Have you ever found yourself babysitting Copilot through every step, first asking it to run the tests, then to fix the lint errors, then to update the docs?

Agent mode helps, but getting out of that loop takes more than just turning it on.

GitHub Copilot can plan and execute multi step tasks, but to make it truly autonomous, you need to shape how it works. Define your team’s standards, give it clear roles, and structure the workflows it repeats.

In this session, we’ll explore how agent mode and agentic workflows let Copilot take on end to end tasks, and how custom instructions, custom agents, and skills help you control its behavior, keeping it aligned, predictable, and reusable.