Before you describe what you want, tell Claude who it is. This isn't a magic trick — it's a fundamental shift in how the model frames its response. A role assignment primes the model to draw on domain-specific knowledge, vocabulary, and conventions appropriate to that role.
Without role: "Write a cold email to a VP of Sales."
With role: "You are a B2B sales copywriter specializing in cold outreach for enterprise software companies. Write a cold email to a VP of Sales..."
Large language models are trained on vast amounts of text across thousands of domains. When you assign a role, you're effectively telling the model which subset of that knowledge to draw from. A prompt that starts with "You are a senior negotiation strategist" produces different output than the same request without that framing — the role-assigned output tends to be more specific, more tactical, and more consistent with how a real expert in that domain would approach the problem.
For sales tasks, the most effective role assignments are specific rather than generic. "You are a sales rep" is less effective than "You are a senior enterprise account executive with 15 years of experience selling complex B2B software to Fortune 500 companies." The more specific the role, the more calibrated the output.
A reliable formula is seniority + domain + orientation: "You are a senior enterprise AE with [YEARS] selling [SOLUTION] to [BUYER]. You prioritize [WHAT THIS BUYER CARES ABOUT, e.g. risk and compliance language]."
The Promptifi prompts all include role assignments — this is one of the reasons they produce better outputs than freeform requests on the same task.
If you use Claude, Promptifi's Claude Skills are ready-made role assignments: paste a Skill's system prompt into a Claude Project instead of writing the role from scratch.