Prompt Frameworks for Business and Creativity

By this point, you know how to shape a prompt, how to control its structure, and how to build multi step interactions with AI. You understand the “grammar” of talking to intelligent systems.

This section focuses on something more advanced and more practical: frameworks.

Prompt frameworks are reusable patterns that turn prompting from a one off activity into a disciplined practice. Instead of inventing a new instruction every time, you work from proven templates that you can adapt to different tasks, industries, and levels of complexity. This is how professionals achieve consistency, quality, and speed when they work with AI every day.

Here, we will introduce a set of frameworks for core areas of modern work:

Analysis and decision support
Strategy and planning
Communication and stakeholder reporting
Marketing and creative development
Learning, research, and problem solving

Each framework can be customised to your role and environment. You can use them directly with general models such as ChatGPT or Claude, embed them in your own workflows, or treat them as design patterns when you configure systems with multiple AI agents, including those inside Cyrenza.

The goal of this section is simple. When you face a real problem in your work, you should not be staring at a blank prompt box. You should be able to recognise the type of problem, select an appropriate framework, adapt it to your context, and obtain a result that is structured, defensible, and ready to use.

1. What Are Prompt Frameworks?

Prompt frameworks are structured patterns you can reuse whenever you work with AI. Instead of writing a new instruction from scratch every time, you apply a formula that already captures what good prompting requires: clarity, order, and intent.

You have already learned the core building blocks of a strong prompt: Role, Task, Context, Format, and Constraints. You have also seen how more advanced prompting uses few shot examples, step by step reasoning, and chained roles. A prompt framework simply combines these elements into a repeatable recipe for a specific type of outcome, for example an analysis, a strategy document, a client email, or a risk assessment.

In practice, a prompt framework tells the model:

Who to be in this situation (role and perspective)
What to produce (task and output type)
What to use (data, background, constraints)
How to think (step by step reasoning or evaluation criteria)
How to speak (tone, audience, level of detail)

Because the pattern is explicit, you move from a loose request to a designed interaction. Instead of saying, “Help me with this,” you instruct, “Act as X, follow these steps, use this context, and present the result in this format.” That shift is what turns general models into dependable tools.

Over time, these frameworks become part of your personal or organisational toolkit. You refine them, reuse them, and adapt them to new situations. The result is that AI outputs become reliable, consistent, and aligned with your standards, not occasional moments of luck.

2. Framework 1 — The CREST Framework (for General Tasks)

The CREST framework is a practical formula for writing clear, high quality prompts in everyday professional work.

It stands for:

Context → Role → Expectation → Structure → Tone

You can think of it as a compact version of the larger prompt architecture you have already learned. CREST keeps all the essential ingredients but arranges them in an order that is easy to remember and fast to apply.

This framework is useful whenever you need the AI to produce something specific and usable, for example an email, a plan, a summary, a brief, or a short piece of analysis.

1. Context — Where are we and what is this about?

Context tells the AI what situation it is stepping into.

Without context, the model has to guess. With context, it can anchor its response in your reality.

Useful context can include:

What the project or problem is
Who the stakeholders are
Any constraints, deadlines, or goals
Any relevant background information or recent events

Examples of strong context statements:

“Our company is preparing for a yearly review with our largest client in the retail sector.”
“This report will be presented to a municipal government that is considering new transport regulations.”
“The audience is non technical senior leadership in a European financial institution.”

Good context shrinks the distance between your world and the model’s general knowledge. It prevents generic answers and makes the result feel specific and relevant.

2. Role — Who should the AI act as?

The role sets the perspective and level of expertise.

When you say “You are a senior legal counsel” or “You are a supply chain analyst” you are telling the model which part of its training to lean on and what voice to adopt.

Examples:

“You are a senior risk manager at a European bank.”
“You are a change management consultant working with a public hospital.”
“You are a secondary school teacher explaining complex concepts in simple terms.”

The role influences:

Vocabulary and depth of explanation
What the AI considers important or risky
How it frames recommendations

A well chosen role makes the output feel closer to advice from a specialist rather than a general internet answer.

3. Expectation — What does success look like?

Expectation tells the AI exactly what you want the result to achieve.

This is more than “write a plan” or “summarise this.” It defines the target.

You can specify:

Purpose: inform, persuade, analyse, decide
Scope: what must be covered and what can be ignored
Outcome: what the reader should understand, decide, or do afterwards

Examples:

“Create a 3 step social media plan for the first week of launch that increases awareness and encourages existing customers to share the campaign.”
“Summarise the attached report in a way that helps a busy executive decide whether to approve the project.”
“Evaluate the strengths and weaknesses of this proposal, focusing on financial risk and operational feasibility.”

The clearer the expectation, the less revision you will need later.

4. Structure — How should the answer be organised?

Structure tells the AI how to present its thinking.

This is where you ask for bullet points, sections, tables, headings, or numbered steps.

Why it matters:

It saves you formatting time.
It makes the output easier to read and share.
It forces the AI to think in a logical order.

