Lesson 71 · 2 min

Give AI a role

Telling AI who to be changes what it notices, which words it uses and how deep it goes.

Ask "Is this contract fair?" and you get a general answer. Ask "You are a lawyer who protects freelancers. Is this contract fair to me?" and the answer starts pointing at payment terms, deadlines and who owns the work.

A role works like a filter. It tells the model which part of everything it learned to draw on, and what a good answer looks like for that kind of person. A teacher explains, an editor cuts, a skeptic hunts for holes.

Make the role specific. "An expert" adds little. "A primary school teacher explaining fractions to a nervous 9 year old" adds a lot, because it brings an audience and a goal along with it.

A role does not give the model real qualifications. A pretend doctor is still a chatbot. For health, money or legal decisions, use the answer to prepare good questions for a real professional.

Do it now

One question, three roles

  1. 1. Pick a small decision you are facing, like a purchase or a message you need to send.
  2. 2. Ask the same question three times, each time with a different role from the prompt below.
  3. 3. Notice what each role spots that the others missed.
You are [ROLE, e.g. a careful accountant / a friendly marketing coach / a skeptical customer].
Here is my situation: [DESCRIBE IT].
Give me your honest view in 5 bullet points, the way that person would.

Quick check

Which role is likely to give the most useful answer?

Which role is likely to give the most useful answer?

More on this

Words from this lesson

Role prompting
Starting a prompt by telling the AI who to act as, such as an experienced editor, to shape tone and focus.
Prompt
The instruction or question you give an AI. Clearer prompts with context usually get much better answers.