You’re employing an LLM to automate the generation of email responses for a customer service...

This is a lifecycle problem, not a wording problem, and Option A gives the team a controllable lifecycle for the agent behavior. A detailed response-composition prompt forces the model to address intent, structure, and tone. Vague “be helpful” language does not bind the output to the customer’s actual concern. The runtime should therefore be built around a prompt contract that tells the model what to extract, which evidence to preserve, and what output format is valid. The selected option specifically A states “Instructing the LLM with a detailed prompt containing instructions on how to format and compose the response in an easy-to-understand structure.”, which matches the operational requirement rather than a superficial wording match. The alternatives would look simpler in a prototype, but asking for final accuracy alone hides whether the intermediate decomposition was valid. For a production build, prompt design is still an engineering control when it defines extraction targets, tool names, parameter examples, and evaluation rubrics. The answer is therefore about engineered control planes, not simply model capability.