Examples of structure instructions:

“Use numbered points with a short explanation under each.”
“Present the answer under three headings: Context, Risks, Recommendations.”
“Create a two column table, with ‘Issue’ on the left and ‘Suggested Action’ on the right.”
“Write in short paragraphs of no more than four sentences each.”

By defining the structure, you turn a loose block of text into something ready to use.

5. Tone — How should it sound?

Tone controls the voice and style.

The same content can sound very different when written for a public audience, a technical team, or a board of directors.

You can specify tone in terms such as:

Formal, neutral, conversational, encouraging, direct
Suitable for senior leadership, suitable for students, suitable for public communication
Cautious, optimistic, urgent, reassuring

Examples:

“The tone should be inspiring, professional, and easy to follow.”
“Use formal language suitable for a regulatory briefing.”
“Write as if explaining to a non technical colleague, clear and calm.”

Tone alignment is critical in organisations where wording can affect trust, reputation, or legal interpretation.

Putting CREST Together

Here is the original example, enriched with CREST fully visible:

Context: “The company is launching a new sustainability campaign focused on reducing plastic waste across all European offices.”
Role: “You are a senior marketing strategist with experience in sustainability communications.”
Expectation: “Create a 3 step social media plan for the first week of launch that raises awareness and encourages employees and partners to participate.”
Structure: “Use numbered points, each with a title and a short explanation under it, and finish with one suggested call to action for the audience.”
Tone: “Inspiring, professional, and easy to follow for a general European business audience.”

You can give this to any capable language model and you will usually receive an output that:

Stays on topic
Uses appropriate language
Is ready to share with minimal editing

A Second Example: Internal Operations Memo

CREST works across domains, not only marketing. For example, an operations manager preparing an internal memo could use:

Context: “We are experiencing delays in our internal ticketing system, which affects how quickly the IT team can respond to staff requests.”
Role: “You are an internal communications specialist.”
Expectation: “Draft an internal memo that explains the issue, sets expectations for response times over the next two weeks, and reassures staff that a permanent solution is underway.”
Structure: “Use three short sections titled ‘What is happening’, ‘What this means for you’, and ‘What we are doing about it’.”
Tone: “Clear, calm, and transparent, suitable for all staff in a European public sector organisation.”

This is the same framework applied to a different kind of task. The pattern remains stable even when the content changes.

How to Use CREST in Practice

Write your prompt as five short lines, one for each element.
Check for gaps. Ask yourself, “Would a new colleague understand this situation from what I wrote?” If not, add context or clarify expectations.
Paste into your AI tool as one coherent instruction.
Refine iteratively. If the first result is close but not perfect, adjust one CREST element. For example, tighten the tone or change the structure to a table.

Over time, CREST becomes automatic. You will find yourself thinking, before every important prompt, “Have I given enough context, a clear role, a specific expectation, a defined structure, and a suitable tone.”

That habit is what separates casual users from professionals who can rely on AI as a stable part of their work.

3. Framework 2 — The IDEA Framework (for Creativity & Innovation)

The IDEA framework is designed for moments when you want the AI to help you think, not just write.

Use it whenever you need:

Fresh campaign concepts
New product or service ideas
Alternative ways of solving a problem
Creative angles for communication or change initiatives

IDEA stands for:

Inspire → Define → Expand → Apply

Unlike purely open prompts such as “Give me some ideas”, this framework guides the model through a creative journey that starts with a spark and ends with a practical outcome.

1. Inspire – Give the model a starting spark

Creativity improves when there is a clear starting point.

In the Inspire step, you give the AI a theme, domain, or raw material to work with. This limits the space in a helpful way and prevents vague or generic suggestions.

You can inspire with:

A topic
- “Remote work culture”
- “Sustainable packaging”
- “Digital literacy for older adults”
A format
- “Internal campaign”
- “Workshop series”
- “Social media content”
A specific audience
- “First year university students”
- “Employees in logistics and warehousing”
- “Small business owners in Europe”

The goal is to give the AI something to stand on before it starts creating.

Example inspire line

“The topic is remote work culture for mid sized European companies.”

2. Define – Clarify the creative goal or problem

Creativity is only useful when it serves a purpose.

In the Define step, you explain what you are trying to achieve. This turns vague “ideas” into ideas that solve a specific problem or move a specific metric.

You can define by:

Stating the problem
- “Employees feel disconnected from each other.”
- “Customer engagement has declined on our main channel.”
Stating the objective
- “We need to increase participation in our training programs.”
- “We want to position our brand as a leader in sustainability.”
Giving constraints
- “The campaign must be digital only and low budget.”
- “We have to follow strict compliance rules and avoid strong claims.”

Example define line

“We need a campaign that helps companies engage remote employees and strengthen team culture online.”

This ensures the AI does not just produce interesting concepts, but relevant ones.

3. Expand – Ask for multiple diverse options

This is where you let the AI generate possibilities.

In the Expand step, you ask for several ideas at once, and you can guide for variety, originality, or different angles.

