Artikel 60
llm

The Role Technique: Expertise at the Push of a Button

Sebastian Rydz13. März 202610 min Lesezeit

Imagine having access to an expert anytime you need one. An experienced lawyer for contract questions, a creative copywriter for your website, or a patient teacher for complex topics. That is exactly what the role technique makes possible. With a single sentence, you transform your AI into a specialist who responds from a specific perspective. Sounds too simple? In this article, I will show you why it works, how to use it correctly, and where its limits lie.

In the previous article, we explored how to solve technical problems. Now we dive into a technique that takes your prompts to an entirely new level. And in the next article, we will look at how you can achieve even better results by providing concrete examples.

Why the Role Technique Works So Well

Large language models like ChatGPT were trained on billions of texts. Among them are textbooks, scientific papers, blog posts, forum discussions, and much more. All of this knowledge is embedded in the model, but it is not always automatically retrieved. When you ask a general question, you get a general answer. But when you say "You are an experienced tax advisor," you specifically activate the knowledge areas related to tax consulting.

Think of an actor stepping into a role. They change their language, their posture, their perspective. Exactly the same thing happens with AI. By assigning a role, you change the entire context in which the AI generates its response. The result: more precise, more professionally grounded, and more relevant answers.

Why is this so effective? Because with a role assignment, you set three things simultaneously:

  • The knowledge domain: The AI focuses on a specific field of expertise.
  • The language style: A scientist formulates differently than a comedian.
  • The perspective: A critic views things differently than an advocate.

A simple experiment illustrates this. Ask the same question once without a role and once with a role:

Without role: "What should I pay attention to in an employment contract?"

With role: "You are an experienced employment lawyer with 20 years of professional experience. What should I pay attention to in an employment contract?"

The second version delivers significantly more detailed, professionally deeper answers with legal nuances that would be missing in the first version. This is not a coincidence but the result of targeted context control.

The Basic Formula: How to Use Roles Correctly

The role technique follows a simple basic formula that you can easily remember:

"You are a [role] with [experience/specialization]. [Your actual task]."

The more detailed your role description, the better the result. Compare these three variants:

Variant 1 (vague): "You are a copywriter. Write me an advertising text."

Variant 2 (better): "You are an experienced advertising copywriter. Write me an advertising text for an organic granola."

Variant 3 (best): "You are an experienced advertising copywriter specializing in sustainable food brands with a casual, modern writing style. Your target audience is health-conscious people between 25 and 40 years old. Write an advertising text for a new organic granola that is regionally produced."

Can you see the difference? With each additional piece of information, you give the AI more context, and context determines the quality of the answer. Here are some elements you can incorporate into your role description:

  • Professional experience: "with 15 years of experience in..."
  • Specialization: "specialized in..."
  • Target audience: "you explain things for beginners / experts / children"
  • Style: "you communicate casually / scientifically / humorously"
  • Context: "you work in a startup / a government agency / a hospital"

The combination of these elements makes your role instruction precise and effective. You control not only what the AI knows but also how it communicates it.

The Best Roles for Typical Tasks

Certain roles have proven particularly effective in practice. Here is an overview of the most useful roles for different task areas:

For content creation:

  • "You are an experienced SEO copywriter" for search engine optimized texts
  • "You are a bestselling author" for compelling stories
  • "You are a social media manager" for posts and captions
  • "You are a technical writer" for manuals and documentation

For consulting and analysis:

  • "You are a management consultant at McKinsey" for strategic analyses
  • "You are a financial advisor" for money topics
  • "You are an experienced HR director" for application tips
  • "You are a market researcher" for target audience analyses

For learning and explaining:

  • "You are a patient teacher" for simple explanations
  • "You are a university professor of physics" for scientific depth
  • "You are a tutor for elementary school students" for child-friendly explanations

For creativity:

  • "You are a Creative Director at an advertising agency" for campaign ideas
  • "You are a screenwriter" for storytelling
  • "You are a UX designer" for user experiences

For programming and technology:

  • "You are a senior software developer" for clean code
  • "You are an IT security expert" for security questions
  • "You are a database administrator" for database design

Important: The role must match the task. An advertising copywriter will not give you good legal advice, just like in real life. Always choose the role that fits what you want to achieve.

Perspective Switching: Using Multiple Roles for One Topic

One of the most powerful applications of the role technique is perspective switching. You examine one and the same topic from different angles by assigning different roles. This helps you identify blind spots and make more informed decisions.

Imagine you are planning a new product. You could have the AI step into the following roles one after another:

The Optimist: "You are an enthusiastic startup founder who believes in the product. What opportunities do you see?"

The Critic: "You are a skeptical investor with years of experience. What risks and weaknesses do you see?"

The Customer: "You are a typical end customer in the target group. What would you expect from this product? What would put you off?"

The Competitor: "You are the CEO of a competing company. How would you react to this product?"

Through these four perspectives, you get a much more complete picture than if you had only sought a single opinion. You recognize opportunities and risks, understand the customer perspective, and anticipate market reactions.

Another example: You want to write a blog article. You could ask the AI to evaluate your draft from three perspectives:

  • As an SEO expert: Are the keywords properly placed? Is the structure correct?
  • As a reader from the target audience: Is the text understandable and engaging?
  • As a strict editor: Are there linguistic weaknesses or logical gaps?

Perspective switching is also extremely useful in everyday life. In conflicts, you can ask the AI to analyze the situation from the viewpoint of all parties involved. For decisions, you can simulate different advisors. You are using a technique known in the real world as "Six Thinking Hats" by Edward de Bono, only digitally and immediately available.

Combining and Refining Roles

Advanced users combine multiple attributes in a single role or refine them during the conversation. Here are some techniques for doing this:

Combining roles: You can merge multiple areas of expertise into one role. For example: "You are a marketing expert who also studied psychology and has expertise in behavioral economics." This produces answers that connect marketing knowledge with psychological insights.

