Updated: Why I Use The RICECO Prompt Framework For AI Marketing
- Chris McLellan
- 6 days ago
- 6 min read
A simple and memorable prompt framework is a key part of building responsible and scalable AI marketing practices.
Editor's Note:
I updated this post to reflect the prompt structure I currently use. I previously relied on the ROSES framework, but as best practices have evolved, I now use RICECO, which you can think of as the ROSES Plus version.
Quick Summary
RICECO is a prompt framework designed to improve the clarity, consistency, and results of AI generated work. It is an evolution of the ROSES framework.
RICECO is an acronym that stands for Role, Instruction, Context, Examples, Constraints, and Output.
RICECO refreshes the 'Role' element. Instead of asking an AI to act like a human persona e.g. "You are a marketing pro", you assign it a system role (e.g. "You are a content generation system") that defines its function.
RICECO, like all prompt frameworks, serves to help marketing teams reduce trial and error, keep outputs on brand, and reuse prompts across tools like Custom GPTs, NotebookLM, Sora, N8N, and Voiceflow.
Most AI power users do not write prompts manually anymore. They use a small helper tool, sometimes called a Prompt Bot, to generate first drafts of RICECO prompts based from a short brief.
This post by Chris McLellan, principle at the Friends Electric AI marketing consultancy, includes a free prompt to kickstart your own Prompt Bot.
Why AI Prompt Frameworks Matter
Prompt frameworks are having a moment. As AI adoption spreads across marketing teams, there’s growing interest in ways to structure prompts to get more consistent, brand-safe results. The goal is clear: reduce trial and error, save time, and make outputs more predictable.
You’ll often see names like PEEL, OCAR, RAC, and FAB tossed around in AI marketing circles.
Each has its niche, whether it’s for generating ad copy, formulating plans, or writing product descriptions. But not every framework is built for versatility.
Why I Use The ROSES Plus Prompt Framework
I currently use the RICECO prompt framework because it’s relatively simple, clear, and covers most of the bases. It’s easy to remember and works across use cases, from drafting content and advising on growth programs to planning campaigns.
Many of the AI marketing practitioners whom I trust use it, too.
What Many Folks Misunderstand About AI Prompts
Most business users underestimate how critical prompt frameworks are.
Without them, outputs become wildly inconsistent, off-brand, or even non-compliant. That leads to wasted time, excess iteration, and compliance headaches
Frameworks like RICECO aren’t just helpful, they’re essential for scaling AI content safely and effectively.
In fact, research suggests that 68% of businesses now provide prompt-engineering training to their staff, both technical and non-technical. And organizations that use structured prompt-engineering processes report 34% higher satisfaction with their AI deployments.
Most well-constructed prompts aren't written from scratch each time. They're stored and reused, often baked into purpose-built bots like Custom GPTs, or saved in prompt libraries that support ongoing tasks.
That’s where frameworks like RICECO shine: they make prompts modular, scalable, and easy to update without starting over.
Defining the RICECO Framework
At its core, ROSES Plus is a flexible structure that helps you think clearly and draft better prompts, faster.
Here's the full breakdown of the framework with explainers for each step:
R: Role
Define the role the AI should take, but treat it as a system role, not a human persona. You are telling the model what function it performs. Examples: a system for analysing documents, a system for rewriting content clearly, a system for evaluating marketing plans. This keeps the model focused and reduces drift.
I: Instruction
State the single task you want completed. Be specific and avoid combining multiple objectives. Examples: summarize this section, create two options, identify gaps, refine the positioning statement. One task at a time produces higher quality work.
C: Context
Provide the background information the AI needs to understand the request. This may include the audience, the purpose of the work, the platform where the output will be used, or any business constraints. Clear context prevents inaccurate assumptions.
E: Examples
Show or describe what good looks like. This can be a short snippet, a reference style, a link, a screenshot, or a brief description. Examples dramatically improve output quality because the AI can pattern match against them.
C: Constraints
List the limits the AI must respect. Examples: no invented facts, no jargon, keep the tone neutral, ask for missing details if unclear, follow Australian healthcare terminology. Constraints protect quality and reduce errors.
O: Output Format
Tell the AI exactly how to structure the final result. Examples: short summary first, then detail; three column table; two options plus a recommendation; bullet list only. Clear formatting instructions save editing time and improve usability.
Pro Move: Build a Prompt Bot
You can use RICECO anytime you need clear and reliable instructions for an AI tool. It works across custom bots, research assistants, image and video tools, and agent platforms. It is especially useful in marketing, where consistency and quality matter.
However, the truth is that few AI power users, including myself, handwrite the first drafts of prompts.
Instead, they create a 'Prompt Bot' helper to do it for them.
When you open the Prompt But, it asks you to provide a short explanation of the prompt you need and how it will be used. It then generates a clean, well structured RICECO prompt which you can copy and paste into a new chat stream, new saved bot, or wherever it's needed.
This is faster, clearer, and usually produces better results than writing prompts by hand.
It can feel slightly meta that the Prompt Bot itself runs on a RICECO prompt, but this is what keeps it consistent and predictable.
After you paste your new prompt, remember that you can improve results by also uploading an example of the kind of output you want, even if it's just a screenshot. This helps the AI understand your expectations for tone, structure, and quality.
Free Resource To Kickstart a Prompt Bot
The following text can copied and pasted into a saved GPT or Gem to create a Prompt Bot that can be used generate FIRST DRAFTS of new prompts, whether they are single use prompts or those intended as saved instructions for custom bots:
Copy this text into saved GPT or Gem to kickstart your own Prompt Bot:
Role
You are a system for generating high quality RICECO Plus prompts. You turn short user briefs into complete, structured prompts that can be used in any AI tool.
Instruction:
Your job is to create a complete RICECO prompt, but you must first collect all required briefing information. Ask the user to clarify the following six items before drafting the prompt:
The purpose of the prompt
Where the prompt will be used
The audience for the output
The key informational sources it should prioritize
The preferred output format
Any constraints or limits the user wants applied
Do not generate the RICECO prompt until all six items are confirmed.
Context:
After receiving the six briefing items, identify the task, audience, expected outcome, informational sources, constraints, and formatting needs. Use only this context when constructing the ROSES Plus prompt.
Examples:
Use any examples or reference styles the user provides. If none are provided, proceed using only the confirmed briefing information.
Constraints:
Do not invent objectives, context, or constraints.
Do not expand the task beyond what the user requests.
Do not add examples unless asked.
Do not proceed until all briefing items are clear.
Ask for clarification whenever needed.
Output Format:
Produce a complete RICECO prompt with the following sections:
Role: describe the system’s function for the task
Objective: define one clear outcome
Steps: list two or three specific actions the AI should take
Evidence: explain which sources the AI should prioritise
Style: define tone, length, and formatting
Constraints: list the limits the AI must respect
Inputs: define what the AI should ask for when information is missing
Output Structure: define exactly how the final answer must be presented
Provide only the finished RICECO prompt unless the user requests additional notes.
Happy Prompting!
Hope this shed some light on my process and informs your next AI marketing project.
Until next time.
I’m Chris McLellan I’m a Certified Chartered Marketer and the founder of Friends Electric, a personal marketing consultancy that blends data, AI, and strategy to accelerate growth.
I help startups, scale-ups, and established brands across Canada, the UK, and the US with AI-enabled marketing, product launches, and integrated campaigns.
👉 Let's connect