Frameworks & Strategy

Prompting frameworks, AI adoption models, and implementation guides for your organization.

Top 5 Prompting Frameworks

Structured approaches to get better, more consistent outputs from any AI model.

R = Role · T = Task · F = Format

This simple framework works well for general prompting in any scenario.

Template: "Act like a [role]. Can you [insert task] in [format] format?"

Example: Act like a chef who has 30 years of experience in cooking. Can you give me a weekly plan with the purpose of losing weight in TABLE format? Give me a shopping list at the end in text format.

This method improves an LLM's reasoning by going step-by-step.

Good for:

  • Analytical tasks
  • Decision making
  • Problem solving

Simply add "Let's think step by step" at the end of your prompt.

R = Role · I = Instructions · S = Steps · E = End Goal · N = Narrowing (Constraints)

Good for:

  • Writing tasks with specific constraints (e.g., blog posts)
  • Tasks with clear guidelines (e.g., business plans)

Example: Role: You are a business strategist with expertise in integrating Automation. Main Task: Write an in-depth X post about automation in modern business. Steps: 1. Start with a simple introduction... 2. Proceed with two main sections... 3. End with a conclusion. Goal: Inform people on why automation is important. Constraints: No hashtags, no emojis, max 500 words.

R = Role · O = Objective · D = Details · E = Examples · S = Sense Check

Good for:

  • When you have examples similar to your desired output

Example: R: You are a seasoned copywriter specializing in crafting viral tweets. O: Write a tweet that aims to go viral on the topic of discipline. D: Very concise, use powerful language, make it relatable. E: "Discipline isn't just a one-time act; it's a lifetime commitment to excellence." S: Do you understand the objective and guidelines?

A prompt that uses recursion to create increasingly better outputs.

Good for:

  • Summaries
  • Improving your favorite prompts
  • Generating usable long-form content via recursion

The technique involves generating entity-dense summaries of identical length, each iteration adding missing entities while maintaining the same word count. Each summary becomes denser and more self-contained.

Claude Skill Prompt (11 Elements)

A structured approach to crafting effective prompts. Elements range from required to optional but powerful.

The Core (Required)

1
Task / Goal
What you want Claude to do. One clear verb. The most important line in any prompt.
2
Context / Background
Why you need this, who it's for, what situation it's solving. Claude reasons better with more context.
3
Input Material
The actual content to work on: transcript, URL, document, data, code. Paste it or attach it.
4
Output Format
How you want the response structured: bullet list, HTML, JSON, prose, table, email, etc.

Make It Sharper (Recommended)

5
Role / Persona
Tell Claude who to be. Shifts tone, depth, and vocabulary to match a specific expert.
6
Constraints / Guardrails
What to avoid, what not to include, word limits, tone restrictions. Prevents unwanted outputs.
7
Examples (Few-shot)
Show Claude what good looks like. One or two examples dramatically improve output quality.
8
Audience
Who will read or use the output. Adjusts complexity, jargon, and assumed knowledge.

Advanced (Optional but Powerful)

9
Reasoning Instruction
Ask Claude to think step-by-step before answering. Improves accuracy on complex tasks.
10
Success Criterion
How you'll know the output is good. Useful for evals and getting Claude to self-check.
11
Fallback / Uncertainty Handling
Tell Claude what to do if it doesn't know or the input is ambiguous. Prevents hallucination.

SPICE Framework for AI Adoption

A five-level model for achieving organizational AI transformation, from strategy to ethics.

05 Ethics for AI Transformation
04 Cultural Shift towards AI-First Mindset
03 Innovation through AI Culture
02 Personal Integration of AI
01 Strategy for AI Adoption

CAIO-Lite Organizational Structure

A practical AI governance model. Not a C-suite hire, but a senior manager or director reporting to the COO.

External Coach
CAIO-Lite
Workflow Owners
Agent Champions

Can be internal (upskilled) or external (contracted)

Tier 1
Tier 2
Tier 3
External

What the CAIO-Lite does:

  • Owns the weekly AI scorecard
  • Coordinates workflow owners and champions
  • Runs monthly sprint planning
  • Removes blockers cross-functionally
  • Presents quarterly results to leadership

Time commitment ramp-down:

Q1-Q2: 50% · Q3-Q4: 30-40% · Year 2+: 25-30%