How Uber uses AI for development: inside look

2026-03-10 · Source: The Pragmatic Engineer · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Advanced, short

Summary

Uber has significantly integrated AI into its engineering workflows, aiming to become a "GenAI-powered" company by eliminating toil and freeing engineers for creative work. This initiative, detailed by Principal Engineer Ty Smith and Director of Engineering Anshu Chada at The Pragmatic Summit, involves a four-layer agentic system built on Uber's Michelangelo platform, accessing internal context, supporting industry agents like GitHub Copilot, and deploying specialized agents for tasks like code review. Uber developed internal tools such as the MCP Gateway for standardized agent-data interaction, Uber Agent Builder for no-code agent creation, and the AIFX CLI for agent provisioning and management. The company also introduced Minion for scaling background agents and new dev tools like Code Inbox and uReview to manage increased AI-generated code. Currently, 92% of Uber developers use agents monthly, with 31% of code being AI-authored, yet AI-related costs have surged 6x since 2024.

Key takeaway

For AI Architects and Directors of AI/ML evaluating internal AI adoption, Uber's experience highlights the necessity of a robust, integrated AI stack and developer-centric tooling. Your strategy should prioritize eliminating developer toil and providing intuitive platforms like Uber's MCP Gateway and Agent Builder to foster organic adoption, rather than relying on top-down mandates. Be prepared for significant cost increases, as Uber saw a 6x rise in AI-related costs since 2024, necessitating early token cost optimization strategies.

Key insights

Uber's comprehensive internal AI stack transforms developer workflows, automating toil and increasing code generation.

Principles

• AI should eliminate toil, not automate all engineering.
• Standardized protocols (MCP) are crucial for agent integration.
• Developer experience tools drive AI adoption and efficiency.

Method

Uber's agentic system uses an internal AI platform, context sources, industry tools, and specialized agents, managed via an MCP Gateway, Agent Builder for workflows, and the AIFX CLI for provisioning.

In practice

• Implement a central gateway for AI agent-data interaction.
• Develop no-code tools for internal agent creation.
• Provide a CLI for unified agent provisioning and updates.

Topics

• AI Agents
• Developer Workflows
• Internal AI Platform
• Code Generation
• Cost Optimization

Best for: AI Architect, CTO, Director of AI/ML, AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• 📈 Data to start your week: Inside the AI boom – jobs, jargon & jittery uptime — AI is poised to automate junior roles, but widespread adoption faces cost and compute constraints.
• Why was MCP created? — MCP standardizes AI-tool communication, reducing M×N integration complexity to M+N implementations.
• State of the Art of Platform Engineering • Abby Bangser & Charles Humble • GOTO 2026 — Platform engineering must provide self-service, API-first, deterministic capabilities to enable scalable, compliant AI agent consumption.
• What Developers Actually Need to Know Right Now — AI's true frontier lies in orchestrating agents to solve real problems with control and traceability.
• Quests, token leaderboards, and a skills marketplace: the elite AI adoption playbook | John Kim — Empowering non-technical teams with AI-building platforms accelerates internal innovation and fosters an AI-first company culture.
• Sam Altman predicts AI agents will integrate any service they want, with or without official APIs — AI agents will autonomously integrate services, forcing all companies to become de facto API providers.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Pragmatic Engineer.