Abstract - In early 2026, the author used Gemini Advanced to plan family visits to Reykjavik and Seville — generating complete dining itineraries through conversational AI, without a single search engine query. This paper argues that AI-mediated restaurant discovery represents a structural shift from a search problem to an inference problem, and that the infrastructure serving inference at commercial scale does not yet exist. We identify the Concierge Agent — a city-level AI combining operational intelligence, occasion matching, booking execution, and stablecoin-settled micropayments — as the venture-relevant opportunity in the emerging agentic discovery economy. We present a six-layer technical architecture, per-cover unit economics, competitive landscape analysis, and principal failure modes. The 2026–2030 window represents a credible founding opportunity for a team combining hospitality operational expertise with infrastructure engineering capability.

About this paper - This paper was written as part of the Wharton AI for Business Programme and draws on 20 years of multi-site hospitality operations experience across Chipotle, Wingstop, and Burger & Lobster. It is intended as a resource for founders, investors, and operators building toward the 2030 agentic commerce landscape. Feedback and correspondence welcome at mike@michaelbaglieri.co.uk