You can instruct the AI to:

Generate a specific number of ideas
- “Generate 6 campaign concepts.”
- “Give me 10 product ideas.”
Vary the type
- “Include a mix of small quick wins and bigger strategic ideas.”
- “Provide some serious options and some more playful options.”
Consider different perspectives
- “Include ideas for leaders, for employees, and for HR.”

This step is about breadth. Quantity comes first, quality can be filtered afterwards.

Example expand line

“Generate 6 distinct campaign concepts that approach engagement from different angles.”

4. Apply – Anchor ideas in real execution

Many idea lists fail because they stay theoretical.

The Apply step forces the AI to connect each idea to a concrete application so that a human can judge feasibility and value.

You can ask the model to:

Explain execution channels
- “For each idea, explain how it could be executed on LinkedIn and in email.”
- “Describe how this could be run as a monthly internal initiative.”
Link to resources or roles
- “Indicate which team would own this and what they would need.”
Tie to outcomes
- “For each idea, add one sentence on what success would look like.”

This turns abstract creativity into actionable creativity.

Example apply line

“For each concept, explain in 3 to 4 sentences how it could be executed on LinkedIn, including the type of posts, frequency, and ideal audience.”

Full Example: Remote Work Culture Campaign

Putting it together:

“Let us use the IDEA framework.

Inspire: The topic is remote work culture in mid sized European companies.

Define: We need a digital campaign that helps companies engage employees online and strengthen a sense of team belonging. The budget is modest and the campaign should be realistic for HR teams to run.

Expand: Generate 6 distinct campaign concepts, including a mix of light touch initiatives and deeper culture programs.

Apply: For each concept, explain how it could be executed on LinkedIn in 3 to 4 sentences. Include the type of content, suggested posting rhythm, and the main outcome it supports.”

This single prompt now gives the model:

A clear theme
A defined goal
A request for multiple diverse ideas
An instruction to show practical usage

The output tends to be significantly more useful than a simple “Give me ideas for a remote work campaign.”

Second Example: Product Innovation in Education

You can use IDEA in completely different domains.

“Use the IDEA framework to help with an innovation task.

Inspire: The topic is digital tools for secondary school science education.

Define: We want ideas for new features that could help teachers explain complex concepts more interactively to students aged 13 to 16. The tools should be usable in schools with limited hardware and variable internet quality.

Expand: Generate 8 feature ideas, some small improvements and some larger concepts.

Apply: For each feature, describe how it would work in the classroom in 3 to 5 sentences and mention one possible risk or limitation that schools should consider.”

Here, the framework guides the AI to:

Respect the realities of schools
Consider constraints and risks
Generate ideas that are not only creative but grounded

How IDEA Relates to CREST

IDEA can be combined with CREST rather than used instead.

CREST focuses on how to package the instruction
IDEA focuses on how to structure creative thinking

A combined prompt might use CREST to define the overall prompt and IDEA inside the Expectation part.

For example:

Context, Role, Tone come from CREST.
The creative process inside the Task uses Inspire, Define, Expand, Apply.

This layered approach gives you both clarity and creativity under control.

Using IDEA Iteratively

You do not need to get everything perfect in one prompt. You can use IDEA in cycles:

Run IDEA to get first round concepts.
Pick two or three promising ideas.
Run a second IDEA prompt focused only on those, for example:

“Inspire: Focus on idea 2 and idea 5 from the previous list.

Define: We want to adapt these for a healthcare context.

Expand: Suggest 4 variations for each idea.

Apply: Explain which stakeholders would be involved in implementation.”

Over time, this creates a structured dialogue where AI helps you explore, refine, and ground creative solutions, instead of just producing one long list.

In summary, the IDEA framework ensures that when you ask for creativity, you receive ideas that are not only interesting but also connected to real problems, real constraints, and real next steps.

4. Framework 3 — The SCQA Framework (for Analysis & Consulting)

The SCQA framework comes from classic management consulting practice. Firms such as McKinsey use it to turn complex situations into clear narratives that decision makers can act on.

SCQA stands for:

Situation → Complication → Question → Answer

It is a way of structuring thinking: first describe what is true, then explain what changed, then ask the right question, and finally propose a response. This makes it an ideal backbone for prompts that involve analysis, strategy, diagnosis, or structured recommendations.

How SCQA Works

1. Situation: Describe the current state

The Situation sets the stage. You explain the context that everyone would agree on if they looked at the facts.

This might include:

Who the organisation or client is
What market they operate in
What products or services they offer
What recent performance has looked like

The key is neutrality. You are not yet describing a problem. You are just answering the question:

“Where are we now”

When you give this to an AI model, you are giving it the foundation on which all later reasoning will sit. Without a clear situation, the AI has to invent context and the output becomes less reliable.

Example situation for a prompt

“Our logistics company operates in three European countries and delivers consumer parcels for major retailers. Over the past two years we have grown quickly and now handle an average of 50 000 deliveries per day.”

2. Complication: Identify the issue or change

The Complication introduces the tension. Something has changed or emerged that makes the situation unsustainable or less desirable.

