The CISCO prompt

Prompting has emerged as a backbone of interacting with AI systems. In the past, I have shown some esoteric use cases of AI through pure prompt engineering such as making simulations for the classroom.

That said, the practice of prompt engineering involves more than just asking questions. It requires taking advantage of the AI's vast resources - the trillions of words present in its memory and priming them with the choice of a few words.

How big a difference can prompting make?

Depending upon your skill, minimal to VERY huge!

Here’s a case that I recently demonstrated on Linkedin which went viral:

I asked Google Bard to create a marketing plan for Tata Harrier (a car by TATA motors) which was inspired from a faculty member who gave this assignment to his students.

This is what plain prompt got me:

If you see, it’s a rather average output. There’s some information, yes, but there are no numbers. no detailed ‘plan’ but just generic do’s and don’ts

Now compare this to what I did next:

This output, complete with quantifiable objectives, plan, KPI and numbers with sources came from One Single Prompt.

That’s the magic of prompting.

So, what’s the strategy here?

The prompting scheme I relied on here is known as the CISCO structure.

The CISCO structure is a methodical approach that simplifies the prompt creation process, making it accessible and easy to refine due to its listing format and can be understood as follows:

Context: This involves explaining the background or situation that the prompt is based on. For instance, if you're asking ChatGPT to generate a blog post about a specific topic, you would provide some information about the topic and why it's relevant

Intent: This outlines the main goal of the prompt. What do you want ChatGPT to achieve with its response? This could be anything from generating a list of ideas to providing a detailed explanation of a complex concept

Style: This refers to the tone, style, and personality you want in the response. Do you want the response to be formal or casual? Should it be informative or persuasive? Specifying the desired style can help guide ChatGPT's output

Command: This involves providing detailed instructions and rules for ChatGPT to follow. For instance, you might instruct ChatGPT to generate a blog post with a specific structure, or to avoid using certain types of language

Output: This outlines the specific format that you want the result produced in. For example, you might specify that you want the output to be a structured blog post with clear headings and subheadings

Practical Application of the CISCO Structure

Let's consider a practical example. Suppose you're asking ChatGPT to generate a blog post about the latest trends in artificial intelligence. Here's how you might structure your prompt using the CISCO method:

Context: "Artificial intelligence is a rapidly evolving field with new trends and developments emerging regularly."

Intent: "The goal of this blog post is to inform readers about the latest trends in artificial intelligence and their implications."

Style: "The blog post should be informative and engaging, written in a conversational tone."

Command: "Start with an introduction that provides some background on artificial intelligence. Then, discuss at least five recent trends in the field, providing a brief explanation of each trend and its implications. Conclude with a summary and some thoughts on the future of artificial intelligence."

Output: "The output should be a structured blog post with clear headings for the introduction, each trend, and the conclusion."

Now, based on this can you guess, what was the approach here? Send it on our community channel and I’ll share the full prompt with you.

By using the CISCO structure, you can craft effective prompts that guide ChatGPT to produce the desired output. However, it's important to remember that the effectiveness of this method can vary depending on the specific task and the version of ChatGPT you're using. Therefore, it's always a good idea to experiment with different approaches to find what works best for your needs.

Want to learn more such techniques? We’ll soon be launching a course exclusively for professionals in academia and learning and development that demonstrates unusual use cases of AI to enhance training, learning, and research outcomes. We would like to hear what you’d want to see in the course. Click here to give us your suggestions.