The Challenge
Travel agents were losing time sifting through hundreds of near-identical hotel listings for corporate clients. Search returned volume, not relevance — and agents couldn't tell why one property ranked above another, which eroded trust in the results.
Business Opportunity
An AI-assisted discovery layer could shorten time-to-booking for agents, improve match quality for corporate policies, and differentiate HNext from competitors still relying on manual filter gymnastics.
The Approach
I designed the experience around agent trust, not novelty. The AI narrows results based on corporate policy, location intent, and historical booking patterns — but always shows why a hotel surfaced, so agents can explain recommendations to clients on live calls.
- Mapped agent search tasks through contextual inquiry and call shadowing
- Defined ranking signals that map to real booking criteria — negotiated rates, policy fit, distance, amenities
- Designed explainable result cards with visible match reasons, not black-box scores
- Built fallback paths so agents can override AI suggestions without breaking flow
- Partnered with engineering on progressive rollout — AI suggestions as an opt-in layer before default ranking
Key design decisions
- Transparency over magic: Match badges ("Corporate rate", "Policy fit", "Near meeting location") instead of opaque relevance percentages
- Agent control: Easy pivot back to manual search when the AI miss fires — no dead ends
- Speed for power users: Keyboard-friendly result scanning and one-click add to comparison tray
Outcomes
Agents reported faster shortlisting on complex corporate requests. Support questions shifted from "where are my results" to requests to extend AI ranking to additional search contexts — a signal the pattern was trusted enough to depend on.