If you are dying to understand the various GenAI prompting methods, how AI interacts with your prompt, and why this is key to optimising your results, this free prompting guide was made for you. This manual describes GenAI chatbots and the different methods of prompting. It was put forward by Frederic Cavazza, a digital transformation expert, consultant and speaker with over 25 years of experience.

I’ve known Frederic Cavazza for years and I’ve even had the pleasure of working on a few engagements with him. As we were on our way to a GenAI client workshop a few months ago, he showed me this guide and I thought to myself: “This is exactly what I’d like to share with my readers and students”.

Hence this translation and adaption of Frederic’s prompting guide, with his kind permission. I tried his tricks myself and I can guarantee that the mega prompt he describes at the end of the guide is really something you should test, copy, paste, adapt and keep in your own prompt library.

Artificial intelligence is booming, and chatbots like ChatGPT are radically transforming the way we interact with digital tools, changing the way we work. This guide aims to introduce you to the art of ‘prompting’, a key skill for engaging effectively with these artificial intelligence platforms and making the most of their potential.

If you’re researching a topic on the Web and you type in a simple key phrase, the result may not be very compelling. On the other hand, if you structure your search well, you’ll get more relevant answers. And so it goes with artificial intelligence tools like chatbots or digital assistants.

The more structured the prompt, the more relevant the outcome

The way you ask AI chatbots questions or give them instructions is conducive to more or less convincing results. This is what is called ‘prompting’, i.e., the art of formulating clear and precise instructions to guide the work of artificial intelligence models.

In essence, a well-structured prompt is like a well-formulated search query. When done properly, it shall provide relevant results.

There is no one-size-fits-all prompt methodology, as use cases differ from one user to another. However, we recommend you use one of these three methods based on your needs.

These three methods are entitled RTF (Role, task, format), CRAFT (Context, role, action, format, tone of voice) and COAT-SITES (context objective, acumen, task, specimen, impediments, tone of voice, encoding, scrutiny). Each technique works best depending on expected results. RTF was made for quick results, CRAFT, for simple questions with more accurate results and COAT-SITES, for clear cut questions and extensive results.

So, what are these methods about? Here they are in more detail.

With RTF, the prompts specify the role, task and format that AI should adhere to. It consists in a simple, “you are…, you must…, your answer must…” Role indicates who the AI bot should impersonate, providing a contextual framework. Task, gives AI the precise action or problem to be solved, guiding AI towards the expected objective. And Format specifies the type and structure of the outcome.

Should you be looking for more accurate results, it might then be a good idea to expand your prompt to incorporate more context. The CRAFT method is therefore what you would have to resort to. CRAFT implies providing specifics to the AI chatbot in your prompt such as, “I’m in charge of…, you are… you have to…, your answer should include…, choose the following tone of voice.”

With the CRAFT method, Context describes the overall situation or requirement. Role, tells AI the character it should enact. Action, specifies what AI must do, directing the LLM. Format, provides examples or details clarifying final expectations. Tone of voice, defines the expected style or category AI must follow, aligning your response with your objective and audience. Expanding your prompt to formulate a better structured and accurate result.

In the case of COAT-SITES technique, your prompt is not only expanding the context of the situation but giving AI strict guidelines to narrow the margin of misunderstanding. Allowing for your results to be more accurate and extensive. This includes Context, Task, Tone of voice, as in the previous methods but also includes Objective, Acumen, Specimen, Impediments, Encoding and Scrutiny.

Objective and Acumen, give AI the tools to reach your expected result with the correct level of expertise. Specimen and Impediments, provide clear examples and guidelines of what is wanted and what should be avoided. By providing models or illustrations, you clarify the expectations, just like defining what cannot be done narrows down your result. Lastly, COAT-SITES encompasses Encoding and Scrutiny. Encoding defines the output format syncing the results to your objective and scrutiny offers a final sanity check to ensure the results comply with your stated guidelines.

Once you have familiarized yourself with these three prompting methodologies, give them a go. Here are some tips and tricks to keep in mind when navigating GenAI chatbots. Frederic details them all in the guide for you.

After you have signed in to a chatbot, enter a few questions into the prompt window to get a feel for how it replies. After you’ve done that, try testing out the suggested prompting methods, saving a specific topic or tricky task for COAT-SITES.

When interacting with your favourite chatbot, remember to select relevant keywords, stick to one question at a time, test and tweak your prompts and follow up. Don’t settle for half-baked answers and consider asking a different chatbot to critique your results.

Lastly, remember at all times that the better formulated your prompt, the better AI can provide accurate and relevant results. So give these methods a test and see how your interaction with GenAI chatbots evolves.

The post GenAI Prompting Guide for Aspiring Experts appeared first on Marketing and Innovation.