Ewan Simpson

Stage 3 prototype

AI-OS Workbench

This is the deeper task-layer worksheet: AI mode, residual effort, Decision Compass risk, proof burden, future effort, and capacity released. It is deliberately transparent. No mystical doom percentage nonsense.

Residual assumptions: Automate = 20%, Assist = 50%, Advise = 70%, Human-only = 100%.

Hours are reported as estimated productive hours per month. Demo and imported default task values are converted from weekly-scale assumptions using 4.33 weeks per month.

Task register and scoring

Score Decision Compass dimensions from 0 to 3. The total suggests the permitted AI posture: 0–5 Autopilot, 6–9 AI–Human Interface, 10–12 Assist, 13–15 Human-only.

Task Bundle AI mode Proof Current hrs/mo Stakes Ambiguity Reversibility Data sensitivity Reputation DC total Permitted posture Future hrs/mo Released hrs/mo