GitHub is adjusting to a software world in which AI agents are producing more code, more often, and from more people. In a recent discussion, GitHub COO Kyle Daigle described how the company is responding to that shift, saying the rise of agentic coding is changing not only developer behavior but also the demands placed on GitHub’s platform.
Daigle said GitHub has long served as the central home for software development, but the company now faces a new scale of activity driven by AI tools. He pointed to a sharp increase in coding-agent usage in 2026 and said the resulting traffic is placing far more pressure on GitHub’s infrastructure than its systems were originally built to handle.
That pressure is visible across the platform. More code is being committed, more builds are being triggered, and more people are participating in development workflows. GitHub has also seen major growth in GitHub Actions, which Daigle described as a general-purpose compute layer for automation and continuous integration. The company is now dealing with a development environment that moves at machine speed rather than human speed.
Daigle said GitHub’s response has been to bring AI into the tools and habits developers already use, rather than forcing teams to adopt entirely new systems. He pointed to a range of surfaces where GitHub is weaving in AI, including Slack, Teams, email, the command line, desktop tools, and cloud-based agents.
The goal, he said, is to make GitHub act more like a context-aware operating layer for work. In that model, agents can pull from company information, check past decisions, and help users look backward across internal context before deciding what to do next.
This approach also reflects a broader shift in how GitHub thinks about software development tasks. Daigle said large, all-purpose skills are giving way to smaller, more focused micro-skills that can be combined by agents to complete more specific jobs. He also noted that AI is changing work beyond engineering, including communication, summarization, marketing, and analyst roles.
GitHub’s own leadership team is using agents internally, according to Daigle. He described a workflow in which he uses multiple agents over a weekend to help with different tasks, as well as AI-generated presentations prepared for executive audiences such as finance and revenue leaders.
He said these tools do not eliminate human work, but they do change the role of executives and chief-of-staff style functions. Instead of replacing judgment, AI helps surface information and draft outputs that humans can refine.
Daigle also said AI has drawn him back into coding more directly after years in leadership roles. That personal shift mirrors GitHub’s broader challenge. The company is trying to support a wave of users who are shipping more code while also maintaining the reliability, trust, and review processes that have long defined the platform.
As AI-generated code becomes more common, GitHub is also confronting questions about open source governance and software supply-chain security. Daigle discussed the strain that floods of agent-produced pull requests can place on maintainers, and how the meaning of a review may change when a large share of contributions come from software agents.
He said GitHub is exploring ways to preserve trust in that environment, including approaches that help teams vouch for code and assess agent-generated contributions. The company is also focused on security measures around dependencies, tokens and automation, especially after its acquisition of npm.
The broader message from GitHub is that AI agents are not a side feature. They are becoming part of the core operating model for software development. For GitHub, that means reworking products, infrastructure and internal practices at the same time the industry is asking its platform to handle far more code than before.