Delegation Redesign

Type: Process design principle
Referenced from: audit-test-automate-ai-delegation

Definition

"AI is a forcing function for better delegation. It rewards clear inputs, bounded tasks, explicit review, and honest judgment about risk."

Before delegating to AI, you must first redesign the work itself: break it into smaller tasks, define what can be checked, and decide what's safe to delegate.

The Redesign Ladder

Most tasks fail the 3-Questions Test initially. Each failure is a signal to redesign:

  1. Add a checklist — Make implicit knowledge explicit
  2. Add tests — Define what success looks like mechanically
  3. Add a human review step — Bound the failure radius
  4. Add a narrower scope — Reduce what the AI needs to handle
  5. Add a permission boundary — Contain what the AI can access

The Progression

Vague work → Inspectable → Delegable → Automatable

Once the work becomes inspectable, some of it becomes delegable. Once it becomes delegable, some of it becomes automatable.

Key Insight

"We do not win by asking AI to 'do the work.' We win by redesigning the work so that humans keep what 'only humans can do', while AI handles everything else."

This is the opposite of "just throw AI at it." The leverage comes from the redesign, not from the AI.

Example

Health provider report:
- Initial state: All AI, but failure radius too large (expulsion risk)
- Redesign: Added human verification step
- Result: AI handles the bulk, human catches the critical errors

Application to dark-factory-kb

The factory pipeline IS delegation redesign in action:
- Tasks are explicitly decomposed (intake → context → information → synthesis → gate)
- Each step has clear inputs/outputs (templates, frontmatter schemas)
- Quality gates act as failure radius bounds
- Human review is embedded at decision points (gate decisions, deploy approval)