API testing lessons from the Insurance Domain
summary
Argues that traditional Given-When-Then API tests fail for insurance rating and shows a working prototype: a simulator that emits step-by-step calculations which runs alongside API calls (Karate can be the client), compares expected vs actual, and produces a git-diff view when numbers change. Demonstrates using Claude to read a rules PDF and produce AI-generated JavaScript calc checked into Git; injecting a backend rate change breaks tests, the simulator pinpoints the offending lookup, and reruns pass.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 3:40Why testing in complex domains
- 6:53Insurance and tax complexity
- 8:00Insurance rating use case
- 11:57Business user testing
- 13:03No-code test data form
- 15:10API response and costs demo
- 16:59Business rules simulator
- 18:43Regression testing and batch
- 19:25Debugging API failures
- 21:12Simulator debugging output
- 22:37AI for business rules
- 24:41BDD with failure analysis
- 27:21Recap of testing approach
- 28:44AI recreates business rules
- 31:36AI test scenarios and data
- 33:23Business rules in JavaScript
- 35:07Diff view for impact
- 36:11Deterministic AI testing
- 37:36Q&A
related talks
54mAutomated Tests: What Courses Won't Teach You
54mShould You Be Doing More API Testing?
43mAI-Generated API End-to-End Tests: A Legacy Rewrite Case Study
31mExploratory Testing on Computer Interfaces (APIs)
47mNext generation of API testing : Unleashing the power of AI
43mExploratory Testing with AI
44mAPI Testing following the Test Pyramid
29mBeyond OpenAPI and AsyncAPI: AI agents wrote the spec, code, and tests - now w…