Building AI for Insurance Underwriting and Rating under State Insurance Law
For P&C pricing actuaries, pricing data scientists, underwriting managers and filing analysts who build or buy underwriting and rating AI and must defend it to a state insurance department, one job task at a time.
Level
Intermediate
Study time
15 hours
Lessons
18 in 6 modules
Exam
20 questions
About this course
A fifteen-hour intermediate course for pricing, underwriting, filing and validation teams at US property and casualty insurers. Cedar Quay Mutual, a fictional Ohio-based regional insurer, must choose a bounded evidence program before a committee deadline. Eighteen preserved and completed technical lessons lead from model interactions and disparity testing to consumer notices, imagery verification, catastrophe-load support, submission reconciliation and a dated model/vendor file.
The supplied-output route requires a calculator and written reasoning; optional Python implementations use synthetic data only. State rules are taught within their line and jurisdiction, federal consumer-report duties are distinguished from conditional numerical-credit-score disclosures, and a NAIC model bulletin is not treated as nationwide law. Learners compare AI and deterministic approaches and test failures that numeric or schema checks miss. Practice feedback is advisory; the course does not authorize rates, bind insurance, confer a professional license or establish workplace competence.
What you will be able to do
Evaluate an interaction selection, disruption analysis and state-specific input-lineage screen using supplied modeling records
Analyze proxy uncertainty, outcome metrics and alternative-model tradeoffs without treating a screening heuristic as insurance law
Diagnose state-line, consumer-report and score-definition notice obligations and imagery verification gaps
Reconcile modeled average annual loss and evaluate Florida and California filing prerequisites with explicit model/version and commitment scope
Reconcile extracted commercial records and justify classification and referral questions from supplied sources
Evaluate a filing draft and assistant boundary using adverse counterexamples and source/authority records
Assemble a dated model/vendor evidence index and prioritize a bounded response to material change
Syllabus
6 modules · 18 lessons · about 15 hours
1
Module 1: Pricing Actuary: Filing a GBM-Informed Personal Auto Rating Plan
Use gradient-boosted models to find risk signal, translate it into rating tables a state rate analyst can review, and keep the plan cost-based and free of prohibited factors.3 lessons
Fitting a GLM and a GBM on a Synthetic Auto Book and Using SHAP to Find the Interaction the GLM Missed
Turning the Finding Into a Filed Interaction Table: Capping, Disruption Analysis and Michigan's Prohibited Auto Factors
Renewal Retention Models and the Price-Optimization Line: Keeping Rate Selection Cost-Based
Decision Lab · 12 minThe Interaction Before the Filing FreezePrepare a source-grounded underwriting/rating evidence recommendation after a material record change.
2
Module 2: Pricing Data Scientist: Unfair-Discrimination and Proxy Testing of a New Rating Variable
Run the outcome testing Colorado and New York expect before a new external variable or telematics score goes into a filed rating plan.4 lessons
Inferring Race and Ethnicity With BIFSG for Testing Only, and Measuring Its Misclassification
Computing Adverse Impact Ratio and Standardized Mean Difference on Premiums, Declines and Nonrenewals
Writing the Circular Letter 7 Three-Step Memo and Pricing the Less Discriminatory Alternative
Telematics Pricing Lead: Testing a Usage-Based Score for a Night-Shift Proxy Under Colorado's Auto ECDIS Rules
Decision Lab · 12 minThe Alternative That Lost Its DenominatorPrepare a source-grounded underwriting/rating evidence recommendation after a material record change.
3
Module 3: Product Manager and Underwriting Manager: Credit-Based Scores and Aerial Roof Scores
Govern two external scores that change individual premiums and renewals, with the notice, recency and dispute controls each one requires.2 lessons
Credit-Based Insurance Scores: State Bars, Reason Codes and the 15 U.S.C. 1681m(a) Adverse-Action Notice Pipeline
Writing the Inspect, Repair, Renew or Nonrenew Rule for Aerial Roof Scores Under State Imagery Bulletins
Decision Lab · 12 minTwo Notices and a Disputed RoofPrepare a source-grounded underwriting/rating evidence recommendation after a material record change.
