A score is seductive. Add the points, find the band, feel organized.
Bounded Operational AI™ scoring becomes useful when it changes what you inspect and build next. It becomes misleading when the total starts doing the job of evidence, permission, or owner judgment.
The THMC model scores workflow architecture across 12 dimensions. It also includes critical-failure overrides because strength in one area cannot cancel a missing boundary somewhere consequential. The question is how reliably a workflow operates, not whether the spreadsheet looks reassuring.
Use the canonical scale consistently
Each dimension receives 0–4 points: 0 is Undefined, 1 is Informal, 2 is Defined, 3 is Operational, and 4 is Verified. Twelve dimensions create 48 possible points.
The dimensions are Business Outcome, Process Quality, Founder Dependency, Permission Boundaries, Access Control, Approval Object, Review Evidence, Escalation Quality, Stop Conditions, Execution & Recovery, Outcome Verification, and Learning & Measurement.
Defined means a rule exists and can be explained. Operational means it is built into the workflow and reliably followed. Verified requires evidence, testing, logs, or outcomes. A document describing a control does not earn the same score as proof that the control held during execution.
Read the bands with the overrides
The v0.1 bands are 0–9: Level 0, Unbounded; 10–18: Level 1, Human Click; 19–28: Level 2, Defined Authority; 29–37: Level 3, Evidence-Based Control; 38–43: Level 4, Operationally Bounded; and 44–48: Level 5, Learning System.
Those raw bands are only the first interpretation. A consequential external action without a reliable stop mechanism caps maturity at Level 1. Broad/shared credentials or materially unnecessary access also cap it at Level 1. So does consequential human approval without a specific approval object.
If consequential external changes are judged only by automation execution, the maximum is Level 2. Irreversible or materially financial or reputational action without an escalation path caps maturity at Level 1. Apply every relevant override; the strictest cap governs the effective level.
Keep maturity and readiness separate
The score assesses the architecture. A readiness gate assesses one specific object at one consequential boundary. A mature workflow can correctly return NOT READY for an incomplete engagement. That is the control working.
THMC’s documented Studio readiness design requires every mandatory proof to pass for the same engagement. Eleven passes and one failure still mean NOT READY. Owner activation remains a distinct decision after readiness. This describes the canonical design; implementation and failure testing require their own evidence.
Imagine a hypothetical workflow with a strong total but the wrong customer linked to a payment. No score should authorize proceeding. Identity evidence fails, so the individual action remains blocked. Averaging that failure away would defeat the reason for having a gate.
Turn the assessment into the next build
Record the workflow, business outcome, risk class, raw score, effective level, override, founder dependency, current authority, primary weakness, next boundary, and evidence required before authority expands. Keep the evidence beside each dimension so another reviewer can challenge the rating.
THMC’s method is Diagnose → Prescribe → Implement → Monitor → Improve. A low score points toward a prescription. A high score should still lead to monitoring and deliberate limits. This is a practical model in development, not an independently validated certification or a compliance standard.
Assess one workflow with the people who run it. Start where consequence and recurring friction overlap. If you need help turning that assessment into a practical implementation, explore Studio Services. For the underlying handoff problem, revisit why business systems keep breaking.
