Uber deploys AI agents across customer support to surface context, automate investigations, and deliver empathetic resolutions

Background

Uber's marketplace complexity — spanning rides, food delivery, and grocery across multiple geographies — generates an enormous volume of diverse customer service interactions. Front-line agents frequently needed to reconstruct context manually from prior interactions and navigate complex policies to resolve issues, limiting the time they could spend on higher-value or emotionally sensitive cases.

What Was Implemented

• AI conversational summaries that surface context from prior customer interactions before an agent responds
• Automated investigations for routine issues (wrong items, route disputes, fare adjustments)
• Empathetic next-best-response suggestions delivered to agents in real time
• Policy-to-action translation: AI converts complex Uber policy text into step-by-step resolution routines
• ChatGPT Enterprise deployed across marketing, data science, product, and engineering functions
• OpenAI-powered AI assistants for driver/earner support

Results

Uber reports qualitative improvements in customer outcomes and operational efficiency, with agents freed to focus on higher-value interactions. No specific quantitative KPIs (percentage reduction in handle time, CSAT improvement, automation rate) were disclosed in the primary source interview.

Lessons

• AI's highest near-term value in customer support may be context surfacing and policy translation rather than full automation.
• A multi-sided marketplace requires AI tailored to each stakeholder segment (rider, driver, merchant, eater) rather than a single universal model.
• Measuring AI impact through "controlled experiments comparing AI-augmented workflows with traditional ones" (as Malkani describes) provides more credible ROI evidence than self-reported metrics alone.