Leading agentic development projects requires new habits because traditional process assumptions — that implementation is the slow, expensive part — no longer hold. When AI agents compress build timelines, the bottleneck shifts upstream to definition and decision-making. Three key habits help: building small, fast decision loops with stakeholders instead of waiting for sprint ceremonies; getting comfortable working directly in the repo to manage context files, specs, and design references alongside code; and using a physical kanban board organized around 'blocked vs. moving' rather than velocity or estimates. The human role becomes about judgment, taste, and keeping the team aligned — the glue that holds a smaller, faster team together.
