible.

Design deterministic reconciliation across booking, payment, cancellation and supplier events with explicit authority and audit rules.