This might be:

A new competitor entering the market
Rising costs or falling margins
Regulatory changes
Operational bottlenecks
Shifts in customer behaviour

The complication answers the question:

“What has gone wrong or what has become more difficult”

For the AI, this is the trigger that tells it where to focus its attention. It narrows the analysis to the specific pressure point.

Example complication for a prompt

“In the last six months our average delivery time has increased by 20 percent, customer complaints about delays have doubled, and fuel prices have risen sharply. At the same time, city centres have introduced tighter traffic restrictions that slow our vans during peak hours.”

3. Question: Define what needs to be solved

The Question is the hinge of the framework. It states clearly what you want to resolve.

Good SCQA questions are:

Specific instead of vague
Framed around outcomes rather than tools
Focused on decisions or options

They answer:

“What decision or problem are we trying to address”

For AI prompting, this is where you state the task in analytical form. It guides the model away from generic advice and towards a targeted problem solving effort.

Example question for a prompt

“How can we reduce delivery delays over the next twelve months while keeping overall costs under control”

You can further refine this by adding constraints, such as avoiding large capital expenditure or protecting service quality.

4. Answer: Ask the AI to propose structured solutions

The Answer in the SCQA story is not something you write in advance in your prompt. Instead, you instruct the AI on how you want it to construct the answer.

This might include:

Number of options or recommendations
Type of structure, for example, step by step plan, scenario comparison, or list of trade offs
Level of detail
Time horizon

By defining the expected shape of the answer, you help the AI produce something you can immediately use, adapt, or present.

Example answer instruction for a prompt

“Provide a four step operational improvement plan.

For each step, briefly describe the action, the expected impact on delivery time and cost, and one possible risk or downside.”

This is the moment where SCQA and prompt engineering meet: the Situation, Complication, and Question sit inside the context and task, while the Answer instruction is part of the format and constraints.

Full SCQA Prompt Example

Here is the full example brought together as a single prompt:

“You are a senior strategy consultant advising a mid-sized logistics company.

Situation: Our company operates in three European countries and delivers consumer parcels for major online retailers. Over the past two years we have grown quickly and now handle around 50 000 deliveries per day.

Complication: In the last six months our average delivery time has increased by 20 percent, customer complaints about delays have doubled, fuel prices have risen, and several city centres have introduced tighter traffic restrictions that slow our vans during peak hours. Management is concerned that we will lose contracts if service levels continue to decline.

Question: How can we reduce delivery delays over the next twelve months without significantly increasing total operating costs

Answer: Provide a four step operational improvement plan. For each step, explain the action, the expected impact on delivery time and cost, one key risk, and how that risk can be mitigated. Present the answer as a numbered list with clear headings.”

This gives the AI:

Clear context
A precise problem
A well defined question
A structured way to respond

The output will usually be far closer to consultant level reasoning than if you had simply written:

“How can we improve our logistics operations”

Using SCQA Across Different Domains

Although SCQA comes from consulting, it is useful in many roles:

Management
- Use SCQA to brief AI on internal challenges and request decision options.
Project management
- Use SCQA to diagnose why a project is delayed and request recovery strategies.
Marketing and communications
- Use SCQA to structure situation analyses and narrative reports.
Public sector and policy
- Use SCQA to frame societal problems and explore policy responses.

For example, a public sector prompt might look like this:

“You are a policy analyst.

Situation: Our city has seen a steady increase in cycling over the past five years, with 30 percent of commuters now using bicycles.

Complication: Accident rates involving cyclists and cars at major intersections have increased by 25 percent in the same period, leading to public concern and media attention.

Question: What policy and infrastructure measures could we introduce over the next three years to improve cyclist safety without discouraging car traffic entirely

Answer: Propose three combined packages of measures. For each package, describe the actions, estimated cost level, expected impact on safety, and one possible public resistance we should plan to address.”

How SCQA Helps When Working With AI

Using SCQA in your prompts has several benefits:

It forces you to clarify your own thinking before asking for help.
It reduces the chance that the model will misinterpret the problem.
It encourages logically structured answers that are easy to present or adapt.
It can be reused as a template across topics and projects.

You can also use SCQA iteratively. For example:

First prompt: Ask the AI to help you define the Situation and Complication from raw notes.
Second prompt: Once the SC part is clear, refine the Question.
Third prompt: Run a focused SCQA prompt with a sharpened Question and a more precise Answer instruction.

In practice, SCQA turns AI from a general assistant into something closer to a structured analyst: it guides the model to frame the problem, not just throw solutions at it.

By combining SCQA with the earlier CREST framework, you can design prompts that are both narratively sound and technically clear. CREST shapes the overall prompt, and SCQA shapes the analytical core.

5. Framework 4 — The PESTEL Framework (for Business Analysis)

The PASTEL framework is a structured way to examine the external environment around a business or sector. It is a variation of the classic PESTLE model used in strategy and policy work. PESTEL helps you and the AI think systematically about forces that lie outside the organisation, but still have a strong influence on performance and risk.

