This insight summarises Why AI pricing still fails hotels — and what needs to change by Tong Yin, founder and CEO of InsightBridge Global, published on PhocusWire in June 2026.
Most "AI revenue management" isn't AI. Yin's blunt opening: many hotel executives believe they've adopted AI in revenue management, when in reality they've adopted automated suggestion systems built on three architectural assumptions that no longer hold.
The three broken assumptions:
- Stable historical demand data — irrelevant in greenfield markets like Saudi Arabia's mega-projects or new Southeast Asian resort destinations, where there is no baseline.
- Clearly defined competitor sets — fragmented markets shift constantly.
- OTA-driven pricing signals — AI-mediated discovery (ChatGPT, Gemini, Perplexity, regional assistants) is breaking the OTA monopoly on demand signal.
Hotels still optimising as if these held may be leaving 8–14% of revenue on the table annually in markets without mature demand curves.
The override signal nobody captures. Revenue managers override system recommendations more than half the time — and the system never learns from it. Override isn't noise; it's tacit market knowledge (a local event, source-market sentiment, a competitive dynamic). Captured and fed back, it compounds: within 12–18 months the engine is genuinely calibrated to its own market. Not captured, it vanishes the moment the revenue manager changes jobs.
The three-layer fix:
- Demand reconstruction from first principles — ingest flight capacity, event calendars, visa changes, AI-platform search behaviour, source-market FX — so the model can price markets that have no past.
- Channel-aware net revenue optimisation — a $1,000 OTA booking ≠ a $1,000 direct booking once commission, cancellation behaviour, and payment costs net out. Stop optimising gross ADR.
- Human-in-the-loop learning — every override becomes a training signal: what did the human see that I missed?
The distribution kicker. Hotels with generic pricing systems default to OTA dependency because their intelligence isn't differentiated enough to justify direct-booking investment. Hotels with adaptive pricing build the intelligence advantage that AI travel agents will reward in recommendations. The next decade of distribution economics goes to human-amplified pricing systems that learn faster than competitors.