Agentic Software Engineering: The Rise of AI Teammates | DevNet Research

In this stream, Oleksii Borysenko sits down with Hao Li, Postdoctoral Researcher with SAIL and MCIS at Queen's University, to discuss Agentic Software Engineering, AI Teammates, and empirical studies of agentic pull requests across GitHub repositories. Explore Hao Li's publications and research: https://leo-lihao.github.io/ AI coding agents are no longer experimental tools. They are actively contributing to real-world open-source projects, with studies reporting acceptance rates comparable to, or even higher than, those of human contributions. Yet history shows that adopting a powerful new technology without rethinking the surrounding processes often leads to failure. In this stream, we'll cover: The foundations of Agentic Software Engineering and why AI agents should be treated as teammates rather than tools. Findings from large-scale empirical studies of agentic pull requests across GitHub repositories — how coding agents behave in practice. The emerging infrastructure reshaping how teams communicate, coordinate, and govern AI teammates, including agent context files (e.g., CLAUDE.md, AGENTS.md) and the Model Context Protocol (MCP) ecosystem.