Uber’s AI push is best understood as a headcount-efficiency strategy. In 2026, the company is spending more on AI while hiring less, and CEO Dara Khosrowshahi said autonomous agents now produce roughly 10% of Uber’s code changes, with humans still reviewing that code before it is merged ^[10].

The core idea: fewer incremental hires, more output per employee

Uber is not presenting AI as a full replacement for engineers today. The reported strategy is to “meter” headcount growth while using AI to raise employee throughput ^[10]. Khosrowshahi has said he wants employees to use AI to increase output by 20%, 30%, 50%, or even 100% ^[10].

That changes the hiring math. Instead of adding people every time the company needs more engineering capacity, Uber can try to add some of that capacity through coding agents, developer tools, GPUs, and workflow automation. Khosrowshahi has also discussed the longer-term possibility of replacing some incremental engineering headcount with AI agents and GPUs, but the current model still keeps human engineers in the loop ^[5][10].

What AI is doing inside Uber engineering

The biggest shift is from AI as autocomplete to AI as an active participant in software delivery. Uber CTO Praveen Neppalli Naga said the company has “leaned in hard” to AI coding, that 95% of Uber engineers use AI tools every month, and that an internal AI agent is making about 1,800 code changes per week ^[13].

That does not mean every change is blindly shipped. The most important control remains human review: Khosrowshahi said AI-generated code is checked by employees before it is added to the repository ^[10].

Uber’s developer-productivity work also appears to span more than code generation. A Developer Productivity Engineering session described Uber as investing in AI across the software development lifecycle to help developers “Ship Quality Faster,” including coding-assistant customization for large monorepos, agentic systems for large-scale code migrations, and AI-powered testing and code-review workflows ^[14].

Agentic development is the real productivity bet

The strategic shift is not just that engineers get better suggestions in an IDE. It is that more work can be assigned to tools that act like agents: taking a larger task, using project context, producing code, and sometimes preparing changes for review.

An inside look from The Pragmatic Engineer reported that 84% of Uber developers were agentic coding users, meaning they used command-line agents or made more agentic requests than simple tab-completion requests in an IDE ^[8]. The same report said 65% to 72% of code was AI-generated inside IDE-based tools ^[8].

Those numbers should not be read as the same metric as Khosrowshahi’s 10% figure. The 10% figure refers to code changes produced by autonomous agents, while the 65% to 72% figure refers to code generated inside IDE tools ^[8][10]. In practical terms, AI may be helping draft a much larger share of code than the share of merged changes that are attributed to autonomous agents.

Why this can reduce hiring needs

If engineers can ship more with the same staffing level, Uber can grow output without growing headcount at the same pace. That is the economic logic behind spending more on AI while hiring less ^[10].

The cost does not disappear; it shifts. Reports have described a surge in Claude Code usage at Uber that burned through a 2026 AI coding budget earlier than expected, while also noting Uber’s use of tools such as Claude Code and Cursor ^[2][3]. Treat those budget reports as reported anecdotes rather than a full accounting of Uber’s AI spend, but they show the tradeoff clearly: software capacity is increasingly being planned as a mix of people, agents, tooling, and compute.

AI is also moving into operations, not just coding

Uber has used AI for years in areas such as ridesharing pricing and matching drivers with passengers ^[20]. More recent reporting says generative AI and agentic AI are also being applied to customer support, driver onboarding, and parts of the engineering development lifecycle, reducing manual intervention in some operational workflows ^[11].

That matters because the productivity story is broader than software engineers writing code faster. If AI can diagnose internal-service problems, streamline support, or reduce manual steps in onboarding, the company can reduce bottlenecks outside the core coding workflow as well ^[11].

What this does—and does not—mean for engineers

The evidence points to a supervised AI engineering model, not a no-engineer model. Uber is using agents to draft and prepare more work, but human employees still review AI-written code before it is merged ^[10]. The company’s public numbers also measure adoption and code activity more clearly than they measure audited productivity gains.

The practical takeaway is that Uber is trying to make each engineer more leveraged. Engineers still own architecture, judgment, review, debugging, and production quality, but more of the drafting, migration, testing, and repetitive implementation work can be handed to AI systems ^[10][14].

For hiring, that means the pressure is most likely on incremental headcount growth. Uber can keep investing in engineering capacity while adding fewer new employees than it otherwise might have needed, as long as the productivity gains from AI tools hold up in real workflows ^[10].