Prompt and Prompt Engineering
What is Prompt and Prompt Engineering?
A prompt is the instruction and context given to a generative AI model. Prompt engineering is the practice of designing those instructions so the model produces accurate, consistent and on-brand results.
A production prompt usually contains more than the user’s question: a role and tone (“you are a support assistant for…”), rules (“answer only from the passages provided; if they do not cover it, say so”), the relevant knowledge or customer data, the format the answer must take, and sometimes worked examples.
Small changes in a prompt can change results a great deal, so in a contact center prompts are treated like any other business configuration: versioned, tested against a set of real questions, reviewed for compliance and changed under control. Many platforms hide prompts from agents entirely and expose them only as well-defined actions such as “summarize this call” or “draft a reply”.
Related: Generative AI, AI guardrails, Large language model (LLM).