Uber高管披露:2026年第三季度,第三方AI代理仅促成4笔订单。
During its Q3 2026 earnings call, Uber confirmed exactly four rides were booked through third-party AI agents — a figure verified across all public API endpoints and excluding internal experiments. This stands in stark contrast to Uber’s reported 2.2 billion total rides that quarter, underscoring the chasm between agent demos and production-grade transactional autonomy. Critical gaps remain: reliable user intent disambiguation, standardized identity and payment handoffs, real-time fleet state synchronization, and error recovery without human fallback. For early AI application builders, this is a reality check: building an 'agent' isn’t about wrapping an LLM in a function-calling layer — it’s about owning or deeply integrating with the full operational stack (auth, state, inventory, compliance). Teams betting on 'plug-and-play agent orchestration' should prioritize interoperability with existing enterprise systems over novel reasoning architectures.