Skip to content

Lift 4: Scale Your Delegation

Where We're Starting

After Lift 3, you've got a deployed, tested observation network backed by automated tests. Your acceptance criteria are both your spec and your safety net. The pipeline works: Explore → Plan → Implement → Verify → Ship.

But you're still working through your backlog one story at a time. Write a story. Hand it to AI. Wait. Verify. Move on. AI builds fast — but you can only focus on one thing at a time. You're the bottleneck, and there's so much more you want to build.

Notice where you've been on the autonomy slider. In Lift 1, you worked synchronously — prompting AI and reviewing every response. In Lift 2, you encoded your judgment into skills so AI could follow your processes without re-explanation. In Lift 3, you added automated tests so verification happens without you watching. Each step gave AI more autonomy because you built more scaffolding around it.

This lift takes the next step. You'll learn to assess which work is ready to delegate, run tasks in the background, and trust the system you've built — acceptance criteria, automated tests, and project context — to verify the results without watching every step.

What You'll Learn

  • How to assess which work is delegation-ready vs. which needs more thought first
  • How to run AI tasks in the background and manage parallel conversations
  • How sub-agents work behind the scenes to handle complex tasks
  • How to batch similar work and trust your tests to verify the results

Sections

  1. Delegation Judgment — Deciding what's ready to delegate and what isn't
  2. Going Parallel — Background execution, sub-agents, and keeping parallel work focused
  3. Trusting Your System — Delegating with confidence because your system catches problems

By the End of This Lift

  • You can assess whether a story is delegation-ready or needs more thought first
  • You understand background execution and how sub-agents handle specialized work
  • You know when and why to start a fresh conversation for each task
  • You can batch similar tasks and delegate them in parallel
  • You're ready for the Final Sprint: more output, less bottleneck