Actuarial Pricing Models & GLM Insurance Pricing with AI

P&C pricing work — actuarial pricing models, frequency-severity GLMs, machine-learning diagnostics, and the rate-level indication — is the technical core of P&C ratemaking. This guide covers where AI actually fits that workflow: not fitting the models, but reviewing GLM insurance pricing output for reasonableness and writing the filing-quality documentation that state departments require. For the modelling theory itself, see Actuarial Pricing Models: GLMs, GBMs & AI for P&C Ratemaking.

Where Pricing Work Actually Gets Stuck

P&C insurance pricing has two distinct phases: the modeling work (running GLMs in R or Python, fitting frequency and severity models, building rating plans) and the documentation work (rate indication exhibits, state rate filings, actuarial opinions, management presentations). Most actuarial teams are reasonably well-equipped for the modeling phase — the bottleneck is everything that happens after the model runs. Reviewing 40 pages of GLM output for reasonableness, writing the filing narrative for a state department, producing the actuarial rate indication exhibit with supporting documentation, and explaining the rating variable choices to compliance — these tasks consume days and are poorly served by existing tools. That is where Claude adds value in a P&C pricing workflow.

GLM Rate Model Review

Once a GLM runs, the actuary's job is to review the relativities and decide whether the model is producing credible, actuarially sound results. A 47-variable commercial auto GLM produces a table of relativities that need to be evaluated for direction, magnitude, monotonicity, interaction logic, and regulatory defensibility. Claude doesn't build the model, but it reviews the output: cross-checking relativities against prior year, flagging reversed signs that suggest multicollinearity, identifying variables where the confidence interval is too wide for the relativity to be credible, and generating the variable-by-variable documentation narrative that goes into rate filings.

• "Review these GLM relativities for our commercial auto book. Prior year relativities are [prior]. Identify: any reversed signs, variables with >50% change from prior year, variables where 95% CI crosses 1.0, and any potential multicollinearity concerns. Format as a review memo."
• "The following variables have relativities that differ from industry benchmarks by more than 20%: [list]. For each, assess whether the deviation is credible given our volume, and draft the actuarial justification for the filing."
• "Our ML-boosted pricing model is replacing a GLM for homeowners. Draft the state department narrative explaining the model change, covering: data inputs, validation methodology, how the model satisfies unfair discrimination requirements, and how relativities were tested for reasonableness."

Rate Indication and On-Level Analysis

The rate indication exhibit is the core actuarial deliverable — the calculation proving that proposed rates are actuarially justified. It requires on-level adjustments to remove the effect of prior rate changes from historical premium, loss development to ultimate, trend factors, and expense loading. Each assumption needs to be documented and defensible. Claude reviews the indication framework, checks whether the on-leveling procedure is consistent with the rate history provided, flags mathematical errors in the development or trend selections, and produces the supporting exhibit narrative in the format state actuaries expect to see.

• "Here is our rate indication for personal auto liability. The on-level premium is [X], ultimate loss estimate is [Y], target LR is [Z]%. Indicated change is [W]%. Review the calculation for errors, check whether the trend and development selections are consistent with each other, and draft the actuarial certification language."
• "Our rate indication shows -8% but we are proposing +4%. Draft the actuarial explanation of why we are not implementing the full indicated change — reference investment income, competitive position, and planned underwriting actions."
• "The state requires a difference exhibit showing why our indicated rate differs from ISO pure premiums. Draft the exhibit and the supporting narrative explaining our book's deviation from industry average."

Loss Cost Trend Analysis

Trend selection is one of the most judgment-intensive parts of P&C pricing. The actuary must select trend periods, decide whether to use pure frequency/severity or pure premium trends, evaluate whether recent accident years are credible, and document the rationale for every selection. This documentation is what state departments scrutinize in rate filings and what appointed actuaries rely on for their opinions. Claude reviews trend exhibits, questions the trend endpoint selection, evaluates whether segmented trends are warranted, and generates the trend section of the actuarial rate filing in the format required by major state departments.

