Uber now taps OpenAI’s large language models to power in‑app assistants, aiming to cut friction for drivers and riders. The move follows months of scaling machine learning across 40 million daily trips in 15,000 cities. By turning traffic, weather, and demand data into instant chat responses, Uber sharpens its real‑time marketplace.

Uber Assistant, the flagship AI helper, delivers real‑time positioning guidance to drivers. The system translates earnings heatmaps and trend curves into easy‑to‑read suggestions, letting drivers ask natural‑language follow‑ups. Early pilots show new drivers achieve profitability in fewer trips, while veterans repeatedly use the tool to fine‑tune shift plans for optimized earnings and route efficiency every day.

To keep responses safe and fast, Uber built a multi‑agent architecture. Lightweight nano models handle quick classification, while larger reasoning models tackle complex queries. An internal AI Guard layer screens prompts, curbs hallucinations, and enforces policy, ensuring the assistant stays reliable across 70+ countries for drivers and riders worldwide in real time operations every day.

Voice booking, powered by OpenAI's Realtime API, lets users request rides with spoken intent. The feature parses context, suggests suitable vehicle types, and syncs visual cues, removing the need for multiple taps. Initial rollouts in the U.S. indicate higher engagement, proving voice can streamline complex trip planning for drivers and passengers across major metropolitan areas.