AI-Powered PNR Intelligence Layer

Extended an existing GDS-to-CRM automation pipeline with a Claude API enrichment layer that parses raw PNR data, generates personalised client trip summaries in the consultant's house voice, and drafts internal briefing notes that previously took 15 to 20 minutes per booking to write by hand.

Implemented a rules-plus-LLM anomaly detector that flags tight connections, missing segments, and ticketing-time-limit risks before the consultant ever sees the booking, so the human review pass starts from a triaged list rather than a raw record.

Designed the layer to fail safely into the unenriched pipeline: if the LLM call errors or returns a malformed structure, the booking still lands in Zoho CRM with the original PNR intact and a flag asking for manual enrichment. Reliability mattered more than coverage.