Complete AI Training
Sign inGet my AI kit

Your job's AI kit

Get your AI kit

Tell us who you are and what you do. We show you your kit right away and email you the link: skills, prompts, AI agents, MCP servers and courses for your job.

500+ jobs ready, and we make a kit for any other job. No payment needed to look.

Share

AI agent for insurance actuaries

Actuarial Model Change Control Agent

Show exactly what a model change does to results before it goes live.

Actuarial Model Change Control Agent: what goes in, what the agent does and what you get

What it does

A model update can quietly change reserves or prices, and without testing, nobody knows by how much until a number looks off. This agent reads a proposed model change and reruns prior period results with the old and new versions. It compares outputs line by line, ranks the differences and explains each one in plain words, tying it to the change. It then checks that the documentation, version notes and test records were updated. If a difference cannot be explained by the change, it loops back, narrows the change and reruns to isolate the cause. The actuary approves the release.

How it works

Follow the arrows from top to bottom. The orange dashed arrow is the loop: when a check fails, the agent goes back and tries again.

Start and resultWhat it doesA check on its own workWaits for your OKGoes back and retries
Yes, continueYes, continueApprovedNoNo 1 STARTS WHEN Model change submitted 2 USES A TOOL Read the change request and list affected components 3 USES A TOOL Run prior period inputs through the old version 4 USES A TOOL Run the same inputs through the new version 5 DOES Compare outputs and rank differences 6 CHECKS THE RESULT Is every difference above threshold explained by thechange? If not: isolate the cause by testing components one at atime and rerun. Back to step 2. 7 DOES Write an explanation for each difference 8 CHECKS THE RESULT Are documentation, version notes and test recordsupdated? If not: list what is missing and return it to the modelowner. Back to step 7. 9 DOES Draft the change control summary 10 YOU APPROVE Actuary approves release 11 RESULT Change control record with results
Read the steps as a list
  1. Model change submitted
  2. Read the change request and list affected components
  3. Run prior period inputs through the old version
  4. Run the same inputs through the new version
  5. Compare outputs and rank differences
  6. Is every difference above threshold explained by the change?If not: isolate the cause by testing components one at a time and rerun. Back to step 2.
  7. Write an explanation for each difference
  8. Are documentation, version notes and test records updated?If not: list what is missing and return it to the model owner. Back to step 7.
  9. Draft the change control summary
  10. Actuary approves releaseThe agent waits here for your OK.
  11. Change control record with results

How it decides

It accepts the change when every difference above the threshold is explained by the stated change and documentation is updated.

  • Investigate any output difference above 0.5 percent
  • Test at least the last 2 reporting periods
  • Require updated documentation before release
  • Isolate causes by changing one component at a time

Make it yours

Every agent is a starting point. You choose these settings for your own situation.

  • Difference threshold (default 0.5 percent)
  • Periods to test
  • Documents required for release
  • Model list

What keeps you in control

It always asks you first

  • Actuary approves release of the new version

Hard limits

  • Never releases a model to production
  • Keeps old and new results side by side

It stops when

  • Done: differences explained and release approved
  • Stop: prior period inputs cannot be reproduced with the old version

Set it up

We guide you through the set-up, step by step

Members get the full set-up guide for this agent. No technical skills needed: you copy, paste and upload.

10 minto set it up in your AI
5 AIsChatGPT, Claude, Copilot, Gemini, Grok
  • One set of instructions to paste into your AI, with the clicks for ChatGPT, Claude, Microsoft 365 Copilot, Gemini and Grok
  • The agent then walks you through connecting your own data, one source at a time
  • A downloadable copy with the flow chart, the rules and the full guide
Get access to this agent

An example run

What happensA change to the lapse table was meant to move reserves by 0.3 percent. The run showed 1.2 percent. The explanation check failed. Testing components one by one showed a changed rounding rule in the discount curve caused 0.8 percent. The model owner confirmed it was unintended and reverted it. The rerun showed 0.3 percent, documentation was updated and the actuary approved.

More agents for insurance actuaries