The single biggest factor in AI output quality is prompt quality. This guide teaches you the RCTFC framework — a practical structure for writing prompts that produce professional results every time.
Most people use AI like a search engine: short, vague queries expecting magic. Then they wonder why they get generic outputs. The difference between a professional who gets consistently impressive results from AI and someone who finds it disappointing almost always comes down to one thing: how they write their prompts.
The RCTFC framework (Role, Context, Task, Format, Constraints) gives you a systematic way to build prompts that work. It applies to Claude, ChatGPT, Gemini, and any other conversational AI.
Five components. Use all five for complex tasks. At minimum, use three (Role, Task, Format).
Tell the AI what role to play. This sets the expertise, tone, and perspective it adopts for the entire conversation.
Weak
"Write a blog post about project management."
Strong
"You are a senior project manager with 15 years of experience in software delivery who writes for an audience of mid-level managers. Write a blog post..."
Why it matters: Roles activate relevant knowledge and shape tone. A finance analyst prompt and a project manager prompt about the same topic will produce meaningfully different outputs.
Give the AI the background it needs. Who is the audience? What has already happened? What constraints exist? The more relevant context, the better the output.
Weak
"Help me write an email to a client about a delay."
Strong
"Our client is a UK law firm expecting delivery of a software integration by Friday. We need to push the deadline by 2 weeks due to an API issue on their end (not ours). The relationship is good but they are under pressure from their own board. Help me write an email..."
Why it matters: Context is everything. Without it, AI produces generic outputs. With it, it produces targeted, situationally appropriate responses.
Be specific about exactly what you want. Use precise verbs: draft, summarise, analyse, compare, rewrite, generate, extract. Avoid vague requests.
Weak
"Help me with my sales presentation."
Strong
"Draft 5 opening slides for a sales presentation. Each slide should have a headline (max 8 words), 3 bullet points (max 10 words each), and a speaker note of 2-3 sentences suggesting what to say."
Why it matters: Precision in the task instruction directly determines how useful the output is. Vague tasks produce vague outputs.
Specify the structure and length of the response. List, table, paragraphs, bullet points, numbered steps, JSON. Don't make the AI guess.
Weak
"Explain the main differences between our two pricing plans."
Strong
"Create a comparison table with these rows: Price, Included features, User limits, Support level, Best for. Two columns: Starter plan and Pro plan. Keep each cell to one concise sentence."
Why it matters: Format specification dramatically reduces editing time. Getting the structure right first time saves multiple rounds of refinement.
Set boundaries: word count, tone, what to include or exclude, what NOT to do. Negative constraints ('do not mention competitors') are as important as positive ones.
Weak
"Write a LinkedIn post about our new product launch."
Strong
"Write a LinkedIn post about our CRM launch. 150-200 words. Professional but conversational tone. Do not use emojis. Do not mention competitors. End with a question to encourage comments. Avoid corporate jargon."
Why it matters: Constraints prevent the most common AI mistakes: too long, wrong tone, generic phrasing, missing your specific requirements.
Once you have the basics, these techniques unlock the next level of AI performance.
Ask the AI to think step by step before answering. Add "think through this step by step before giving your final answer" to complex analytical tasks.
"Analyse our Q3 marketing spend. Think through what worked, what didn't, and why — before giving your recommendations."
After getting an output, ask the AI to argue against it. This surfaces weaknesses and produces more balanced analysis.
"Now argue the opposite. What are the strongest reasons NOT to pursue this strategy?"
Show the AI exactly what you want by pasting examples. 'Write something like this' is more effective than describing the style.
"Here are three LinkedIn posts that got strong engagement for us: [paste examples]. Write a new post in the same style about our upcoming webinar."
Don't try to get the perfect output in one shot. Get a first draft, then refine with specific instructions: shorter, more formal, cut the jargon.
"This is good. Now: make it 30% shorter, replace the word 'leverage', and make the opening sentence more direct."
Ask the AI to evaluate its own output before you do. 'Rate this on a scale of 1-10 and explain what would make it better.'
"Rate this email from 1-10 for persuasiveness and clarity. What would make it a 9 or 10?"
Too short and vague
e.g. "Write me something about AI."
Fix: Add role, context, task, format, and constraints. A good prompt is 50-150 words.
Starting over instead of iterating
e.g. Getting a mediocre draft and starting again with a new prompt
Fix: Keep the conversation going. "This is 80% there — now fix these specific issues: [list]."
Not specifying the audience
e.g. "Write a report on our market position."
Fix: Add audience context: "for our board of directors who have limited technical background".
Accepting the first output
e.g. Using whatever the AI gives you without review or refinement
Fix: Treat the first output as a draft. Always do at least one round of iteration.
Forgetting to specify what NOT to do
e.g. Getting a 2,000 word response when you wanted 200 words
Fix: Add explicit constraints: "maximum 200 words", "do not include introductory phrases", "no bullet points".
Ready to Learn AI?
Reading the framework is the first step. The AI Course London workshop gives you a full morning of hands-on prompt engineering practice — working on your real business problems with immediate feedback.
Book Your Place — £599Sunday 19th April 2026 • Richmond, SW London • 8 people max
Prompt engineering is the skill of writing instructions to AI models that consistently produce useful, high-quality outputs. It's not magic — it's precision. The more clearly you specify what you want (who, what, how, and for whom), the better the output. The RCTFC framework (Role, Context, Task, Format, Constraints) is a practical structure for building good prompts. Most business professionals can dramatically improve their AI outputs within a few hours of deliberate practice.
No. Prompt engineering for business use is entirely natural language — you're just writing clear, structured instructions in English. There's a separate discipline of 'technical prompt engineering' for developers building AI-powered products, which can involve JSON, system prompts, and API parameters. But for a business professional using Claude or ChatGPT day-to-day, the skill is entirely about clear communication and structured thinking.
Long enough to include all necessary context, short enough to stay clear. For most business tasks, a good prompt is 50-150 words. Complex tasks with many constraints or a lot of context may require 300+ words — this is fine. What kills quality is not length but vagueness. A 300-word specific prompt will almost always beat a 30-word vague one.
For recurring tasks (weekly reports, social media posts, client emails), absolutely — build a prompt library. Keep your best-performing prompts in a document and iterate on them. Teams benefit enormously from shared prompt libraries: everyone gets consistent quality and you build on each other's improvements rather than starting from scratch each time.
AI models have some inherent variability — the same prompt doesn't always produce identical output. But more often, the apparent inconsistency is because small differences in context produce large differences in output. Try to make your prompts self-contained: don't assume the AI remembers context from a previous session, and specify explicitly rather than relying on implied meaning. 'Professional tone' means different things in different contexts — 'formal British English, no contractions, third person' is more reliable.
You can learn the core principles in a day and get to a professional level. The RCTFC framework and the five advanced techniques in this guide cover the majority of what you need. Getting genuinely expert takes months of practice — but you'll see 80% of the benefit from the first few hours of deliberate learning. Our workshop includes a dedicated prompt engineering session with hands-on practice, so you leave with the skill, not just the theory.
Tools we use & recommend
Start with the right AI tools
The AI used in our workshop. Best for writing, analysis, and complex reasoning.
OpenAI's flagship model. Strong for research, browsing, and image generation.
AI built into your workspace. Great for notes, docs, and project management.
Some links may be affiliate links. We only recommend tools we actually use.