Mob programming, or ensemble coding, is one of the most effective formats for transferring technical skills inside a team. Junior engineers learn by watching how more experienced developers use their tools and by being included in design and direction decisions. It works well, until you add agentic AI to the mix.
Technical coach Emily Bache says the introduction of agentic AI tools has been a genuine shock to this workflow. The classic mob programming dynamic breaks down in predictable ways, and she has been searching for new interaction patterns that let a group code together with agentic AI without falling into those failure modes.
Why the Failure Happens
Bache points to the core tension: agentic AI moves fast and takes over the keyboard in ways that do not fit the structured rotation and deliberate conversation that mob programming depends on. The collaborative rhythm that makes ensemble coding valuable gets disrupted when an agent is autonomously generating and executing code while the group is still trying to reason through the problem together.
What She Is Exploring
Rather than abandoning mob programming or abandoning agentic AI, Bache has been developing new patterns of interaction designed to preserve the knowledge transfer and group decision making benefits while still taking advantage of what agentic tools offer. The full breakdown is available as a video on the Modern Software Engineering channel.
If your team uses ensemble coding and has been noticing friction since adding AI agents to the workflow, this is worth a watch.
