When AI Writes the Code, Who Verifies the Logic?

What happens when a non-developer uses AI to build a mission-critical financial verification engine, reverse-engineers transaction formats across 22 banks, and tests it against 69,000+ real transactions?

This talk shares the unfiltered story of building a production transaction verification pipeline with AI coding tools — and the crucial realization that while AI can generate syntactically correct code effortlessly, validating the underlying domain logic and edge cases remains the real challenge.

Key takeaways:

• Where AI-generated code excels and where domain assumptions silently break
• Practical techniques for designing verification suites and fraud-detection logic
• The evolving role of engineers and product managers in an AI-assisted development era