Human-in-the-loop clarity
The share of AI-touched decisions with an explicit autonomy level and defined oversight.
Formula
Oversight clarity coverage
(AI-touched decisions with autonomy level and oversight defined / total AI-touched decisions) * 100
Unit
percent
Direction
Higher is better
Cadence
Monthly for AI programs.
Numerator
AI-touched decisions with explicit autonomy and oversight
Denominator
Total AI-touched decisions
Worked Example
AI decision oversight review.
AI-touched decisions
31
Decisions with explicit oversight
14
Calculation
14 / 31 * 100
Result
45.2%
Interpretation
Humans may not know when to review, approve, or intervene.
Recommended action
Define oversight and autonomy levels before increasing AI scope.
Why it matters
This metric helps leaders see whether the ai amplification layer is healthy enough to support execution and AI readiness.
What it reveals
Whether humans know when AI supports, recommends, acts, or requires approval.
Manual assessment method
Classify AI-touched decisions by autonomy level and identify oversight requirements.
Lapemo calculation path
Supported by autonomy level, gate thresholds, supervisors, and AI-touched decision types.
Interpretation Bands
Healthy
85-100%
Strong enough for normal operating review.
Watch
70-84%
Usable, but gaps should be assigned owners.
Risk
50-69%
Weak enough to slow execution or create risk.
Critical
<50%
Do not scale the related workflow without intervention.
Required inputs
- AI-touched decision list
- Autonomy level
- Oversight owner
- Approval rule
Data sources
- AI governance checklist
- Decision Rights Matrix
- Agent accountability checklist
Common pitfalls
- Saying human-in-the-loop without naming the loop
- Ignoring autonomy levels
Recommended actions
- Define autonomy tiers
- Assign oversight owners
- Add approval gates
Measure Before You Scale
Measure the operating model before AI amplifies it.
Use the Metrics Library to understand what to diagnose, what to monitor, and where Lapemo can turn framework metrics into operating intelligence.