Skip to content

Case Studies

Seven problems, and what actually shipped.

Every case study below is reconstructed from the real artifact — the PRD, the Linear project, the ticket history. Same four questions each time: what was broken, how it was designed, how it was sequenced, what changed.

In flightTPO / SM Assessment · Sep 2026

AI-Powered Food Order Support Assistant

Support agents were drowning in manual lookups across five systems for every ticket. Delivered a human-in-the-loop AI assistant — context aggregation, PII masking, grounded 3-sentence summaries, categorized next-best-actions — with agents always owning the final decision.

−30%
AHT reduction
+10 pts
FCR improvement
  • Enterprise GPT API (zero-retention tier)
  • React / Vue SPA agent portal
  • REST integrations (5 systems)
  • OAuth 2.0 + RBAC