System Prompt: Sales Call Summariser

Use this after discovery calls, demo calls, or founder-led sales conversations when somebody needs the commercial takeaways without rereading the whole transcript.

It is particularly useful for small teams where the same call needs to inform sales follow-up, pipeline updates, and product or marketing feedback.

Sales summaries go bad when the model collapses everything into a generic recap. This prompt forces extraction by commercial category instead: goals, objections, buying signals, stakeholders, and next steps.

The fact-versus-inference distinction matters because sales teams often mistake model confidence for buyer intent. This prompt makes uncertainty explicit.

Add your sales methodology if you use one. MEDDICC, BANT, SPICED, or a custom qualification framework all fit well here.

If the output feeds a CRM or automation tool, tighten the structure into named fields so downstream updates are easier.

This should not replace human judgement on deal quality. It is a compression layer, not a forecast model.

Short or low-quality transcripts also create false precision. A missing objection in the transcript is not proof that the prospect had none.

Claude is strong at separating explicit statements from softer inference.

GPT works well if you want the output shaped into CRM-friendly fields. Model-agnostic overall, but sales taxonomy should be made explicit.