Infor's Bet: Hotels Will Build Their Own AI, and the Real Rivals Are Google and Anthropic
Infor's David Poprawka argues hotels will assemble their own AI agents atop existing systems rather than buy generic products — but Mews's semantic-layer warning and Oracle's own 50-agent build show the risk in doing it without shared data.
Hotels will increasingly build their own AI rather than buy it off the shelf — and the vendors they should worry about aren’t other hospitality-tech companies, they’re Google and Anthropic. That’s the argument David Poprawka, Infor’s Hospitality Innovation Strategist, made in a HITEC 2026 interview, positioning Infor Portico Hospitality AI not as a chatbot but as a connectivity and orchestration layer sitting above a property’s PMS, POS, and RMS — letting operators assemble custom agents on their own data rather than wait for a vendor’s roadmap. He demoed the case live: a predictive-housekeeping agent using Flight Radar data to reprioritize room prep around guest arrival delays, and a $35 million revenue opportunity at a tribal casino from real-time VIP identification.
The “build vs. buy” framing only holds up if the underlying data actually supports it, which is where Mews’s own argument about siloed agents gets uncomfortable. In a recent post introducing its Semantic Layer, Mews describes a scenario where a revenue agent quotes a strong rate for a Tuesday and a sales agent books it — both individually correct — while a group hold that neither agent could see had already claimed the inventory. Without a shared semantic layer modeling what a reservation actually is, hotels assembling their own agent fleet risk building exactly the contradiction-prone stack Poprawka is pitching as the alternative to legacy vendors.
Oracle’s own HITEC interview is the tell that even the incumbents agree with the direction, if not the framing. Tanya Pratt, Oracle Hospitality’s VP of OPERA Cloud Strategy, says her team has already shipped roughly 50 AI agents into OPERA Cloud — but admits agent supply is currently outrunning operator demand, and that AI’s ROI case isn’t proven the way cloud’s total-cost savings once were. Infor is betting operators will want to own that agent layer themselves rather than wait for Oracle’s roadmap to catch up to their specific workflow. Whether that bet pays off depends on whether hotels can stitch their own agents together without the semantic-layer failure Mews is warning about.
Source: Hospitality Net — Infor's David Poprawka interview Auto-generated brief — verified before publishing.