YOUR AI ISN’T UNDERPERFORMING. YOUR PROMPTS ARE.

R — ROLE

Tell the model who it is. Before anything else, define the expert persona the model should adopt. A marketing analyst thinks differently than a financial advisor, even when looking at the same data. Setting the role up front frames the perspective, vocabulary, domain expertise, and communication style for everything that follows.

Be specific. “You are a marketing expert” is vague. “You are a B2B SaaS marketing strategist with 15 years of experience in demand generation and account-based marketing” gives the model a much sharper lens. The more precisely you define the role, the more targeted and relevant the output.

Ask yourself: Have you clearly defined who the model should be and what expertise it should bring?

A — ACTION

Define the task clearly. State exactly what you need the model to do. Be specific, unambiguous, and direct. A well-defined action prevents misinterpretation and wasted iterations. “Help me with email” is not an action. “Write three subject line variants for a product launch email targeting enterprise CTOs” is.

The Action step tells the model what task it’s performing — not how to perform it (that comes later in the Prompt step). Think of it as the assignment headline. Clear actions lead to clear outputs.

Ask yourself: Could the model complete the task from your description alone, without asking follow-up questions?

P — PRIME

Ask what it knows about the topic. This is the step most people skip, and it’s the one that makes the biggest difference. Before providing your full prompt, ask the model what it already knows about the subject. This does two things: it surfaces the model’s existing knowledge so you can see what’s accurate and what’s not, and it reveals gaps you need to fill with your own context.

After you send the Prime question, wait for the model to respond. Read what it says. Then, in your next message, provide your full prompt — correcting any mistakes, filling in gaps, and adding the specific context the model needs. This two-turn approach is what separates RAPPEL from single-shot prompting frameworks.

Ask yourself: Have you checked what the model already knows before loading it with instructions?

MULTI-TURN CHECKPOINT

Wait for the model to respond to your Prime question. Read its answer. Then continue.

P — PROMPT

Write your full, detailed prompt. Now that you’ve seen what the model knows (and doesn’t know), write the complete prompt. Include all context, data, constraints, examples, and step-by-step instructions. This is where you provide everything the model needs to produce the output you want.

Unlike reasoning models, non-reasoning models benefit from explicit instructions. Tell them how to structure the output. Tell them what tone to use. Tell them what to include and what to leave out. The more specific your prompt, the closer the output matches your expectations on the first try.

Ask yourself: Does your prompt include all the context, data, and instructions the model needs to produce the output you want?

E — EVALUATE

Evaluate and refine the output. Don’t accept the first response as final. Review the model’s output critically. Check for accuracy, completeness, tone, and quality. Ask the model to revise, expand, or correct specific parts. Iterate until you’re satisfied with the result.

The Evaluate step is where most of the real value gets created. Your first output is a draft. Good prompting is a conversation, not a transaction. Each round of feedback makes the output sharper, more accurate, and more useful.

Ask yourself: Have you reviewed the output against your original requirements and refined where needed?

L — LEARN

Encode the process for reuse. Once you have output you’re happy with, don’t just close the window. Ask the model to document the method, create a reusable prompt template, or summarize what worked well in this interaction. This turns a one-time conversation into a repeatable process you can share with your team.

The Learn step is what separates professionals from casual users. Over time, you build a library of tested, refined prompts that consistently produce quality results — instead of starting from scratch every time.

Ask yourself: Have you captured what worked so you or your team can reuse this process?