Uber is undergoing a major shift in how it builds software, and that shift is being driven by aggressive adoption of AI coding tools. According to the company’s chief technology officer, AI has moved from a helpful assistant to a core part of the engineering workflow, with autonomous agents now handling a meaningful share of day-to-day coding. This transformation is changing what it means to be a software engineer inside the company, and offers an early look at how AI may reshape technical roles more broadly.
The company’s internal metrics highlight the scale of this change. Around 95% of engineers at Uber now use AI tools every month, a level of adoption the CTO describes as a “reset moment” for engineering. Uber’s internal coding agent is generating roughly 1,800 code changes per week entirely on its own, up from less than 1% of total changes only a short while ago to about 8% today. In many cases, engineers no longer write individual lines of code; they review, refine, and approve what AI systems produce.
This evolution is driven by what the company calls agentic AI. Instead of simply suggesting snippets or autocompleting lines inside an editor, these AI agents can take on entire tasks with minimal supervision. Engineers hand off work such as implementing small features, refactoring modules, or addressing certain issues, and the agents generate complete code changes that are then reviewed by humans. As a result, the workflow looks less like back-and-forth assistance and more like delegation to a capable junior teammate who happens to be an AI.
Within Uber’s development environment, AI-generated code is becoming the norm rather than the exception. In some tooling setups, a majority of committed code now originates from AI systems, with estimates ranging around 65–70% in certain IDE-based workflows and even higher in command-line agents. The company is also seeing strong adoption of “agent-style” workflows, where a large share of AI users regularly choose to delegate entire tasks instead of relying solely on line-by-line suggestions.
For human engineers, this shift is redefining their responsibilities. Rather than focusing primarily on writing every line of code themselves, they are increasingly responsible for architecting systems, shaping requirements, guiding AI tools, and rigorously reviewing AI-generated outputs. Their work now leans more heavily on design decisions, risk assessment, debugging tricky edge cases, and ensuring that the finished software aligns with business goals and user needs. In effect, execution is becoming more automated, while oversight, judgment, and coordination remain firmly human.
The cultural side of this change is significant as well. Uber’s leadership notes that the strongest adoption is coming from engineers who are experimenting with AI in their daily work, not just from top-down mandates. Many developers see these tools as a way to move faster and focus on more interesting problems, even as the broader industry debates issues such as workload, burnout, and long-term career impact. The company’s approach suggests that curiosity, adaptability, and willingness to learn new workflows are becoming critical traits for modern engineers.
Looking ahead, Uber’s CTO anticipates an even more agent-driven future. He has described a vision in which AI systems take on larger portions of the software lifecycle, from writing and testing code to managing deployments, while human experts concentrate on defining problems, validating solutions, and orchestrating multiple agents. Hiring for engineering roles remains active, but the expectation is that day-to-day work will continue to shift toward what could be called “AI agent engineering” rather than traditional coding.
Uber’s experience demonstrates how quickly AI can move from optional productivity booster to foundational infrastructure in software development. As autonomous coding agents take on more of the execution, engineers are being pushed up the value chain, toward architecture, strategy, and cross-functional collaboration. For any organization watching this trend, the message is clear: AI will not simply change tools, it will change roles, workflows, and the very definition of what it means to build software.
Read more such articles from our Newsletter here.
