Part 3 · Lesson 1 of 4
⏱ 17 min read

The Art of Asking: Prompting Basics

📖 Lesson 3.1⏱ 17 min read🎯 Part 3: Getting Better Results

You've been using AI for two parts now. You've got results — some good, some a bit flat, some surprising. This part is about understanding why some prompts work brilliantly and others fall flat, and building the technique to get dramatically better results every time.

The difference between a mediocre AI user and a genuinely skilled one isn't intelligence or technical knowledge. It's the ability to frame a question well. That's a learnable skill — and this lesson is where you build it systematically.

Why prompt quality matters so much

AI doesn't read your mind. It responds to exactly what you write. When you give it a vague instruction, it makes assumptions — and those assumptions are often not what you had in mind. When you give it a precise, rich instruction, it has what it needs to produce something genuinely useful.

Think of it like briefing a very capable contractor. A good contractor can build almost anything — but only if you tell them what you want, in what style, by when, and for what purpose. The same contractor given a vague brief will produce something generic. Given a detailed brief, they'll exceed your expectations.

The CRAFT framework

CRAFT is the prompting framework at the heart of this course. It stands for Context, Role, Action, Format, and Tone. You don't need all five elements in every prompt — but understanding each one and when to use it is what separates good prompts from great ones.

ElementWhat it meansExample
ContextThe background — who you are, what the situation is, what matters"I'm preparing for a job interview at a small marketing agency. I have 5 years of experience but this would be my first management role."
RoleWho you want AI to be — an expert, a coach, a particular type of professional"Act as an experienced interview coach who specialises in helping people step up to management."
ActionThe specific task — what you actually want it to do"Help me prepare answers to the three most likely questions about managing people for the first time."
FormatHow the response should be structured — length, layout, structure"For each question, give me a concise suggested answer followed by a bullet point of what makes the answer strong."
ToneThe register and style — formal, warm, direct, simple, technical"Keep the answers confident but honest — I want to sound capable without overstating my experience."

Here's that full CRAFT prompt assembled:

Full CRAFT prompt — interview prep

I'm preparing for a job interview at a small marketing agency. I have 5 years of experience but this would be my first management role. Act as an experienced interview coach who specialises in helping people step up to management. Help me prepare answers to the three most likely questions about managing people for the first time. For each question, give me a concise suggested answer followed by a bullet point of what makes the answer strong. Keep the answers confident but honest — I want to sound capable without overstating my experience.

Compare that to the prompt most people would write:

❌ Typical first attempt

"Help me prepare for a job interview"

✅ With CRAFT applied

The full prompt above — specific, contextualised, structured. The response will be night-and-day more useful.

You don't always need all five elements

CRAFT is a toolkit, not a checklist. For simple tasks, two or three elements are enough. For complex, high-stakes outputs — something you need to be really good — all five are worth including.

  • Quick tasks: Action + Format is usually sufficient. "Summarise this in three bullet points."
  • Everyday writing: Context + Action + Tone. "I need to email my landlord about a leak. Polite but firm, about 120 words."
  • Complex, important outputs: All five. Use the full framework when the quality of the output really matters.
🎯
The most commonly missed element: Context

Most people include the Action — the thing they want done — but skip the Context. Context is what turns a generic response into one that fits your situation. Even one sentence of context ("I'm writing this for my elderly mother who has never used a smartphone") transforms what AI produces.

Negative constraints: telling AI what not to do

One of the most useful prompting techniques most people never use is the negative constraint — telling AI explicitly what to avoid. This is particularly valuable when you have a clear sense of what you don't want.

Using negative constraints

Write a short bio for my LinkedIn profile. I'm a freelance graphic designer with 8 years of experience, specialising in brand identity for small businesses. Do not use the phrases "passionate about", "results-driven", or "leverages". Don't start with "I". Keep it to 80 words and make it sound like a real person wrote it, not a CV.

Those "do not" instructions prevent AI from defaulting to the generic corporate phrases that make most LinkedIn bios instantly forgettable. Negative constraints give AI guardrails that result in more distinctive, personalised output.

