Skip to content
All posts

Prompt Framework: RISEN

Whether you’re new to prompting or a prompting veteran, frameworks can be super helpful to get more reliable and consistent results from your AI models.

The core idea behind RISEN is ensuring you provide the AI with a comprehensive, structured "recipe" for completing a task. 

No more vague instructions or half-baked expectations. 

RISEN gives you a framework to capture all the essential elements:

  1. Role: Defining who the AI will be - an advisor, a creator, a problem-solver, etc. This establishes the perspective and capabilities you need.
  1. Input: The specific data, information or instructions you're providing as the starting point. Think of this as setting the scene.  
  1. Steps: The detailed directions you want the AI to follow. A clear, step-by-step process for reaching the desired outcome.
  1. Expectation: Explicitly stating what you want the final result to look like. The goal or deliverable you're aiming for.
  1. Narrowing: Any restrictions, limitations or specific areas of focus to guide the AI's efforts.

If you’re thinking – "That sounds like a lot of work upfront!" I get it and you’re not wrong.

But trust me, the time invested in crafting RISEN-structured prompts pays dividends in the form of higher-quality, more reliable AI outputs.

Here's a real-world example:

Role: Chief AI Scientist

Input: A blog post about LLMOps

Steps: Read this blog post. Provide me with a 300-word summary on what the blog entails.

Expectation: a concise, informative summary detailing what LLMOps is, how to implement LLMOps, and how to create alignment with LLMs and my business

Narrowing: Focus on making this summary relevant for a general business audience, avoiding overly technical jargon.

Why you should try a framework

Frameworks like RISEN transform the AI from a black box into a collaborative partner, where you're both working towards a clearly defined goal. The more thoughtful and intentional you are with your prompts, the more valuable the AI's contributions will be.

When I’m experiencing hallucinations with an AI, I lean on frameworks to help the AI narrow in on my request.

Of course, RISEN isn't a silver bullet. There will always be some trial and error, and you have to be diligent about monitoring for potential mistakes or biases. 

But as a framework for prompt engineering, it's an absolute game-changer.

So the next time you're working with an AI assistant, whether it's Claude or any other model, give RISEN a try. I guarantee it will elevate the quality, consistency and reliability of your outputs. 

And don't be afraid to experiment - the more you use it, the more intuitive it becomes.