When you give the AI a Role, you are telling it who to be for this conversation. You are essentially casting it in a part — like a director telling an actor what character to play.
This works because AI has been trained on text written by all sorts of different people — teachers, doctors, comedians, marketing professionals, scientists, journalists. When you tell it to adopt a specific role, it draws on all the text written by people in that role and responds accordingly.
The result? A completely different tone, vocabulary, level of expertise, and style — all from the same tool.
🎭
Think of it like this
If you asked a random stranger on the street for fundraising advice, you would get a vague answer. If you asked someone who had spent 20 years helping charities raise money, you would get something far more useful. Giving the AI a Role is like choosing which expert you want to speak to.
🔄
Same task — completely different results
Here is the same basic task — "help me write a social media post about my fundraiser" — given to the AI with three different Roles. Notice how different each response would be:
Role: Enthusiastic youth worker
"Hey everyone!! 🙌 I'm on a mission to raise £600 for my DofE expedition and I need YOUR help! Every £5 gets me closer to the Lake District — link in bio to donate. Let's do this!! 🏔️"
Role: Professional fundraising copywriter
"This summer, I'm taking on my biggest challenge yet — a Duke of Edinburgh expedition through the Lake District. I'm raising £600 to make it happen. Your support, however small, makes a real difference. Donate here: [link]"
Role: Storytelling journalist
"Six months ago, I signed up for something that terrified me. Now I'm 80% of the way to making it happen. My DofE expedition to the Lake District is real — and with your help, I'll be lacing up my boots this summer. Here's how you can be part of the story."
Same task. Same person. Same expedition. Three completely different posts — each suited to a different audience and purpose. None of them are wrong — they are just different tools for different jobs.
✍️
How to write a good Role
A good Role has two parts: who the AI is, and who it is helping. You do not need both every time, but including both usually gives better results.
1
Who the AI is: Give it a specific job title or area of expertise. "A fundraising expert" is better than "an expert." "A secondary school teacher who explains things simply" is better than "a teacher."
2
Who it is helping: Tell it about you. "Helping a 14-year-old who has never done this before" changes the level of explanation it will give. "Helping an experienced fundraiser" changes it again.
Examples of well-written Roles
✅ "You are a friendly fundraising coach who specialises in helping UK teenagers raise money for school expeditions."
✅ "You are an experienced letter-writer who helps young people communicate confidently with local businesses."
✅ "You are a social media manager who creates engaging content for teenagers aged 13–18."
✅ "You are a patient maths teacher who explains things step by step for students who find the subject difficult."
⚠️
Avoid being too vague
"You are an expert" tells the AI almost nothing. An expert in what? For whom? The more specific your Role, the more targeted and useful the response will be.
🧩
Quick check
🎯 Knowledge Check
Which of these is the best Role to give the AI if you want help writing a speech to deliver at a school assembly about your fundraising campaign?
"You are an expert"
"You are a public speaking coach"
"You are a public speaking coach who helps teenagers deliver confident, engaging speeches to their peers"
"You are a Hollywood screenwriter"
✏️ Write your own Role
Think about something you need help with for your fundraising — a letter, a social media post, a plan, a speech. Now write a Role for the AI that would help you with that task. There is no right or wrong answer — just practise being specific.
What do you need help with?
Now write your Role:
👀
Up next: T and D — Task and Detail
In Lesson 4, we cover the T and D together — because they work as a pair. Task is the what, Detail is the who, where, how much, and why. Getting these right is what turns a generic AI response into something you can actually use.