Life Insurance Reserving AI — VM-20, PBR & Actuarial
How life insurance actuaries use Claude AI for principle-based reserving (PBR): VM-20 net premium reserve, deterministic reserve cash flow testing, CTE70 stochastic reserves, prescribed economic scenarios, and VM-31 actuarial memorandum documentation.
Educational content, not professional advice — AI output and figures here can be wrong. Verify before you rely on it. Full disclaimer →
Life Insurance Reserving and AI
VM-20 principle-based reserving requires actuaries to document assumptions, test prescribed economic scenarios, run stochastic models, and write extensive actuarial memoranda — work that is analytically intensive and documentation-heavy. Claude with FinSkilz accelerates the documentation and analytical explanation work: structuring assumption memos, explaining stochastic results, drafting VM-31 memorandum sections, and performing sensitivity analysis on reserve drivers.
Net Premium Reserve (NPR) Framework
- "Explain the VM-20 net premium reserve calculation for a 20-year level term policy: male, age 45, non-smoker, face amount $500,000, issue date 2023. The NPR uses: (1) 2017 CSO mortality table (gender-distinct); (2) 4.0% valuation interest rate; (3) net premium = benefit premium calculated using CSO and valuation rate. Walk through the NPR formula step by step: calculate the net single premium, annual net premium, reserve at end of policy year 1, year 5, year 10."
- "Compute the NPR exemption test for our term insurance block: total reserve under NPR = $42.6M. Is this below the simplified method threshold? The exemption allows companies to use the simplified method (Commissioner's method) if NPR × adjustment factor is below the deterministic reserve by less than [threshold]. Show the NPR vs deterministic reserve comparison and whether this block qualifies for the simplified method."
Deterministic Reserve and Cash Flow Testing
- "Structure the deterministic reserve cash flow testing for a UL product block: the DR requires projecting policyholder cash flows (premiums, benefits, lapses, expenses) under the prescribed economic scenario. Assumptions: mortality per VM-20 Section 3, lapses per company experience study, expenses per pricing assumptions with prescribed margins. Discount using the net asset earned rate (portfolio rate method). Explain the 6 prescribed scenarios and how to determine which produces the highest reserve."
- "Perform the deterministic reserve sensitivity analysis: base case DR $85.4M. Sensitivity tests required by VM-20: (1) mortality +10%: $87.2M; (2) lapse rates -20%: $89.6M; (3) interest rate +100bps: $82.1M; (4) expenses +5%: $86.0M. Identify the largest reserve driver and describe the business impact. Draft the sensitivity table for the VM-31 actuarial memorandum."
Stochastic Reserve and CTE70
- "Interpret the stochastic reserve output from our VM-20 model: 10,000 scenarios generated using Academy prescribed economic scenarios. Sorted reserve values by scenario outcome. The 3,000th worst scenario (70th percentile) has reserve $94.2M. CTE70 = average of scenarios 1 through 3,000 = $98.7M. The deterministic reserve under the worst prescribed scenario is $85.4M. The stochastic reserve (CTE70) binds: SR $98.7M > DR $85.4M > NPR $78.3M. Required reserve = $98.7M. Explain this result in the VM-31 memo language."
- "Explain the Academy prescribed economic scenarios for VM-20 stochastic testing: the 16 deterministic scenarios cover interest rate paths (up, down, level, volatile) and equity return scenarios. The stochastic scenario generator produces interest rates using a mean-reverting model with prescribed parameters. Why does VM-20 use prescribed scenarios rather than company-generated scenarios? What is the regulator's rationale for standardizing the interest rate model?"
Assumption Documentation and VM-31
- "Draft the mortality assumption section for the VM-31 actuarial memorandum: our company uses the 2015 VBT (Valuation Basic Table) for best estimate mortality, applying a 90% factor for preferred lives and 115% for standard lives. VM-20 margins: +7.5% margin on best estimate mortality for the NPR; mortality margins for DR are as specified in VM-20 Section 9.C. Describe the assumption derivation, credibility weighting with industry tables, and VM-20 compliance."
- "Draft the lapse assumption documentation: our policyholder experience study shows shock lapse at policy year 10 (end of level term period) of 72%. First-year lapse 8%, renewal years 4-6% annually. We are applying a VM-20 prescribed margin of +20% on lapse rates above 10% shock. Explain how anti-selective lapsation could affect reserves and why the lapse margin exists. Draft the VM-31 lapse section in regulatory actuarial memo style."
Where to Start
Start with the assumption documentation — it is the most text-intensive part of VM-31 and where AI adds the most immediate value. Describe your mortality table, lapse experience study, and expense assumptions, then ask Claude to draft the regulatory-style documentation section. For stochastic model outputs, paste the CTE distribution and ask Claude to interpret the results and draft the narrative for the appointed actuary's memo.
Connect Claude to live financial data via MCP — EDGAR, FDIC, BIS, CME and 18 more.
New guides & tools — free
Get notified when we add new MCP servers, finance AI guides, and eval results.