There is a step in the development process for large language model (LLM)-assisted tooling that most teams skip because it's tedious, time-consuming, and doesn't produce results visible to end users: Verifying that what the model is saying is actually correct. Not fluent, not coherent, not