R.I.C.K. in Action: Real Prompts for Real Work

Start with an observation

You’ve learned each letter of R.I.C.K. on its own. Now watch them work together on real jobs. As you read each example, try covering the “Why it works” section and predicting it yourself. Ask: which line in this prompt is doing the most work?

These examples are adapted from Chapter 5 of Rage with the Machine, where the R.I.C.K. Model comes from.

Example 1: An announcement people will actually read

Weak prompt: “Write an email about the new schedule.”

ROLE:
You are an experienced internal communications writer who specializes in clear, direct messages.

INSTRUCTIONS:
- Task: Write an internal email announcing our move to a four-day workweek.
- Source: No external source. Use the details provided in Context.
- Action: Draft a complete, ready-to-send email.

CONTEXT:
- Audience: 22 full-time staff across operations, programs, and administration.
- Situation: Leadership approved this change after a six-month review. It starts the first of next month.
- Goal: Inform staff clearly, answer the pay question directly, and keep morale positive without overdoing it.
- Background: Staff have asked about work-life balance for a year. Some worry about pay. Pay stays the same: different schedule, same compensation.
- Constraints: No jargon. No corporate speak. Don't imply more changes are coming.

KEY OUTPUT:
- Format: Professional email with subject line, greeting, body, and closing.
- Tone: Positive but grounded. Warm and clear.
- Quality check: Confirm the email answers what's changing, when it starts, that pay is unchanged, and who to contact with questions.
- Include: A clear point of contact for questions.
- Avoid: Exclamation points, vague language, and anything that sounds like a PR announcement.

Why it works: The Background names the real worry (pay), and the Quality check guarantees the email answers it. Without those two lines, the AI would write a cheerful announcement that skips the one question everyone actually has.

Example 2: Social media with options

Weak prompt: “Write a post about my business.”

ROLE:
You are a social media copywriter who creates content for small service businesses.

INSTRUCTIONS:
- Task: Write three versions of a LinkedIn post announcing my new AI workflow consulting service for nonprofits.
- Source: No external source. Use the details in Context.
- Action: Draft three distinct posts, each with a different opening strategy.

CONTEXT:
- Audience: Nonprofit directors who want to use AI practically but don't know where to start.
- Situation: I'm launching this service and want to start conversations on LinkedIn.
- Goal: Get qualified people to follow, comment, or reach out.
- Background: I'm a solo consultant. My approach is practical and no-hype, focused on saving documentation and reporting time.
- Constraints: No emojis. No buzzwords. Don't promise guaranteed results. Each post under 150 words.

KEY OUTPUT:
- Format: Three posts labeled Version 1, Version 2, Version 3.
- Tone: Professional but human. Confident without being salesy.
- Quality check: Each version must open differently: one with a question, one with a pain point, one with a short story.
- Include: A clear call to action at the end of each.
- Avoid: "Game-changing," "revolutionary," and generic hooks like "Are you struggling with..."

Why it works: Asking for three labeled versions turns one guess into a comparison. The Avoid list switches off the clichés AI tends to reach for.

Example 3: Messy notes to clear action items

Weak prompt: “Summarize these notes.”

ROLE:
You are a detail-oriented executive assistant with a clean, direct writing style.

INSTRUCTIONS:
- Task: Turn these meeting notes into a structured summary with decisions and action items.
- Source: The raw meeting notes pasted below.
- Action: Extract and organize. Do not rewrite or add interpretation beyond what is in the notes.

CONTEXT:
- Audience: The director and two program leads who attended.
- Situation: Notes from a 90-minute planning session. Some are incomplete.
- Goal: A clean reference document people can act on immediately.
- Background: The director prefers bullets over paragraphs.
- Constraints: Use only what's in the notes. Don't fill gaps with assumptions. Flag anything unclear.

KEY OUTPUT:
- Format: Three sections: (1) Key Decisions, (2) Action Items as a table with Task | Owner | Deadline, (3) Open Questions.
- Tone: Professional. Tight. No filler.
- Quality check: Every action item needs an owner and a deadline. If either is missing, write "Owner TBD" or "Deadline TBD."
- Include: A flag next to anything ambiguous.
- Avoid: Small talk and sidebar discussion.

Why it works: The Source, Action, and Constraints all point the same direction: stick to the notes. The Quality check turns missing information into visible “TBD” labels instead of invented names and dates.

