This insight summarises Forget Me Not, by Terence Ronson, Founder and Managing Director of Pertlink Limited, published on Hospitality Net on 2 October 2026.
The problem has a name. Twilio's 2026 Customer Insights Series for Asia-Pacific (7,652 consumers and 660 business leaders) found that 70% of consumers abandoned an AI interaction because it lost context, and 65% had to repeat themselves after an AI-to-human handoff. Yet 84% of brands believe their AI recognises returning customers. For hotels the stakes are higher: "A bank that forgets you is irritating; a hotel that forgets you is failing at the one thing it sells."
Five kinds of amnesia. Ronson names session, channel, handoff, stay and portfolio amnesia: from a chatbot asking for a booking reference it was given four messages ago, to a guest who is loyal at one property and a stranger at a sister hotel.
A symptom, not the disease. The h2c 2026 AI Opportunity Study (113 chains, 8,200+ properties) shows 91% of chains use AI, but only 28% have an enterprise-wide strategy and 13% report measurable ROI. 69% still enter guest preferences manually, only 42% have a central guest profile database in use, and 60% of AI use runs through external tools. PMS platforms that create a new profile per booking make it worse. As Ronson puts it: "The memory exists; it cannot travel."
The six-move playbook:
- Resolve the guest before trying to remember them: merge duplicates and match OTA-masked bookings.
- Make every conversation write back structured preferences to the guest record.
- Build a governed Guest Memory Layer, not another silo, with mandatory vendor write-back.
- Standardise the warm handoff with a full context packet for staff.
- Disclose, consent and show the memory so guests can view, correct and delete it.
- Measure the forgetting with Repeat-Yourself Rate (RYR) and Manual Touches Per Stay (MTPS).
Why this matters for hotels. Recognition is hospitality's core product, and AI layered on fragmented data only forgets faster. Ronson's principle, "Remember centrally, reveal minimally," gives operators a way to personalise without becoming intrusive.