🚀 Try this right now

Take something you've asked AI before and didn't quite get right. Rewrite the prompt using the CRAFT framework — add context it didn't have, specify the format you actually wanted, and include at least one negative constraint. Compare the result.

📌 Key takeaways from this lesson

CRAFT — Context, Role, Action, Format, Tone — is the framework that separates good prompts from great ones. You don't need all five elements every time, but you should be conscious of each one. Context is the most commonly missed element and has the biggest impact. Negative constraints ("do not use...") are powerful and underused.

Part 3 · Lesson 2 of 4
⏱ 14 min read

Going Back and Forth

📖 Lesson 3.2⏱ 14 min read🎯 Part 3: Getting Better Results

One of the biggest mistakes beginners make is treating each AI conversation like a single transaction: ask a question, receive an answer, close the tab. The most powerful AI work happens across a conversation — building, refining, redirecting, going deeper. This lesson is entirely about that skill.

AI remembers your conversation

Within a single conversation, AI holds everything that's been said in context. It knows what you asked before, what it answered, what you said about yourself, and what direction the conversation has been going. This means every follow-up message can build on everything that came before — without you having to repeat yourself.

This is the feature most people underuse. Instead of asking one perfect question, you can have a genuine conversation — exploring, clarifying, pushing further — exactly as you would with a knowledgeable person.

The five most useful follow-up patterns

These are the follow-up phrases that unlock more value from almost any AI conversation:

1. Make it different

Adjusting tone, length, or style

That's good but a bit formal. Can you rewrite it in a warmer, more conversational tone?

Adjusting length

Can you cut that down to about half the length while keeping the key points?

2. Go deeper on one part

Expanding a specific section

I'd like to know more about the third point you made. Can you expand on that specifically?

3. Give it new information

Adding context mid-conversation

Actually, I should have mentioned — I only have a budget of £200 for this. Does that change your recommendations?

4. Get alternatives

Requesting variations

Can you give me three different versions of that opening paragraph? I want to see different approaches before I decide.

5. Challenge it

Stress-testing the response

What are the weaknesses in the plan you just outlined? What could go wrong?

Starting a new conversation vs continuing an existing one

Know when to continue and when to start fresh. Continue when you're building on a topic and want AI to remember the context. Start fresh when you're moving to a completely different task — accumulated context from a previous conversation can sometimes subtly colour responses in ways that aren't helpful.

🔄
The iteration mindset

The best AI users don't expect a perfect response on the first attempt — they expect a good first draft and a productive conversation from there. Shift your mental model from "ask and receive" to "explore and refine" and your results will improve dramatically.

Using AI as a thinking partner

One of the most powerful and underused applications of conversational AI is as a thinking partner — someone to think out loud with, test ideas against, and explore options with before committing to a decision.

AI as a thinking partner

I'm thinking through a decision and want to use you as a sounding board. I'll explain the situation and I'd like you to ask me clarifying questions rather than immediately jumping to advice. Is that OK?

This prompt reframes the conversation from "give me answers" to "help me think". AI will ask probing questions, surface assumptions you haven't examined, and help you arrive at your own well-reasoned conclusion. It's remarkably effective for complex personal and professional decisions.

🚀 Try this right now

Pick any response AI has given you in any previous conversation — one you thought was decent but not quite right. Go back to it and use one of the five follow-up patterns above. Notice specifically how quickly a good response becomes a great one with just one well-crafted follow-up.

📌 Key takeaways from this lesson

AI conversations are iterative, not transactional. The five most useful follow-ups are: adjust it, go deeper, add new information, get alternatives, and challenge it. The iteration mindset — expecting a first draft, not a final answer — is what separates skilled AI users. AI as a thinking partner is one of the most underused applications.