Roles with constraints: Sometimes it helps to set boundaries for the role: "You are a nutritionist who never recommends fad diets and always argues based on scientific evidence." This way, you control not only what the AI does but also what it avoids.

Adjusting the role mid-conversation: You can switch or refine the role in the middle of a conversation. For example: "Now please switch to the role of a critic and question everything you just said." This creates a self-critical analysis of its own statements.

Connecting role with target audience: "You are a financial advisor who explains complex topics so that even a 16-year-old can understand them." The combination of expert role and target audience ensures professionally correct yet easily understandable answers.

Role with format specification: "You are a business coach who always structures answers in three clear steps and concludes with a practical example." This way, you influence both the content and the form of the answer simultaneously.

Feel free to experiment with different combinations. The better you define the role, the more accurate the results become. It is like searching for the perfect tool: Sometimes you need not just a hammer, but a hammer with a rubber head for delicate surfaces.

The Limits of the Role Technique: An Honest Look

As powerful as the role technique is, it also has limits. It is important to know these so you can use the technique realistically:

The AI does not become a real expert: Even if you assign the AI the role of a doctor, it does not replace a real doctor. The AI simulates expert behavior based on training data. It has no real professional experience, no intuition, and no sense of responsibility. For medical, legal, or financial questions, you should always have the answers reviewed by a real professional.

Facts can be wrong: Role assignment improves the style and structure of the answer, but it does not protect against incorrect facts. An "AI historian" can still cite wrong dates. Always verify important facts independently.

Over-specific roles can be limiting: If you define the role too narrowly, the AI may overlook relevant aspects that lie outside that role. A "neurologist" will not suggest psychosomatic causes for your headaches, even though they might be relevant.

The role does not change the underlying model: The AI retains the same strengths and weaknesses regardless of the assigned role. It does not suddenly become better at mathematics or gain knowledge of current events beyond its training date.

Ethical boundaries: The AI will refuse to fill certain roles that violate its usage guidelines. This is a good thing and protects against misuse.

Despite these limitations, the role technique remains one of the most effective tools in your prompting toolkit. You just need to know where the limits are and evaluate the results accordingly.

The Role Technique in the Prompt Generator

At optiprompt.io, you will find our free prompt generator, which already has the role technique integrated. You do not have to formulate a role description from scratch every time because the generator supports you in this process.

Here is how to use the prompt generator with the role technique:

  1. Open the prompt generator at optiprompt.io.
  2. Select your task or describe what you need.
  3. The generator suggests suitable roles based on your task.
  4. Customize the role: You can adjust experience level, specialization, and style.
  5. Generate your prompt and copy it directly into your AI tool.

The prompt generator automatically combines the role technique with other proven prompting methods that we discuss in this article series. This gives you optimally structured prompts without having to think about every detail yourself.

Especially practical is the ability to try different roles for the same task. You can see how the generated prompt changes when you switch from "expert" to "beginner-friendly explainer." This helps you develop a feel for which role works best in which situation.

Exercise: Solve the Same Task with Three Different Roles

Now it is your turn! This exercise impressively demonstrates how strongly the role influences the result. All you need is an AI tool of your choice and about 15 minutes.

Your task: Have the AI explain why regular exercise is important. Use these three roles one after another:

Round 1: The Sports Medicine Doctor

"You are an experienced sports medicine doctor with 20 years of practice at a university hospital. Explain to me why regular exercise is important. Focus on the physiological mechanisms."

Round 2: The Motivational Coach

"You are a charismatic motivational coach who can inspire people. Explain to me why regular exercise is important. Motivate me to start right away."

Round 3: The Grandmother

"You are a loving grandmother giving advice to her grandchild. Explain to me why regular exercise is important. Speak from your life experience."

Compare the three answers: You will notice that the content, tone, level of detail, and argumentation differ significantly. All three answers are correct, but they address different needs. The sports doctor delivers facts, the coach delivers motivation, and the grandmother delivers emotional warmth.

Bonus task: Go to optiprompt.io and use the prompt generator to create an optimized prompt for each of the three roles. Compare whether the generator prompt delivers better results than your manually formulated one.

Reflection: Which role delivered the most useful answer for this task? In which situation would you use which role? Write down your insights because you will continue to build on this knowledge in the upcoming articles and exercises.

Summary and Outlook

The role technique is one of the most effective and simultaneously simplest methods for getting better results from AI tools. With a sentence like "You are an experienced..." you specifically activate the appropriate knowledge and style. Through perspective switching and role combinations, you unlock even deeper insights.

The key takeaways from this article:

  • A role simultaneously sets the knowledge domain, language style, and perspective.
  • The more detailed the role description, the better the result.
  • Different roles for the same task deliver different but valuable perspectives.
  • Roles can be combined, refined, and adjusted during the conversation.
  • The technique has limits: The AI does not become a real expert, and facts must be verified.
  • The prompt generator at optiprompt.io helps you find the perfect role for your task.

In the next article, "Working with Examples: Showing Instead of Explaining," you will learn another technique that makes your prompts even more precise. Instead of just telling the AI what to do, you show it with concrete examples what the result should look like. The combination of the role technique and examples is particularly powerful. Until then: try out different roles and discover which experts are hidden inside your AI!

Autor

Sebastian Rydz

Das OptiPrompt Team teilt Wissen und Best Practices rund um KI und Prompt Engineering, um dir zu helfen, bessere Ergebnisse mit KI-Modellen zu erzielen.

Bereit, deine Prompts zu optimieren?

Erstelle mit OptiPrompt professionelle Prompts in Sekunden – kostenlos starten.