Example 4: Designing a workflow

Weak prompt: “Help me plan a workflow for intake.”

ROLE:
You are an operations consultant who specializes in service-based workflows for community organizations.

INSTRUCTIONS:
- Task: Design a consistent intake workflow for a community mental health organization.
- Source: No external source. Use the details in Context.
- Action: Map the full process from first referral to first-contact confirmation.

CONTEXT:
- Audience: Two intake coordinators and one clinical supervisor who will use it daily.
- Situation: Referrals arrive by phone, email, and online form. Everyone handles them differently. Nothing is documented.
- Goal: One simple process every staff member follows, no matter who's working.
- Background: They serve adults in crisis and non-crisis situations. They use a spreadsheet and email. The clinical supervisor reviews all crisis referrals before action.
- Constraints: Simple enough for a new staff member on day one. Note where AI could help, but the workflow must work without AI if needed.

KEY OUTPUT:
- Format: Numbered steps. Each step: What happens | Who does it | What the output is.
- Tone: Plain English. Operational.
- Quality check: Confirm crisis referrals have a clear escalation path with supervisor review built in.
- Include: Where AI could support the work, and where human judgment is required.
- Avoid: Vendor recommendations and anything requiring a budget they may not have.

Why it works: The Constraint “must work without AI” keeps technology from becoming a single point of failure. The Quality check guarantees the highest-risk path, crisis referrals, is handled safely.

Example 5: Learning something new

Weak prompt: “Explain this to me.”

ROLE:
You are a patient, encouraging tutor who makes complex ideas clear for beginners.

INSTRUCTIONS:
- Task: Explain the concept of machine learning in plain English.
- Source: No external source. Use general knowledge.
- Action: Explain the concept, then give three everyday examples I'd recognize.

CONTEXT:
- Audience: Me, a high school junior with no computer science background.
- Situation: I have a class discussion on this next week and want to explain it without sounding like I memorized a definition.
- Goal: Understand it well enough to explain it in my own words.
- Background: I use Google, Netflix, and TikTok but have never thought about how they work.
- Constraints: No technical jargon. Assume no prior knowledge.

KEY OUTPUT:
- Format: One plain-English paragraph, then three examples of 2–3 sentences each, then one discussion question.
- Tone: Friendly and clear, like a tutor, not a textbook.
- Quality check: Read the explanation back as a first-time student. If anything needs prior knowledge, simplify it.
- Include: A discussion question I could use in class.
- Avoid: "Simply put," and technical terms without immediate plain-English definitions.

Why it works: The Goal, explaining it in your own words, shifts the AI from lecturing to teaching. The Quality check makes it test its explanation against your real level.

Think deeper: what these prompts reveal about the person writing them

Look back at the five examples and notice something: the AI isn’t the expert in any of them. The person writing the prompt is. They knew that staff were worried about pay. They knew crisis referrals needed a supervisor. They knew the director hates long paragraphs. They knew the student needed to sound like they understood, not like they memorized.

None of that came from the AI. The AI supplied speed, structure, and polished language. The human supplied judgment, knowledge of people, and standards. That’s the real division of labor in good AI work, and it’s why learning R.I.C.K. makes you more valuable, not less.

Questions to sit with:

• In which example would a mistake cause the most harm? What in the prompt protects against it?
• If two people used the exact same AI tool for Example 4, what would make one person’s result safer and more useful than the other’s?

Practice: make one your own

1. Choose the example closest to something you need this week.
2. Copy it and replace every detail with your own situation.
3. Run it and review the result with the 5-question check from AI Fundamentals.
4. Note which field you changed the most, and why.
5. Save your version. You’ll start a prompt library in the next lesson.

ROLE:
You are [role].

INSTRUCTIONS:
- Task: [what needs to be done]
- Source: [material to use, or "no source provided"]
- Action: [the specific thing the AI must do]

CONTEXT:
- Audience: [who this is for]
- Situation: [what is happening right now]
- Goal: [what success looks like]
- Background: [what the AI needs to know]
- Constraints and rules: [limits and rules]

KEY OUTPUT:
- Format: [structure]
- Tone: [how it should sound]
- Quality check: [what to verify before finishing]
- Include: [must-haves]
- Avoid: [must-nots]