PESTEL stands for:

Political → Economic → Social → Technological → Environmental → Legal

Used as a prompt framework, PASTEL turns vague questions such as

“Analyze the market for renewable energy in Africa”

into a clear, stepwise analysis that covers all major external factors.

How PESTEL Works

When you use PESTEL with an AI model, you are asking it to walk through six lenses in a disciplined way. Each lens guides the model to pick up different types of information and trends.

You can think of it as asking:

“What is shaping this industry politically, economically, socially, technologically, environmentally, and legally”

Below is how to describe each element when you build your prompt.

P - Political

This covers how governments and public policy affect the industry.

Examples of what AI should look at:

Government stability and policy direction
Subsidies, incentives, or taxes that affect the sector
Trade agreements or sanctions
Public investment priorities
Geopolitical tensions that may disrupt supply or demand

In a prompt, you might say:

“Under Political, focus on government support, stability, and major policies that influence this sector.”

E - Economic

This focuses on the economic conditions that shape opportunity and risk.

Examples:

GDP growth and income levels
Inflation and interest rates
Access to capital and investment trends
Currency stability
Cost structures and margins in the sector

In a prompt, you might refine it as:

“Under Economic, consider growth rates, capital availability, and cost drivers that matter for investors and operators.”

S - Social

This looks at people and society.

Examples:

Demographic trends such as age, urbanisation, or migration
Consumer attitudes, values, and expectations
Education levels and skills
Public awareness of issues related to the sector
Trust in institutions and brands

You can guide the AI with instructions like:

“Under Social, highlight changes in behaviour, preferences, and demographics that could raise or reduce demand.”

T - Technological

This covers the state and speed of technology in the area you are analysing.

Examples:

New technologies reshaping the industry
Adoption rates of digital tools and platforms
Research and development intensity
Infrastructure, for example, connectivity, data, or energy systems
Risk of technological disruption from adjacent sectors

In a prompt, you might say:

“Under Technological, focus on innovations, adoption barriers, and technologies that may accelerate or disrupt this sector.”

E - Environmental

This focuses on the physical environment and sustainability.

Examples:

Climate conditions and climate risk
Resource availability and scarcity
Environmental regulations and standards
Public pressure related to sustainability and emissions
Physical vulnerabilities such as floods, droughts, or extreme weather

For the AI, you can specify:

“Under Environmental, examine climate risks, resource constraints, and sustainability pressures relevant to the sector.”

L - Legal

This lens looks at laws, regulations, and compliance requirements.

Examples:

Industry specific regulations and licensing
Labour laws
Data protection and privacy rules
Health and safety standards
International regulations that affect cross border activity

In your prompt, you might say:

“Under Legal, review the main regulations, compliance obligations, and legal risks that affect how companies operate.”

Using PESTEL in a Prompt

A robust prompt could look like this:

“You are a senior business analyst preparing a briefing for European policymakers.

Perform a PESTEL analysis for the African renewable energy industry.

For each factor - Political, Economic, Social, Technological, Environmental, Legal - identify:

One key trend that is shaping the sector
One major opportunity for investors or operators
One significant challenge or risk

Present the answer under six headings, one for each PESTEL factor. Under each heading, use three bullet points labelled Trend, Opportunity, and Challenge. Keep explanations clear, factual, and under 60 words per bullet point.”

This prompt tells the AI:

The role it should act in
The framework it must follow
The output structure
The level of detail and style

The result is a well organised external analysis that can be used directly in presentations, reports, or strategic discussions.

Why PESTEL Works Well With AI

Using PESTEL as a prompt framework gives you several advantages:

It prevents the AI from focusing only on one dimension, for example, technology, and ignoring politics or law.
It forces a balanced view, which is important in policy, strategy, and investment work.
It makes outputs easier to compare over time or across regions, because the structure remains constant.
It reduces the risk of vague or repetitive answers, since each letter has a clear purpose.

You can also adapt the framework:

Narrow it to one or two letters for a quick scan, for example, just Technological and Legal.
Add extra constraints, such as “focus on the next five years” or “emphasise issues relevant to small and medium enterprises.”
Combine it with SCQA by framing a Situation and Complication first, then instructing the AI to use PESTEL to deepen the analysis.

Example Variations

Here are two additional prompt patterns that build on PESTEL.

For a single company:

“You are an external strategy advisor. Use the PESTEL framework to analyse the external environment for a mid sized European solar panel manufacturer that wants to expand into West Africa. For each PESTEL factor, briefly describe one key external force and explain in two sentences how it affects the expansion decision.”

For policy and regulation:

“You are advising an African regional development bank. Using PESTEL, identify the main external drivers that will influence public investment in renewable energy between now and 2035. Highlight where supportive policy could unlock private investment. Provide your answer as a numbered list under each PASTEL heading.”

By using PESTEL in this way, you transform the AI into a structured analyst rather than a generic commentator. The framework gives both you and the model a disciplined way to think about markets, industries, and strategic environments.

