A vague backlog item is poor input for agents as for developers — now that agents act on your backlog, requirement quality matters even more. When backlog work stays manual, or AI lives in disconnected chats, teams miss agentic workflows' real value.
In this hands-on workshop, you turn backlog work into a governed, AI-assisted workflow with GitHub Copilot, the Azure DevOps MCP Server, and agent skills, in a prepared GitHub Codespaces environment — no local setup. Through guided labs, you build a hierarchical Epic → Feature → PBI backlog from a spec, refine items and acceptance criteria, and add linked test cases with real data — no manual editing. The pattern stays the same: AI proposes and refines; you review, steer, and approve — nothing ships without sign-off.