Six
conditions.
The headline figure is only valid under normal load. Push the system — surge, demotion, escalation, infrastructure outage, critical risk — and autonomy moves. The worst case, Condition E, drops it to 48.43%, an exposure of 27.55pp.
75.98%Baseline · Condition A
48.43%Worst case · Condition E
27.55ppExposure
96.08%Ceiling baseline
Autonomy by condition
A → FSub-tests
11 probesConditions
mechanism| # | Condition | Autonomy | Δ | Mechanism |
|---|---|---|---|---|
| A | NORMAL LOAD | 75.98% | 0pp | No transformation. Expected daily workload, complete and valid submissions. |
| B | PEAK LOAD (3x to 10x) | 73.05% | -2.93pp | Volumes surge 8x on routine stages, 1.5x elsewhere, AND 20% of Class C demotes to Class B because preconditions fail under backlog pressure. |
| C | POOR INPUT QUALITY | 70.07% | -5.91pp | 30% of Class C effort demotes to Class B because a precondition fails. Modelled as a mix transformation, not a factor change. |
| D | REGULATORY CONFLICT | 72.04% | -3.94pp | 10% of Class C escalates to Class D because two or more applicable rules conflict and no agent may resolve by choosing. |
| E | SYSTEM FAILURE | 48.43% | -27.55pp | Knowledge Spine outage. Class C eliminated and Class A halved on the five regulation-dependent stages; effort transfers to Class D. |
| F | CRITICAL-RISK CASE | 66.01% | -9.97pp | Life-safety stages forced to risk 3.00 AND 40% of their Class C reallocated to Class D and E, because human authority is required regardless of confidence. |