6. Framework 5 — The ARC Framework (for Persuasive Communication)

The ARC framework is designed for any situation where you want the AI to help you persuade, motivate, or move someone to take action. It works especially well for marketing messages, internal announcements, stakeholder updates, fundraising decks, and public communication.

ARC stands for:

Attention → Reason → Call to Action

You can think of it as a simple journey:

Get people to look up.
Give them a clear reason to care.
Tell them exactly what to do next.

Used as a prompt framework, ARC helps the AI create communication that is focused, intentional, and action oriented, instead of vague or purely descriptive.

How ARC Works

1. Attention

The first step is to capture interest.

In this phase, the AI should create an opening that makes the reader stop and think. This can be done through:

A surprising statistic
A provocative question
A short, vivid statement about a problem
A relatable scenario

Examples:

“Every week, the average employee loses 8 hours to manual admin.”
“What would your team achieve if email was no longer a full time job?”
“Most small businesses are working harder than ever, yet growing slower than ever.”

In your prompt, you can guide this explicitly:

“For the Attention part, start with one short sentence that highlights a surprising or painful fact for small business owners.”

This tells the AI that the opening must earn attention, not just introduce the topic politely.

2. Reason

Once you have attention, you must give the audience a reason to care.

This is where the AI explains:

Why the situation matters
What is at stake if nothing changes
How your proposed idea, product, or action helps

The Reason portion can be logical, emotional, or both. It should connect the opening to a meaningful benefit or insight.

Examples:

“Manual reporting does not just waste time. It also delays decisions and hides valuable patterns in your data.”
“When teams are buried in repetitive tasks, creativity and innovation are the first things to disappear.”
“Smart automation tools can handle routine work quietly in the background, so your team can focus on customers, strategy, and growth.”

In your prompt, you might write:

“For the Reason part, explain in 2 to 3 sentences how AI automation reduces wasted time and improves focus on high value work. Use clear, concrete language.”

This gives the model a clear target: explain impact, not features.

3. Call to Action

The final step is the Call to Action. This is where you convert interest and understanding into a specific next step.

A good Call to Action is:

Clear
Simple
Easy to follow
Matched to the context (for example, click, reply, book, read, sign up)

Examples:

“Download the free guide to see which tasks your team can automate in one week.”
“Reply to this email with the word ‘demo’ and we will schedule a 20 minute session.”
“Visit our resource page to see real examples from companies your size.”

When prompting the AI, you can specify:

“For the Call to Action, end with one sentence inviting readers to take a realistic, low friction next step, such as learning more or booking a short call.”

This keeps the message from ending weakly or vaguely.

Example Prompt Using ARC

Here is how you might instruct an AI model using the ARC framework:

“Write a short LinkedIn post promoting AI automation for small and medium sized enterprises using the ARC framework.

Attention: Open with a surprising statistic about how many hours employees lose to manual work each week.
Reason: In 3 to 4 sentences, explain how AI based automation can free teams to focus on growth, customers, and strategic work, instead of repetitive tasks. Keep the language concrete and non technical.
Call to Action: Close with one clear sentence inviting readers to learn more about our automation solution on our website.

Keep the post under 180 words, use a professional but approachable tone, and write for a European business audience.”

The model now knows:

The structure (Attention, Reason, Call to Action)
The content focus (AI automation for SMEs)
The tone and audience
The length constraint

The result is usually far more persuasive and usable than a generic instruction such as:

“Write a LinkedIn post about AI automation.”

Adapting ARC to Different Channels

ARC is flexible. You can apply it to:

Emails

Attention: A subject line and opening sentence that highlight a problem or opportunity.
Reason: A short explanation of how you can help and why it matters now.
Call to Action: A direct request to reply, schedule, or review something.

Presentations

Attention: A striking first slide with a key fact or question.
Reason: A few slides explaining the situation and your proposed solution.
Call to Action: A closing slide with a clear decision or next step.

Internal Communication

Attention: “Here is the change that will affect your team.”
Reason: “This is why we are doing it, and how it benefits you and the organisation.”
Call to Action: “Here is what we need you to do by Friday.”

When prompting the AI, you can always embed ARC explicitly:

“Use the ARC structure. Label each part with a heading: Attention, Reason, Call to Action.”

This produces output you can quickly adapt, reorganise, or paste into your channel of choice.

Why ARC Works Well With AI

AI models are strong at language generation, but they do not automatically organise messages into persuasive flow. If you simply ask for “a post” or “a paragraph”, you often get something that sounds nice but lacks direction.

ARC fixes that by:

Giving the AI a narrative spine to follow
Ensuring there is always a clear point and a clear next step
Helping you check the quality of the output quickly:
- Does it grab attention
- Does it give a solid reason
- Does it ask for something concrete

You can also combine ARC with other frameworks:

Use CREST to set context and tone.
Use ARC inside the content section for external communication.
Use SCQA for analytical reports and ARC for the executive summary or cover letter.

