The RACIO Prompt Framework — How to Talk to AI
Why Most People Get Poor AI Results
Poor AI results almost always trace back to poor prompts. The tool isn't the problem — the question is. This chapter gives you a systematic framework for writing prompts that get professional-grade output every time.
Compare these two prompts for the same task — writing a quarterly sales summary:
Beginner: "Write a quarterly sales summary report."
Result: A generic template with placeholder sections, zero relation to your actual business.
Expert: Full RACIO prompt with role (experienced sales director who tells stories with data), specific numbers (Q3 target 3M, achieved 2.6M, 87%), audience context (management presentation tomorrow), format requirements (executive summary under 3 sentences + core data + highlights/challenges + Q4 outlook), and tone guidance (professional but not stiff, under 600 words).
Result: A polished, ready-to-present document that reflects your actual Q3 performance.
The difference isn't the tool — it's the information density. The more specific context you give AI, the more specific and useful the output. This is the single most important principle in this book.
The RACIO Framework
RACIO is a five-element prompt framework that covers every dimension AI needs to produce high-quality, targeted output:
R — Role: Who is the AI playing?
Assign a persona: "You are a senior marketing director with 15 years of B2B experience." Role-setting shapes tone, depth, and perspective. More specific roles produce more expert output.
A — Action: What exactly should AI do?
Use specific verbs: "Write / Analyze / List / Summarize / Compare / Rewrite." Never use vague requests like "help me with this." If another person reading your prompt wouldn't know what to produce, AI won't either.
C — Context: What background does AI need?
AI knows nothing about your specific business, industry, customers, or situation. You must supply this. Include: your industry, the use case, the target audience, relevant data or constraints.
I — Instructions: What are the constraints?
Word limits, output format (list/paragraphs/table), tone requirements (formal/conversational), forbidden phrases or approaches. Clear constraints dramatically reduce revision cycles.
O — Output: What form should the final result take?
An article, an outline, a table, a script, a code snippet? Specifying the output form helps AI structure its response correctly from the start.
RACIO in Action — 5 Industry Cases
Case 1 — Admin: Overdue payment reminder email. The RACIO version specifies relationship context (3-year partner), exact amount, previous communication history, and desired tone (polite but firm, give them a face-saving explanation). The output is a ready-to-send email, not a template to fill in.
Case 2 — Sales: Pre-meeting client analysis. With role, product details, client profile (200-person regional logistics company, VP with operations background, 30-minute meeting), and meeting constraints in the prompt, AI produces specific probing questions, realistic objection scenarios, and concrete response strategies — not generic sales advice.
Case 3 — HR: 30-day onboarding plan. Specifying company size (50 people), solo HR, high early attrition, and flat culture produces a lightweight, executable week-by-week plan with clear ownership — not a Fortune 500 enterprise onboarding manual.
Case 4 — E-commerce: 11.11 campaign planning. Budget ($50k RMB), team size (2 people), GMV target (1.2M), and channel constraints produce a specific day-by-day plan with budget allocation — not a theoretical marketing strategy.
Case 5 — Learner: Rapid concept mastery. Specifying background (no product experience), time constraint (2 hours), goal (participate in meeting conversations), and output format (analogy + example + 3 jargon terms) produces an immediately applicable learning output, not a textbook definition.
5 Advanced Prompt Techniques
1. Few-shot examples. Show AI an example of what you want. "Write in this format: [example]" works better than describing the format in words. AI excels at pattern-matching from examples.
2. Ask AI to ask you first. For complex tasks where you're not sure what you need: "Before you start, ask me 3-5 questions you'd need answered to do this well." This surfaces the key variables you forgot to include.
3. Break big tasks into steps. Don't ask for a complete business plan in one shot. Ask for the structure first, confirm it, then write section by section. Each step will be higher quality than the all-at-once version.
4. Specify what you don't want. "Don't use industry jargon," "don't exceed 300 words," "don't open with a cliché," "don't include more than 3 recommendations." Negative constraints are as valuable as positive ones.
5. Iterate in conversation. AI conversations have context. "Make paragraph 2 more persuasive" or "give me 3 alternative versions of the opening" builds on what's already been established. Iterating in one conversation is more efficient than starting fresh with a new prompt.
Your Prompt Checklist
Before you send any prompt:
- Role defined? "You are a [specific role] with [X years] of experience in [domain]..."
- Action specific? Clear verb: write / analyze / list / compare / summarize...
- Context provided? Industry, audience, relevant data, use case...
- Constraints stated? Word limit, format, tone, things to avoid...
- Output format clear? Article / outline / table / bullet list / code...
5 Most Common Prompt Mistakes
- Too vague: "Write a report" / "Give me some advice" — AI can only guess, output is generic
- No context: Forgetting to tell AI your industry, audience, or business — the output will be irrelevant
- Asking for too much at once: Dumping an entire project in one prompt — each section will be shallow
- Not iterating: Giving up after one disappointing output instead of refining in the same conversation
- Treating AI like a search engine: "What is X?" — AI is best at doing things, not just answering lookup questions