How the AI works

Claude is the intelligence in the system. How you structure its work is a set of decisions in their own right: whether a request runs now or in a batch, whether it is a single call or an agent, what it can reach and the safeguards and human checks around it.

The first of those decisions comes before any of the others: whether a part of the job needs judgement at all, or whether it has a right answer and wants ordinary code instead. That one is worked through in When to use AI and when not to, along with what each option really costs. The whole set is gathered in Prompt engineering with Claude.

Part of Designing the system.