4
Module 4: Property Pricing Actuary: Catastrophe-Model Output in Homeowners Rates
Turn modeled loss into a filed catastrophe load and meet the Florida and California rules on which models may be used and how.2 lessons
From Modeled Average Annual Loss to a Territory Catastrophe Load With Mitigation Credits
Florida's Commission-Accepted Models and California's Model Approval, Reinsurance-Cost and 85% Distressed-Area Rules
Decision Lab · 12 minThe Model Window ClosesPrepare a source-grounded underwriting/rating evidence recommendation after a material record change.
5
Module 5: Commercial Underwriter and Underwriting Assistant: Submissions, Loss Runs and Class Codes
Build extraction and triage where the LLM reads documents, code does the arithmetic and the underwriter makes every decision.3 lessons
Submission Intake, Clearance and Appetite Triage: A JSON Schema, a Seeded Schedule of Values and Field-Level Scoring
Summarizing a Five-Year Loss Run: LLM Extraction, Code Reconciliation and a Narrative That Cites Every Number
Suggesting Workers' Compensation and General Liability Class Codes From a Business Description With Human Confirmation
Decision Lab · 12 minThe Balanced SOV That Was Still WrongPrepare a source-grounded underwriting/rating evidence recommendation after a material record change.
6
Module 6: Filing Analyst, Operations Lead and Model Validator: Filings, Authority and a Defensible Model File
Draft filings without changing a number, encode authority an agent cannot exceed, and keep a filed rating model monitored and documented for a regulator request.4 lessons
Answering a Six-Part Objection Letter With an LLM and a Number-Matching Guardrail
Encoding Underwriting Guidelines and Referral Authority, and Red-Teaming an Assistant That Must Not Bind
Monitoring a Filed Rating Model: Input Drift, Actual-to-Expected by Segment and a Vendor Field Rescale
Assembling the NAIC Bulletin Section 4 File for a Rating Model and Its External-Data Vendor
Decision Lab · 12 minThe File Counter and the Allowed ToolPrepare a source-grounded underwriting/rating evidence recommendation after a material record change.
Capstone simulation · 30 minThe Cedar Quay Committee AllocationAllocate a finite evidence budget across six underwriting/rating workstreams after a vendor and calendar change.
Exam and certificate
The exam unlocks once you have completed every lesson and played each Decision Lab to an ending. It draws 20 questions from a larger bank, spread across the modules, with the answer options shuffled for each attempt. You have 40 minutes and need 80% to pass. After a failed attempt you can try again after 24 hours, up to 3 attempts in total. Feedback is given per module, never as the answer key.
Passing earns a certificate of completion and assessment from the Wirewalk Institute of AI with a unique ID that anyone can check at institute.wirewalk.com/verify. It is not an accredited qualification or a vendor certification. How certification works.
The Wirewalk Institute of AI is independent. This course teaches publicly documented products and is not affiliated with or endorsed by Akur8, Earnix, Guidewire, Duck Creek, Cytora, Federato, Sixfold, LexisNexis Risk Solutions, TransUnion, FICO, ZestyAI, CAPE Analytics, Moody’s, Verisk, Karen Clark & Company, Cotality, Cambridge Mobile Telematics, Arity, Nearmap, or EagleView. Product and company names are trademarks of their respective owners.
A free one-hour briefing built on eight real financial-services jobs, not an overview: spreading financials for a credit memo, writing adverse-action reasons from an ML score, drafting a SAR narrative, setting fraud-score thresholds, checking an earnings-call summary, reviewing an advisor's AI meeting note, summarizing a commercial loss run and reviewing a photo-based auto estimate. Each lesson shows where AI helps, where it must not decide, and the rule that governs that task.
Practice banking work on a self-contained fictional bank packet: repair service drafts, reconcile credit facts and reason records, summarize investigation leads, compare review burdens and prepare a scoped recommendation. Supplied outputs let you complete the work without a model account, code installation or real customer records.
Follow one composite $46 billion bank through a year-long AI program with public promises and OCC examinations on the calendar. Build and test the systems bank AI teams own, from an ML credit model to a servicing assistant, prove what they are worth in savings the CFO can book, and make the strategic trade-offs in six Decision Labs and a capstone simulation.