Over time, you can build your own variations, for example:

Attention → Problem → Solution → Call to Action
Attention → Story → Insight → Call to Action

The core idea remains the same: guide the AI through a deliberate persuasive journey, instead of letting it wander.

Used in this way, the ARC framework turns AI from a general writing assistant into a focused communication partner that helps you inform, influence, and mobilise your audience with clarity and confidence.

7. Framework 6 — The DEEP Framework (for Refinement & Quality Control)

Even a strong first draft from an AI model often needs work. The DEEP framework gives you a structured way to take any rough output and turn it into something you would confidently send to a client, colleague, or executive.

DEEP stands for:

Diagnose → Evaluate → Enhance → Present

You can use this framework yourself when editing, or you can explicitly ask the AI to apply it to its own output.

How DEEP Works

1. Diagnose: Identify what is wrong or missing

The first step is diagnosis. Before you fix anything, you need to understand the problems.

Diagnosis asks questions such as:

Is anything unclear
Is anything factually weak or unfounded
Is the structure confusing or repetitive
Is something important missing for the intended audience

When you involve an AI model, you can ask:

“Diagnose the following text. Identify unclear explanations, missing details, and any sections that may be too generic for a senior audience. List your observations clearly.”

This turns the AI into a critic instead of a writer. It forces the model to read its own work with a more analytical lens.

You can also add your own diagnosis. For example, you might note that the tone is too casual, the examples are not relevant for Europe, or the explanation is too technical for non specialists. These observations can be included in the next step.

2. Evaluate: Decide what needs to change and why

Once you have a list of issues, you move to evaluation. Here you judge how serious each issue is and what kind of change is required.

Evaluation focuses on:

Priority: Which problems matter most for this audience
Purpose: Does the text support the goal of the document
Quality: Is the argument strong, balanced, and coherent

You can ask the AI:

“Evaluate the previous diagnosis. Which three issues are most important to fix for a European policy audience, and why. Explain your reasoning briefly.”

This step pushes the model to think about impact, not only about surface errors. For example, it may decide that a missing definition is more serious than a slightly long paragraph, because the text is meant for non technical readers.

Evaluation is where you align the draft with the real world purpose of the document.

3. Enhance: Rewrite, restructure, and strengthen

After you know what must change, you move to enhancement. This is the active editing phase.

Enhancement can involve:

Rewriting unclear sentences
Reordering sections for logical flow
Adding definitions, examples, or transitions
Adjusting tone to match the audience
Tightening long passages without losing substance

Here is a typical instruction:

“Using your evaluation, enhance the text.

Clarify any vague statements.
Add one concrete example where you mentioned impact.
Improve transitions between paragraphs.
Keep the length within ten percent of the original.

Present the revised version in full.”

Because you have already done diagnosis and evaluation, enhancement is not random. The AI is no longer guessing what to improve. It is working from a clear improvement plan.

You can repeat this step if needed, for example:

“Enhance the revised version again, focusing only on improving clarity for non specialist readers. Avoid technical jargon and explain key terms in simple language.”

4. Present: Deliver a clean final version

The last step is presentation. This is where you prepare the text for actual use.

Presentation involves:

Cleaning up formatting
Ensuring headings and sections are consistent
Checking that tone matches the context (for example, formal for a regulator, warm but professional for internal teams)
Making sure there are no leftover bracketed notes or prompt language in the final text

You can instruct:

“Present the final version as a clean document with clear headings, short paragraphs, and bullet points where appropriate. Remove any internal notes or analysis. The text should be ready to send to a senior stakeholder.”

Presentation is what turns an improved draft into something that looks complete, deliberate, and publication ready.

Example: Using DEEP on a Weak Summary

Imagine you have this AI generated summary of a report on AI in education:

“AI is changing education. It can help students learn and teachers save time. Schools should use AI tools to improve things. There are also risks, like bias and privacy, so people should be careful.”

You can apply DEEP as follows.

Prompt

“Apply the DEEP framework to the summary below.

Diagnose: List weaknesses in clarity, depth, and specificity.
Evaluate: Identify the three most important issues to address for an audience of European education policymakers.
Enhance: Rewrite the summary to correct these issues, adding one concrete example and one specific risk. Keep it under 180 words.
Present: Provide the final summary in clean paragraph form, suitable for inclusion in a briefing note.”

The output will usually:

Replace vague statements with specific findings
Add a concrete example, perhaps adaptive learning software or early warning systems for dropouts
Name a specific risk, such as biased scoring or opaque algorithms, rather than saying “there are risks”
Use a tone that fits policy level communication, not casual commentary

The same approach works for emails, reports, slide notes, executive summaries, or even marketing copy.

Using DEEP As an Ongoing Habit

You can use DEEP in two ways:

As a direct instruction to the AI
- “Use DEEP to review and improve this email to a client.”
- “Use DEEP to refine this two page report for clarity and tone.”
As a mental checklist for yourself
- Diagnose: What feels off
- Evaluate: What truly matters for this audience
- Enhance: How do I fix those points
- Present: Does this look and read like a finished piece

Over time, DEEP becomes a natural part of your workflow. You will find yourself reading every AI output with these four questions in mind:

What is wrong or missing
How serious is it
How do we fix it
Is this ready to send

Why DEEP Matters in Professional Contexts

In business, education, law, healthcare, or government, the cost of a sloppy or unclear message is high. AI can accelerate your work, but it can also multiply small mistakes if they go unchecked.

