Zest新AI助手可通过短信整合多人口味、发起餐厅投票并自动完成预订。
TechCrunch AI confirms Zest has launched an AI agent that 'can combine everyone's tastes, put restaurant options to a vote, and make the reservation' — fully operational over SMS. This is among the rare, media-verified examples of a production AI agent completing a real-world transactional loop (not just a demo or API test). Contrast this with ('Basically no one is using AI agents for their DoorDash orders') and (DoorDash reports only 'several hundred' AI-generated orders), and Zest's success highlights three critical factors for vertical agents: low-friction access (no app install), bounded task scope (meal booking), and trusted action execution (live reservation). For founders, this validates the 'narrow-scenario, deep-closed-loop' strategy over generic agent ambitions. Prioritize domains with strong payment or fulfillment infrastructure — and where users will pay per task (e.g., $2.99 per booking) — rather than chasing broad 'agentic intelligence' without monetizable outcomes.