• "Here is our frequency and severity trend data for commercial GL for accident years 2018-2025. Claim counts are [data], average severity is [data]. Recommend a trend selection, evaluate whether the 2020-2021 COVID period should be excluded or included, and draft the trend selection documentation."
• "Our severity trend has accelerated from 4% to 9% over the past two years. Draft the management presentation explaining the drivers — social inflation, litigation funding, nuclear verdicts — and the implications for our pricing adequacy."
• "We need to justify our medical severity trend of 7.5% to the state of [X], which has challenged our trend assumption as too high. Draft the response memo with supporting data references."

Credibility Weighting and Small Segment Pricing

When a rating segment lacks sufficient volume for full credibility, the actuary must blend the segment's own experience with a complement. The choices — which complement to use, what credibility formula applies, how to handle zero-claim segments — require documentation that filing actuaries and state reviewers will examine. Claude helps select and document the credibility procedure, ensures consistency with prior filings, and drafts the credibility methodology section. For admitted lines in states with filing requirements, this documentation is not optional — it is a required component of the actuarial work product.

• "We have 340 earned car-years in our high-value vehicle segment. Full credibility requires 1,082 exposures. Our complement of credibility is our statewide loss cost. Draft the credibility exhibit using the limited fluctuation method and the credibility blending documentation for the filing."
• "Twelve of our 47 rating territories have fewer than 50 claims. We are proposing to use a credibility-weighted blend of the territory's own experience and statewide average for each. Produce the credibility table and the actuarial justification."

Rate Adequacy Monitoring and Rate Action Memos

Between formal rate filings, pricing actuaries monitor rate adequacy continuously — tracking the booked loss ratio against the target, watching for emerging trends, and flagging when the book is approaching inadequacy. This monitoring feeds management dashboards and, when rate action is needed, the actuarial memo that triggers the rate change process. Claude drafts rate adequacy monitoring summaries, generates the rate action recommendation memo, and produces the management-level explanation of why rates need to change, framing technical adequacy concepts in terms CFOs and underwriting heads can act on.

• "Our booked combined ratio for commercial property is 98.3% vs. target 94%. Frequency is in line but severity is trending 11% vs. our 7% filed assumption. Draft a rate action memo recommending a [X]% rate increase, including: current adequacy position, trend deterioration analysis, and timing recommendation."
• "Produce a one-page rate adequacy dashboard for each of our five product lines using the following data: [data]. Format for a quarterly pricing committee presentation — flag any line where the adequacy position has deteriorated by more than 2 points."

State Rate Filing Narratives

Filed rates require actuarial support filings. Many states (CA, FL, NY, TX) require detailed actuarial narratives explaining the rating methodology, variable selection, model validation, and actuarial basis for the proposed rates. Writing these narratives is time-consuming and requires precise regulatory language. Claude generates filing-quality narratives from structured actuarial data, formats them to state-specific requirements, and drafts the responses to state actuary objections — the back-and-forth that can extend a filing review from 30 days to six months.

• "Draft the actuarial support filing for our personal auto rate filing in California. The filing includes a 4% rate increase based on a +7% indication discounted for competitive reasons. Required sections: executive summary, rate indication exhibit description, trend analysis, development methodology, and actuarial certification."
• "The Florida OIR has objected to our hurricane wind trend assumption as inadequate. Draft the response addressing their specific comments: [comments]. Include references to AIR/RMS model output and the historical loss cost data."
• "Produce the SERFF filing description for our commercial GL rate revision. Keep it under 500 words. Explain what changed, why, and how the actuarial basis was derived."

Related Actuarial Guides

• Actuarial Pricing Models — GLMs, GBMs & AI for P&C Ratemaking
• IBNR Reserves — Chain-Ladder & Bornhuetter-Ferguson with Claude
• P&C Loss Reserving AI — Development Factors & Reserve Analysis
• IFRS 17 CSM Roll-Forward — Actuarial Disclosure with Claude
• Best AI Tools for Actuaries 2026 — Full Comparison
• Claude for Actuarial Analysis — Complete Guide