Question

A pharmaceutical company deploys a generative AI assistant that answers the same drug-safety questions for two audiences: licensed physicians who need dense clinical detail, and patients who need plain-language explanations. The underlying knowledge is identical and already grounded correctly. Leadership wants the fastest, lowest-effort way to tailor the tone and depth of each response to the intended audience without retraining or hosting separate models. Which approach best meets this requirement?

Accepted Answer

Include a role/persona instruction in the system message that defines the target audience and required reading level for each request

Answer

Fine-tune two separate models, one on physician-oriented text and one on patient-oriented text

Answer

Rebuild the retrieval-augmented generation index with two separate document collections per audience

Answer

Increase the model temperature so responses vary enough to suit different readers