Part 3 · Lesson 3 of 4
⏱ 16 min read

Giving AI a Role

📖 Lesson 3.3⏱ 16 min read🎯 Part 3: Getting Better Results

The R in CRAFT stands for Role — and it's the element that produces some of the most dramatic improvements in response quality. When you tell AI who to be, it doesn't just change the words it uses. It changes the depth of knowledge it draws on, the structure of its thinking, and the entire frame through which it approaches your question.

Why roles work

AI has been trained on text written by people across every profession, background, and area of expertise. When you give it a role, you're effectively directing it to draw on the patterns, vocabulary, thinking structures, and approaches associated with that expertise. It's not pretending to be something it isn't — it's focusing its capabilities through a specific lens.

Without a role, AI tends to give a general, well-rounded response. With a precise role, it gives you the response that a domain expert would give — more specific, more nuanced, more actionable.

Examples across different contexts

🩺 Medical context
"Act as a GP explaining this diagnosis to a patient. Use plain language, explain what it means day-to-day, and tell me what questions I should ask my doctor."
⚖️ Legal context
"Act as a UK solicitor explaining the key points of this tenancy agreement to someone renting for the first time. Flag anything unusual."
💰 Financial context
"Act as an independent financial adviser explaining the pros and cons of ISAs versus pensions for someone in their 40s saving for retirement."
✍️ Writing context
"Act as a professional editor. Review this piece and give me specific feedback on structure, clarity, and where I lose the reader's attention."
👩‍💼 Career context
"Act as a senior recruiter who has reviewed thousands of CVs. Tell me exactly what's weak about this CV and how to fix it."
🔧 Technical context
"Act as a patient IT support specialist helping someone who is not technical. My laptop is doing [this]. What do I try first?"

Making roles more specific

Vague roles produce generic responses. Specific roles produce expert ones. Compare:

❌ Vague role

"Act as a doctor and tell me about this medication."

✅ Specific role

"Act as a UK pharmacist speaking to a patient who has just been prescribed metformin for the first time. Explain what it does, common side effects to expect, what to watch out for, and when to contact their GP."

The specificity — UK pharmacist, speaking to a new patient, first prescription — shapes every aspect of the response: the vocabulary used, the level of assumed knowledge, the practical focus, even the advice about when to seek further help.

The critic role

One of the most useful and least obvious role applications is asking AI to be your critic. Most AI outputs are naturally positive and constructive — it tends to build on your ideas rather than challenge them. The critic role deliberately inverts this:

The critic role

Act as a tough but fair critic. I'm going to share an idea I'm excited about, and I want you to tell me everything that could go wrong, every weakness in my reasoning, and every assumption I might be making that isn't justified. Don't soften it — I need the honest version.

This is enormously useful for business ideas, creative projects, plans, and arguments you're preparing. It surfaces the objections you'll face before you face them.

The important caveat

Roles change the frame and focus of AI responses — they don't give AI actual qualifications or real professional authority. An AI asked to "act as a solicitor" is drawing on legal text it was trained on, not giving you actual legal advice. Always be clear in your own mind: AI in a role is a knowledgeable starting point, not a licensed professional.

🚀 Try this right now

Think of an area where you'd genuinely benefit from expert input — something you'd normally need to pay a professional to advise on, or something you've been researching on your own. Ask AI using a precise role. Notice specifically how the response differs from what you'd get without the role instruction.

📌 Key takeaways from this lesson

Roles direct AI to draw on specific expertise, vocabulary, and thinking patterns. Specific roles produce expert responses; vague roles produce generic ones. The critic role is particularly valuable for stress-testing ideas and plans. Roles change the frame of a response — they don't confer actual professional authority.

Part 3 · Lesson 4 of 4
⏱ 18 min read

Prompting Practice: Before & After

📖 Lesson 3.4⏱ 18 min read🎯 Part 3: Getting Better Results

The best way to embed prompting skills is seeing them applied to real situations — and practising them yourself. This lesson is structured differently from the others: it's a series of before-and-after transformations, showing weak prompts turned into strong ones, followed by your Part 3 exercise where you do the same thing with your own prompts.

Transformation 1: The career question

❌ Before

"How do I get a promotion?"

