Definition
Percentage of AI-recommended insurance products where the coverage type, limits, deductible, and endorsements are appropriate for the customer's disclosed risk profile and financial situation.
The Measurement Gap
AI recommendation engines optimize for conversion rate and premium volume, but may not optimize for suitability. An engine can achieve strong conversion while systematically recommending inadequate limits, unaffordable deductibles, or coverage types that miss the customer's primary risk exposures. Legacy distribution metrics report conversion rates, policy counts, and retention, but do not measure whether products sold are appropriate for the customers who bought them. A distribution channel can report excellent business metrics while building a book of improperly matched policies that creates E&O exposure and market conduct risk when losses reveal inadequate coverage.
Decision Supported:
Is the AI recommendation engine placing customers in appropriate coverage, or is it optimizing for conversion and premium while creating suitability gaps that surface as E&O claims and market conduct findings?
Formula
Plain Language: Sample policies where the AI recommendation engine influenced product selection. Have a suitability reviewer assess whether coverage matches the customer's disclosed risk profile. Score on three dimensions: coverage type appropriateness, limit adequacy, and deductible affordability.
Formula:
PSAS = (1/N) x Sum of (Coverage_score_i + Limit_score_i + Deductible_score_i) / 3
Variables:
- N: number of sampled AI-recommended policies
- Coverage_score_i: 1 if coverage type matches disclosed risk exposures, 0 if coverage has gaps relative to disclosed risks (e.g., flood zone property without flood coverage)
- Limit_score_i: 1 if limits meet minimum adequacy standards for the risk, 0.5 if marginally adequate, 0 if clearly inadequate (e.g., liability limits below disclosed asset value)
- Deductible_score_i: 1 if deductible is affordable relative to disclosed financial situation, 0.5 if marginally affordable, 0 if clearly unaffordable (e.g., $5,000 deductible for a household with $2,000 savings)
Units:
0 to 1. 1.0 means every sampled recommendation was suitable. Scores below 0.80 indicate systematic suitability gaps.
Data Requirements:
| Data Element | Source System | Classification |
|---|---|---|
| AI product recommendations (product, limits, deductible) | Recommendation engine / quoting platform | NATIVE |
| Final sold policy details | Policy administration system | NATIVE |
| Customer risk profile (disclosed assets, exposures, financial situation) | Application data / quoting platform | NATIVE |
| Suitability review results | Structured review program (new process required) | MANUAL |
| Agent override records (AI recommendation vs. final sale) | Quoting platform / CRM | NATIVE (if tracked) / NOT CAPTURED |
Proposed formula, pending validation pilots.