๐Ÿค– Build Your Own AI Agent โ€” 10 modules, from zero to a 24/7 AI employee working for you๐Ÿค– Build Your Own AI Agent โ€” 10 modules, from zero to a 24/7 AI employee working for you๐Ÿค– Build Your Own AI Agent โ€” 10 modules, from zero to a 24/7 AI employee working for you๐Ÿค– Build Your Own AI Agent โ€” 10 modules, from zero to a 24/7 AI employee working for you๐Ÿค– Build Your Own AI Agent โ€” 10 modules, from zero to a 24/7 AI employee working for you๐Ÿค– Build Your Own AI Agent โ€” 10 modules, from zero to a 24/7 AI employee working for you
Back to course
Module 2 ยท ~8 minutes

Module 2: The R-C-T-F Framework

Prompt Engineering Fundamentals

READ
Riley Goodside โ€” one of the first people hired specifically as a "prompt engineer" by Scale AI โ€” built his entire early career on one insight: structure beats creativity. While everyone was crafting clever one-liners for ChatGPT, Riley was building repeatable frameworks that worked every time.

You don't need to be creative with prompts. You need to be structured. That's where R-C-T-F comes in.

The Framework: Role, Context, Task, Format

R-C-T-F is four components that, when combined, consistently produce high-quality AI output. Think of it as a fill-in-the-blank template for any prompt.

Role โ€” Who should the AI be?
"You are a senior copywriter with 15 years in B2B SaaS." This isn't roleplay for fun. Research from Microsoft shows that assigning a relevant expert role measurably improves output quality โ€” the model draws on different patterns when it "thinks" as a specialist.

Context โ€” What's the situation?
Background, constraints, audience, history. Everything the AI doesn't know but needs to. This is where most people under-invest. You know your situation intimately. The AI knows nothing until you tell it.

Task โ€” What specifically do you want?
One clear action. Not "help me with marketing" but "write 5 subject lines for our Black Friday campaign targeting existing customers." Specificity is everything.

Format โ€” How should the output look?
Bullet points? Table? Email? 200 words or 2,000? Markdown headers? Numbered list? Don't leave this to chance. If you don't specify format, the AI defaults to whatever it feels like โ€” usually a bland essay.

Quick Check

Complete the R-C-T-F framework: Role, Context, ___, Format

R-C-T-F stands for Role, Context, , Format

R-C-T-F in Action

Without framework:

Help me prepare for my job interview

With R-C-T-F:

Role: You are an experienced tech hiring manager at a FTSE 100 company.

>
Context: I'm interviewing for a Senior Product Manager role at Revolut next Tuesday. The job description emphasises "data-driven decision making" and "cross-functional leadership." I have 6 years of PM experience, mostly at early-stage startups. My biggest weakness is that I haven't worked at scale before.

>
Task: Give me the 10 most likely interview questions they'll ask, with brief guidance on what a strong answer looks like for each. Highlight where my startup background is an advantage and where I need to address the scale gap.

>
Format: Numbered list. Each question followed by 2-3 bullet points of guidance. Bold the questions.

The second version gives you an actionable interview prep sheet. The first gives you a generic listicle you could find on Indeed.

Quick Check

Which prompt better uses the R-C-T-F framework for interview preparation?

Which prompt is better?

When to Skip Parts

R-C-T-F isn't dogma. Quick tasks don't need all four:

  • Simple question: Skip Role and Format. "What's the corporate tax rate in the UK for 2024?" works fine.

  • Creative brainstorming: Skip Format. Let the AI surprise you with structure.

  • Expert consultation: Role and Context matter most. Task can be open-ended.


The framework is scaffolding, not a cage. Use what serves the task.
Quick Check

When is it OK to skip parts of the R-C-T-F framework?

The Compound Effect

Here's what most courses won't tell you: R-C-T-F isn't about any single prompt being perfect. It's about your average prompt quality going up dramatically. When you have a framework, your worst prompts are still decent. Without one, your prompts are wildly inconsistent.

