AWS found bugs in 60% of software requirements. Its fix isn’t more AI — it’s a 50-year-old logic engine.

AWS added Requirements Analysis to Kiro, its agentic development platform, to catch spec problems before code is generated.

The feature rewrites vague requirements, converts them into formal logic, and uses an SMT solver to find contradictions, ambiguities, undefined behavior, and missing details.

Instead of waiting for code bugs, Kiro surfaces issues early as quick questions for developers to answer.

The move highlights AWS leaning on formal methods and automated reasoning to improve AI-assisted coding accuracy.