Giving AI a Role

The "Act As" Technique

One of the most effective and widely used prompting techniques is role prompting — assigning the AI a specific persona before asking it to complete a task. You have probably seen examples like "Act as a marketing expert" or "You are an experienced copywriter." This is not a gimmick. It works, and understanding why helps you use it much more effectively.

Why Roles Change Output

When you assign a role, you are giving the AI a lens through which to interpret your request. A role shifts the vocabulary, the depth of expertise, the assumptions the AI makes about what you need, and often the structure of the response.

Consider asking for advice on a business problem. The same question sent to "an experienced accountant" versus "a startup founder who has bootstrapped three companies" will produce meaningfully different answers — different priorities, different language, different framing. Both are valid. The role tells the AI which version to produce.

Being Specific Makes All the Difference

Generic roles produce generic output. The more specific your role description, the more distinctive and useful the response.

• Generic: "You are a marketing expert."
• Specific: "You are a direct response copywriter with fifteen years of experience writing email campaigns for e-commerce brands. You write copy that is clear, benefit-focused, and always leads with the reader's problem."

The specific version gives the AI a fully formed character to inhabit — not just a job title, but a point of view, a style, and a set of priorities. The output will reflect all of that.

What Roles Are Good For

Role prompting is particularly powerful in a few situations:

• Expertise you don't have — "Act as a nutritionist" when asking for a meal plan; "Act as an experienced HR manager" when drafting a difficult email to a team member
• Tone control — "Act as a patient teacher explaining this to a complete beginner" vs. "Act as a senior developer reviewing this code"
• Perspective-taking — "Act as a sceptical investor and critique this business idea"
• Writing style — "You write in a warm, conversational style with short sentences. Avoid jargon."

Combining Role with Task and Context

A role alone is not a complete prompt. Its power comes from combining it with your task and context. Here is what that looks like:

"You are an experienced UX writer who specialises in SaaS onboarding. I am building a project management app for solo freelancers. Write the welcome message they see the first time they log in — warm, motivating, and under fifty words. Avoid corporate language."

The role (UX writer, SaaS onboarding specialist) + task (welcome message) + context (solo freelancers, first login) + constraints (under 50 words, no corporate language) produces output that would have taken a professional thirty minutes to draft.

You Can Also Give Yourself a Role

Role prompting does not only apply to the AI. You can also tell the AI who you are — which is a form of context that shapes how the AI responds to your questions.

• "I am a first-year law student. Explain this contract clause in simple terms."
• "I run a small e-commerce store and have no accounting background. Walk me through what VAT means for my business."

By telling the AI your level of expertise and your situation, you get explanations pitched at exactly the right level — not too basic, not too technical.

One Caution: Keep Roles Realistic

Roles work best when they are grounded in real-world expertise. "You are the world's greatest copywriter who has never failed" is weaker than "You are an experienced direct response copywriter." Grounded, specific roles produce more consistent, useful output than inflated or fictional ones.