The DEEP framework:

Protects quality while you scale your use of AI
Encourages critical thinking rather than blind trust
Gives you a repeatable method for turning rough drafts into refined outputs
Aligns well with other frameworks in this module, such as CREST and SCQA

When combined with earlier ideas such as context windows and layered prompting, DEEP helps you maintain editorial control. The AI generates, you direct and refine.

That is the central theme of this curriculum: AI provides speed and reach, but humans provide standards.

8. How to Use Frameworks Together

In real work, strong outputs rarely come from a single prompt. They come from a sequence of well designed prompts, each doing a specific job. That is where frameworks become most powerful.

Prompt frameworks are modular tools. You can think of them as building blocks that can be combined in different orders depending on your goal. The aim is not to memorise every framework, but to learn how to chain them together so that each one improves the result at a different stage.

A typical professional workflow might look like this:

SCQA to think and structure the problem
You start by using the SCQA framework (Situation, Complication, Question, Answer) to clarify the problem and structure your thinking.

Example prompt:
“You are a strategy consultant. Using the SCQA framework, help me describe the current situation for our European retail business, the main complication we face, the key strategic question, and a high level answer in bullet points.”

This step gives you a clear narrative: what is happening, what is going wrong, what must be decided, and where the answer should point. It is a thinking step, not a writing step.
CREST to frame a full task for the AI
Once the problem is structured, you can move to CREST (Context, Role, Expectation, Structure, Tone) to design a complete prompt for the main piece of work.

Example:
“Context: Use the SCQA summary you just created for our retail business.

Role: You are a senior strategy advisor.

Expectation: Draft a two page internal briefing that explains the situation and outlines three strategic options.

Structure: Use headings for Situation, Complication, Strategic Options, and Recommendation.

Tone: Clear, concise, and suitable for a board level audience.”

Here SCQA provides the content foundation, and CREST provides the instruction scaffold.
ARC for a concise executive summary or external message
Next, you might use ARC (Attention, Reason, Call to Action) to turn the same core insight into a short, persuasive piece such as an email, slide headline, or LinkedIn post.

Example:
“Using the briefing you produced, apply the ARC framework to write a 120 word executive summary.
- Attention: Open with one striking fact about recent performance.
- Reason: Explain in one or two sentences why a strategic response is needed now.
- Call to Action: Close with a clear next step for the leadership team.”
This step translates analysis into a message that moves people to act.
DEEP for refinement and quality control

Finally, you apply the DEEP framework (Diagnose, Evaluate, Enhance, Present) to polish the outputs before using them.

Example:
“Apply the DEEP framework to the executive summary. Diagnose issues with clarity and tone, evaluate which are most important for a senior European audience, enhance the text to correct them, and present the final version as a clean paragraph suitable for a board pack.”

DEEP acts as a quality filter and ensures the final result is sharp, readable, and aligned with your audience.

Thinking in Layers, Not Single Prompts

The key idea is that different frameworks serve different stages of the work:

SCQA helps you understand and structure the problem.
CREST helps you brief the AI for a complete task.
IDEA or PESTEL might help you explore options or analyse environments.
ARC helps you communicate persuasively.
DEEP helps you refine and finalise.

You can mix and match them:

For a policy paper: PESTEL for external context, SCQA for framing the problem, CREST for drafting, DEEP for polishing.
For a marketing campaign: IDEA for concept generation, CREST for campaign brief, ARC for ad copy, DEEP for final review.
For internal change communication: SCQA to frame the situation, ARC for the main message, DEEP to ensure clarity and sensitivity.

Frameworks do not confine you, they give you structure and repeatability. As you gain experience, you will begin to improvise: combining elements from several frameworks in a single prompt, or running them in sequence over the same piece of work.

Over time, this layered approach becomes second nature. You stop treating AI as a black box and start treating it as a system you can design: first you shape the thinking, then the message, then the quality. That is how frameworks scale with you from simple tasks to complex, high stakes work.

Next: The Science of Iteration and Refinement

Even the strongest prompt will not always produce the ideal result on the first attempt.

This is not a failure of the model or of the user. It is simply how interactive systems work.

Professionals do not expect “one shot perfection.”

They read the output, compare it to their intent, and then adjust the instruction. They treat every exchange as a small experiment:

give a clear instruction,
observe the response,
correct or tighten the prompt,
try again with what they have learned.

Over time this turns a single good answer into a repeatable pattern that can be trusted and reused.

In the next section, we will look at this process in a structured way. You will learn how to refine prompts step by step, how to interpret responses with a critical eye, and how to give feedback that steadily improves the model’s behaviour for your use cases. The goal is not just a better answer, but a better system that produces quality reliably.