Back to the assessment

How we score AI insurance gaps.

Rule-based and reproducible. The engine starts from a base coverage matrix (10 risk categories × 8 policy types), applies endorsement-specific exclusion triggers (CG 40 47 / CG 40 48), use-case amplifiers, and industry weights, then produces a colour-coded heatmap with cited exclusions.

Cell-score formula

score(risk, policy) =
    base_coverage(policy_type, risk_category)
  × cg_40_47_48_applicability(policy, jurisdictions, use_cases)
  × use_case_trigger(use_cases, risk_category)
  × industry_multiplier(industry, risk_category)

color = {
  ≥ 0.80: green       — Policy likely responds
  0.50 – 0.79: amber  — Ambiguous; broker conversation required
  0.20 – 0.49: red    — Policy likely excludes / limits
  < 0.20: dark_red    — Material gap; affirmative coverage recommended
}

base_coverage values come from a curated matrix maintained by Buzzi insurance counsel.

Product fit

fit_score(product, user) =
    0.4 × use_case_match(product.coverage_types, user.use_cases)
  + 0.4 × risk_category_match(product.coverage_types, user.top_gaps)
  + 0.2 × industry_match(product.markets, user.industry)

# Top 5 returned, ordered by fit_score desc.

Integrity

Four commitments.

We are not an insurance broker.

No placement role anywhere in the conversation. Buzzi is a software company.

No commissions.

We never receive payment from any carrier or broker for surfacing their products.

No carrier sponsorships.

No carrier pays for placement on this tool, methodology page, or in any product card.

Every exclusion is cited.

CG 40 47, CG 40 48, silent-AI, war exclusion, IP training exclusion — each traces to a specific endorsement or court case.

Sources

Where the rules come from.

  • Endorsements — Verisk filings (CG 40 47 / 48) and carrier-specific affirmative-AI riders. Refresh quarterly with insurance counsel.
  • Case law — Bartz v. Anthropic (2025), Air Canada v. Moffatt (2024), Mata v. Avianca (2023), plus state insurance department filings.
  • Carrier marketing + fact sheets — public product pages from Armilla, Munich Re, AXA XL, Vouch, Testudo, Hiscox, Travelers, Chubb, Beazley, AIG.
  • Framework alignment — NIST AI RMF (Govern / Map / Measure / Manage) applied to liability classes.

Found a number that's wrong? Email hello@buzzi.ai. We publish corrections within 48 hours.