Create a copy-paste prompt pack for an actuary working on [ACTUARIAL WORK AREA] in [JURISDICTION OR BUSINESS CONTEXT]. The intended users are [USER LEVEL], and the approved materials available are [AVAILABLE MATERIALS]. Focus on these work needs: [PRIORITY TASKS]. Produce 8 distinct prompts, each 90-150 words, for ChatGPT, Claude, or Gemini. Cover the actual workflow where relevant: framing the question; documenting data lineage and exclusions; checking assumptions; testing calculations or model logic; explaining results to non-actuaries; preparing review questions; and recording limitations. Label each prompt with its use case and list the exact inputs the user must supply. At least one prompt must request a data-quality review that separates observed issues from hypotheses. At least one must request a model or spreadsheet review that identifies test cases, boundary conditions, reconciliation checks, and controls without asserting the model is correct. At least one must translate technical results into a decision memo that distinguishes results, uncertainty, and recommendations. Include a prompt for a peer-review checklist appropriate to the stated work area.
Do not provide pricing, reserving, capital, valuation, regulatory, or professional-sign-off conclusions as if generated text can replace credentialed actuarial judgment. Do not instruct the model to use confidential policyholder data, protected personal data, or proprietary model logic unless the user is authorized and the data is approved for that environment. Before answering, verify that each prompt has a concrete deliverable and that none asks the AI to fabricate experience data, assumptions, citations, regulations, or calculations. Ask up to 3 clarifying questions only if a required input is missing.
Fill in
| Placeholder | What to enter | Example |
|---|---|---|
| [ACTUARIAL WORK AREA] | Name the work area, such as life valuation, P&C reserving, health pricing, pensions, or ERM. | P&C personal auto reserving. |
| [JURISDICTION OR BUSINESS CONTEXT] | Provide the relevant market, regulator, product line, or internal business setting. | United States admitted carrier, quarterly internal reserve review. |
| [USER LEVEL] | Describe who will use the prompts, such as actuarial analysts, associates, or a mixed team. | Actuarial analysts and an ACAS reviewer. |
| [AVAILABLE MATERIALS] | List approved source materials, such as aggregated experience studies, model documentation, or policy wording. | Aggregated paid and incurred triangles, claim-count reports, prior quarter memo, approved model documentation. |
| [PRIORITY TASKS] | List the specific tasks the prompt pack should prioritize. | Data validation, development-method review, variance explanations, reserve committee memo, peer review. |
How to use
- Choose a work area and list only approved, de-identified materials you can use in an AI tool.
- Paste the full prompt and save the resulting prompts in your team’s approved knowledge base or working-paper format.
- For each output, verify calculations independently and trace statements back to source data or documentation.
- Follow up with: “Adapt prompt 4 for [SPECIFIC DELIVERABLE] and add the control evidence a reviewer would expect.”
Variations
Reserve review
Use when preparing a structured review of reserve indications.
Create a prompt for reviewing [LINE OF BUSINESS] reserve indications using [AVAILABLE AGGREGATED DATA]. It should request a workpaper outline covering data reconciliation, triangle diagnostics, selected methods, assumptions, sensitivity tests, actual-versus-expected comparisons, and changes from [PRIOR PERIOD]. Require separate sections for evidence, judgment, and open questions for the credentialed reviewer. Do not calculate or select an ultimate without supplied inputs. Include checks for immature accident periods, large losses, calendar-year effects, and inconsistent paid versus incurred patterns.
Assumption memo
Use when documenting an assumption for review or governance.
Write a prompt that turns [SOURCE EVIDENCE] into an actuarial assumption memorandum for [PURPOSE]. Require sections for the assumption definition, scope, source-data period, transformations, rationale, alternatives considered, sensitivity, limitations, governance owner, and monitoring triggers. Ask the AI to distinguish facts from judgment and flag unsupported causal explanations. The memo must not state regulatory compliance or professional approval. Include placeholders for [JURISDICTION], [MODEL OR PROCESS], and [REVIEW DATE].
Executive briefing
Use when a non-technical leader needs decision-ready context.
Create a prompt to explain [ACTUARIAL RESULT] to [EXECUTIVE AUDIENCE] for a decision about [DECISION]. Use only these supported facts: [FACTS AND UNCERTAINTY]. Request a one-page briefing with a headline, three key drivers, a range or uncertainty statement, decision implications, and questions requiring management judgment. Avoid technical formulas unless essential, avoid false precision, and do not imply the analysis guarantees an outcome. Require a final check that recommendations are clearly separated from observed results.
Tips
- Use AI first for structure, challenge questions, and draft explanations; keep data extraction, method selection, calculations, and sign-off inside your approved actuarial controls.
- Provide data definitions and valuation dates, not just tables; a triangle without its basis, treatment of reopenings, and reconciliation context invites wrong conclusions.
- Ask for alternative explanations of material movement, then validate each against evidence before including it in a reserve or pricing narrative.
- Keep a record of the prompt, source version, output, edits, and reviewer decisions when AI-assisted text enters a formal workpaper.
FAQ
Can ChatGPT calculate reserves or set assumptions?
It can help organize a calculation approach or identify tests, but it should not replace controlled models, validation, or actuarial judgment. A qualified actuary remains responsible for selections and conclusions.
Can I paste policyholder or claim data into an AI tool?
Only if your employer has approved that tool, data classification, and use case. Prefer aggregated, de-identified extracts and follow privacy and information-security rules.
Are these prompts limited to 2024 work?
No. The search phrase is retained, but the prompts are designed for current work and should be tailored to current regulations, documentation standards, and internal controls.