USE CASE
The scenario: A rep has a high-stakes call with a skeptical buyer. Silent rehearsal does not test whether the rep can listen, respond naturally, and recover when the conversation changes direction.
What they do: In a new Claude chat, the rep provides the buyer role, deal stage, known facts, likely concerns, meeting objective, and a short scoring rubric. They activate Claude Voice mode, ask Claude to stay in character, and run six to eight conversational turns. Claude may challenge weak answers, but it must not invent undisclosed facts about the real buyer or account.
Scoring: After the role-play—not during it—the rep asks for a 1–5 score on listening and relevance, discovery depth, objection handling, clarity, evidence use, and next-step control. Feedback must cite specific lines from the practice transcript, separate observed behavior from coaching inference, and recommend the two highest-value changes for the next attempt.
Important boundary: Voice mode is a practice surface, not a buyer simulator with privileged knowledge. Use a fictionalized persona when real account details are sensitive, never imitate a real person's voice, and follow your organization's AI and data policy. Claude Voice mode is a beta capability, so availability and admin controls should be checked before a team rollout.
The result: The rep enters the real meeting having practiced aloud, seen where responses broke down, and repeated the hardest moment with a concrete coaching target.
A practical AI workflow for B2B sales — the prompts below run this play end to end, in Claude, ChatGPT, Gemini, or Copilot.
2,900+ TESTED PROMPTS · 15 SALES STAGES
Every prompt in the library is vetted, tested, and built for a real sales job — like this one.
Browse the libraryFree to browse. Pro from $12/month.