AI decision risk score
A risk rating of a decision type being AI-assisted.
Formula
AI decision risk score
weighted score across stakes, reversibility, model confidence, data trust, human oversight, and governance tier
Unit
score
Direction
Lower is better
Cadence
For every high-impact AI-assisted decision type.
Worked Example
AI-assisted eligibility decision review.
High stakes
30
Low reversibility
20
Confidence gap
15
Weak oversight
20
Weak lineage
10
Calculation
30 + 20 + 15 + 20 + 10
Result
95 / 100 risk
Interpretation
This decision should not be automated without stronger gates and oversight.
Recommended action
Require human approval and improve evidence lineage before deployment.
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
Where AI involvement in high-stakes, low-reversibility, or low-confidence decisions requires stronger oversight.
Manual assessment method
Review AI-assisted decision types by stakes, reversibility, model confidence, human review, and governance tier.
Lapemo calculation path
Conceptually useful, but weighting and definition need a product decision before platform use.
Interpretation Bands
Healthy
0-25
Risk is low or controlled.
Watch
26-50
Risk should be reviewed before scale.
Risk
51-75
Risk is likely to affect outcomes.
Critical
76-100
Reduce autonomy, scope, or exposure before proceeding.
Required inputs
- Decision stakes
- Reversibility
- Model confidence
- Information trust
- Oversight level
- Governance tier
Data sources
- AI governance checklist
- Decision record
- Risk register
- Model evaluation notes
Common pitfalls
- Letting model confidence overpower high stakes
- Ignoring reversibility
Recommended actions
- Lower autonomy
- Add approval gates
- Improve evidence and lineage
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.