Zest Launches SMS-Based AI Concierge That Books Restaurant Reservations

Restaurant discovery startup Zest has introduced a personal AI concierge that lives entirely inside SMS threads, letting users discover and book tables without ever opening a standalone app. The agent draws on each user's actual dining history — imported via Plaid from credit-card transactions — to tailor suggestions to proven preferences rather than generic reviews.

Users text the concierge at a dedicated number, specify cuisine, occasion, dietary needs, or neighborhood, and receive curated options with real-time availability. The agent can also be added to group chats, where it merges everyone's taste profiles, runs an in-thread vote, and secures a reservation for the entire party through integrations with OpenTable, Resy, and SevenRooms.

Why it matters for GPU / AI infrastructure

Zest's shift from app-first to conversation-first architecture highlights the growing inference demand for always-on, low-latency language models that must orchestrate multiple external APIs (Plaid, booking platforms, messaging gateways) in a single user turn. Scaling such agents cost-effectively requires GPU clusters optimized for bursty, multi-step reasoning workloads — precisely the infrastructure profile AiGpu cloud is built to serve.