Agentic development often starts with one AI agent handling an entire task. But as that task grows, the agent may cross too many boundaries, mix responsibilities, and leave the developer untangling the result.

Using a small Battleship game repository, this session compares two approaches to the same development challenge: one generalist agent versus a fleet of focused agents responsible for implementation, testing, refactoring, documentation, and review.

We’ll examine how clear task boundaries, scoped context, validation, and deliberate hand-offs can make agentic development easier to control. Along the way, we’ll explore the coordination challenges of working with multiple agents and the developer’s role in guiding the workflow.

You’ll leave with a practical method for deciding when one agent is enough, when an agent fleet is helpful, and how to keep the developer in control.