✅ After

"I'm a senior administrator in an NHS trust and I've been in this role for four years. My next step would be a team manager position. Act as a career coach who works with NHS staff. What are the three most important things I should be doing over the next six months to strengthen my case for promotion? Be specific and practical — I want concrete actions, not general principles."

What changed: Context (NHS, four years, specific next step), Role (career coach who works with NHS staff), Action (three most important things), Format (concrete actions not general principles), implied Tone (practical, direct). The response will be dramatically more useful.

Transformation 2: The health question

❌ Before

"Tell me about high blood pressure."

✅ After

"I'm a 58-year-old woman who has just been told my blood pressure is consistently around 145/92. My GP has suggested lifestyle changes before considering medication. Act as a patient health educator. What does this blood pressure reading actually mean for my health, what lifestyle changes have the strongest evidence behind them, and what questions should I ask at my follow-up appointment? Please don't give general health advice — focus on what's relevant to my specific situation."

What changed: Specific numbers (145/92 not just "high"), age and sex (relevant to risk), current situation (GP's advice), Role (patient health educator, not generic doctor), specific Action with three distinct parts, negative constraint (not general advice). The response will be personalised, evidence-focused, and actionable.

Transformation 3: The creative request

❌ Before

"Write me a speech for my daughter's wedding."

✅ After

"Help me write a father-of-the-bride speech. My daughter Emma is 29 and marrying her partner of 6 years, James. A few things that capture who she is: she's fiercely independent, built her own business from scratch, terrible at accepting help from anyone, and has a brilliant dark sense of humour. James is thoughtful, steady, and the only person she's ever asked for help. The wedding is fairly relaxed and informal — outdoor venue, 60 guests, close family and friends. I want the speech to be warm and funny with one or two genuinely moving moments. About 4 minutes long. Don't make it sound like a generic wedding speech — make it sound like me talking about her."

What changed: Rich character detail (independent, built her own business, dark humour), relationship detail (the only person she asks for help), setting and audience, specific emotional beats wanted, length, strong negative constraint (not generic). The output will be personalised, structured, and genuinely moving.

The single most powerful prompting habit

If you take one thing from Part 3, let it be this: when a response isn't quite right, don't start again — interrogate why.

Ask yourself: What did I not tell it? What format did I not specify? What tone did I assume but not state? What role would have focused it better? What negative constraints would have prevented the generic parts? Usually the answer is one of these five things. Add that information and try again. After a few rounds of this, you'll stop making the same omissions in your first prompts.

📌 Key takeaways from this lesson

Strong prompts contain specific context, a precise role, a clear action, a defined format, and a stated tone. Negative constraints prevent generic output. When a response falls short, diagnose which CRAFT element was missing rather than starting from scratch. The habit of interrogating weak responses is what builds prompting skill fastest.

✏️
Part 3 Exercise — Prompt Transformation

Three tasks that build directly on what you've learned. Work through them in order — each one uses the previous lesson's techniques.

Task 1 — Transform a prompt. Think of a task you've tried AI on before with disappointing results. Rewrite the prompt using CRAFT — add context, a role, specify the format, include at least one negative constraint. Run both versions and compare the outputs side by side.
Task 2 — Have a proper conversation. Start a conversation with AI on a topic that genuinely interests you. Make at least five follow-up messages, using at least three of the five follow-up patterns from Lesson 3.2. Notice how the conversation deepens and the quality of what you get evolves.
Task 3 — Use the critic role. Share an idea, plan, or decision you're considering with AI, using the critic role from Lesson 3.3. Ask it to tell you what could go wrong, what you're assuming, and what you haven't thought of. How does this change how you think about the idea?
🎉
Part 3 Complete!

You now know how to write prompts that actually work — using context, roles, and format instructions to get results that most AI users never see. The CRAFT framework alone is worth the entire course.

Part 4 covers the final piece: using AI safely, critically, and confidently — and earning your certificate.

Start Part 4 →
Lesson marked complete