Role (Persona) Prompting

Ever notice how the same model can sound like a brilliant teammate… or a vague motivational poster?

A simple fix is role (persona) prompting: you tell the LLM who it should be before you tell it what to do.

Role prompting works because it narrows the model’s “style and solution space.” Instead of guessing which tone, depth, and format you want, the model anchors to an archetype: senior engineer, sales coach, HR partner, legal reviewer, executive communicator, and so on.

What to include in a strong persona

A good role prompt usually contains:

• Seniority (junior vs senior vs principal)
• Domain (backend, marketing ops, recruiting, accounting)
• Priorities (speed, correctness, compliance, customer empathy)
• Output preferences (bullets, tables, JSON, step-by-step)

Example 1: Non-technical (Sales coach)

You are a SaaS sales coach.
Write a follow-up email after a first call.
Audience: VP of Operations.
Goal: book a 20-minute demo.
Constraints: 90–120 words, confident but not pushy, include one CTA.

Example 2: Technical (Senior Python engineer)

You are a senior Python engineer.
Review the code below for performance and correctness.
Return:
1) Top 3 issues (ranked)
2) A minimal patch
3) One regression test
Only output bullets and code blocks.

Code: <PASTE HERE>

Takeaway

Role prompting is one of the fastest ways to improve LLM output quality. Pick a persona that matches your use case, specify what they value, and combine it with clear constraints.

When you do, the model stops sounding “generic” and starts sounding like the coworker you actually wanted.