How assigning a role or persona to AI dramatically improves output quality. The psychology behind it, best practices, and when it backfires.
"You are a senior software engineer" produces fundamentally different code than a bare instruction. Role prompting — assigning a persona to the AI — is one of the simplest and most effective prompt engineering techniques. Here's why it works and how to do it well.
When you say "You are a senior data scientist," several things happen inside the model:
You are a marketing expert. Help me write ad copy.
Better than nothing, but still vague.
You are a direct-response copywriter with 15 years of experience
writing Facebook ads for DTC e-commerce brands.
Much better — activates specific expertise.
You are Sarah, a VP of Marketing at a $50M DTC skincare brand.
You've grown the brand from $2M to $50M primarily through
Meta ads and influencer partnerships. You're known for your
data-driven approach and your ability to write hooks that
stop the scroll. Your team calls you "the conversion queen."
The most effective — creates a consistent, detailed persona.
Bad: "You are a financial advisor." Good: "You are a CFP who explains financial concepts using everyday analogies. You never use jargon without explaining it. You present options as 'if you value X, do A; if you value Y, do B.'"
A "world-class expert" role might make the AI sound authoritative about things it's wrong about. Add: "If you're uncertain or the question is outside your expertise, say so directly."
"You are Shakespeare" produces Shakespearean language but not necessarily good content. The role should enhance the output, not become the output.
"You are a creative maverick who follows all the rules" creates tension. Pick one primary archetype.
"You are an expert at basic arithmetic. What is 2+2?" — the role adds nothing for trivial tasks.
Consider this business problem from three perspectives:
1. As a CFO focused on financial impact
2. As a CTO focused on technical feasibility
3. As a CMO focused on market opportunity
For each perspective, give your assessment and recommendation.
Then synthesize all three into a unified recommendation.
First, present the strongest case FOR this strategy as a
supportive advisor. Then, switch roles: present the strongest
case AGAINST it as a skeptical board member.
Finally, as a neutral analyst, weigh both sides.
The most effective prompts combine a role with output format:
You are a McKinsey senior partner presenting to a board.
Format your response as:
1. Executive summary (3 sentences)
2. Key finding (supported by data)
3. Recommendation (specific, actionable)
4. Risk (and mitigation)
Use the pyramid principle: conclusion first, then supporting evidence.
The role shapes WHAT the AI knows. The format shapes HOW it communicates. Together, they produce output that's both expert and usable.
| Need | Role to Use |
|---|---|
| Better code | "Senior staff engineer at Google" |
| Better writing | "NYT editor with 20 years experience" |
| Better analysis | "McKinsey senior partner" |
| Better creativity | "Award-winning creative director at Wieden+Kennedy" |
| Better teaching | "MIT professor known for making complex topics simple" |
| Better feedback | "Veteran book editor who's brutally honest but constructive" |
Browse our System Prompt collection for 40+ ready-to-use role prompts covering every domain.
Sign in to join the discussion.
No comments yet. Share your thoughts on this article.