Shopify's internal AI training programme reportedly saw a 60% improvement in employee-rated output quality after teaching a structured prompting framework. Not because the framework was magic โ€” because it forced people to think before typing.

Quick Check

The main benefit of R-C-T-F is that it makes your BEST prompts dramatically better.

---
TRY IT

The Full R-C-T-F Template

Role: You are a [SPECIFIC EXPERT] with [RELEVANT EXPERIENCE/SPECIALTY].

Context: [SITUATION DESCRIPTION]. The audience is [WHO]. The goal is [WHAT SUCCESS LOOKS LIKE]. Key constraints: [LIMITATIONS].

Task: [SPECIFIC ACTION โ€” one clear deliverable].

Format: [STRUCTURE โ€” bullet points/table/essay/etc.]. Length: [WORD COUNT]. Style: [TONE].
Example:
Role: You are a senior financial analyst specialising in UK small business accounting.

Context: I run a sole trader consulting business turning over ยฃ85K/year. I'm approaching the VAT threshold and trying to decide whether to register voluntarily now or wait. I'm worried about cash flow impact on my smaller clients.

Task: Give me a clear recommendation on whether to register for VAT now, with the key financial implications of each option.

Format: Start with your recommendation in one sentence, then break down the pros and cons in a table, then give me the specific numbers I should calculate. Keep it under 400 words.

The Role Deep-Dive

Role: You are [EXPERT TYPE]. You've spent [X] years working specifically on [NICHE]. You're known for [CHARACTERISTIC โ€” e.g., "giving direct, no-BS advice" or "finding creative solutions others miss"].

I'm going to describe my situation. Before giving advice, ask me 3 clarifying questions to make sure you fully understand my needs.

Here's my situation: [DESCRIBE]
Example:
Role: You are a startup pitch coach. You've spent 12 years helping founders raise Series A rounds. You're known for being brutally honest about what investors actually care about.

I'm going to describe my situation. Before giving advice, ask me 3 clarifying questions to make sure you fully understand my needs.

Here's my situation: I'm preparing a pitch deck for a B2B SaaS product that automates invoice processing. We have ยฃ200K ARR, 15 customers, and growing 20% MoM. I'm targeting ยฃ2M seed round.

The Format Enforcer

[Your task/question here]

STRICT FORMAT REQUIREMENTS:
- Output as [FORMAT TYPE]
- Use [SPECIFIC STRUCTURE]
- Maximum [X] words per section
- Include [SPECIFIC ELEMENTS]
- Do NOT include [WHAT TO EXCLUDE]
Example:
Analyse the competitive landscape for UK meal kit delivery services.

STRICT FORMAT REQUIREMENTS:
- Output as a comparison table with these columns: Company, Price Range, USP, Biggest Weakness, Market Position
- Include the top 5 players only
- Below the table, write a 100-word "So What" summary identifying the biggest gap in the market
- Do NOT include companies that have shut down
EXERCISE
The R-C-T-F Conversion Challenge:

Take these three vague prompts and rewrite each one using the full R-C-T-F framework:

1. "Help me write a LinkedIn post"
2. "I need advice on my business"
3. "Create a meal plan for me"

For each, invent realistic context (your industry, situation, constraints). Run both the vague and R-C-T-F versions through ChatGPT or Claude. Screenshot the difference โ€” you'll never go back to vague prompts.

---

KEY TAKEAWAYS
  • 1R-C-T-F = Role, Context, Task, Format โ€” a repeatable framework for any prompt
  • 2Role assignment measurably improves output quality โ€” it's not just roleplay
  • 3Context is where most people under-invest โ€” the AI knows nothing until you tell it
  • 4You don't need all four components every time โ€” match the framework to the task complexity
  • 5Structured prompting raises your *floor*, not just your ceiling โ€” your worst outputs get dramatically better
TEST YOUR UNDERSTANDING4 exercises
0 / 4
Exercise 1 of 4

What does R-C-T-F stand for?

Complete all exercises to unlock the Mark Complete button