Yesterday I was hosting a webinar with Microsoft team about AI and HR. While demoing a great model-driven app one of my colleagues created (Denisa 💛), I made a beginner's mistake, and that led to an error in the plan we had. Luckily we were able to recover fast, and even prove the point on why prompt engineering is a big deal. During the demo, I was using this app to retrieve information about the candidates through Copilot but one thing led to another and I forgot to prompt correctly. However, in this post, I will show you how AI Prompts are here to help you, which is a little bit of a different story.
What are prompts?
"Prompts are the way to communicate with large language models (LLMs), the driving force behind generative AI technologies like ChatGPT and copilots." (Nirav Shah, VP, Dataverse)
In short, these are the different inputs that we will give to our solutions in order to retrieve data. If you have been using platforms like Copilot or Gemini, you know that it can take up to 5 or 6 messages to get a desired and valid outcome. Being able to prompt correctly will help us retrieve the data in our first attempt.
AI Builder recently released a feature that helps you create your own AI Prompts and use them in apps or automation. I wrote a short blog about it, you can find more details here: Your own AI Prompt library (anainesurrutia.com)
There are four existing models that you could use and tailor as you like:
Summarize text
In this case, I added a song and I want the prompt to summarize the main topics included in it:
All prompts can be tested before publishing, so you can see if they are good or not.
And the results we get are the following:
Extract information from text
Template prompt can be found below.
Classify text
Template prompt can be found below
Respond to a complaint
Template prompt can be found below
There is also a fifth case where you could create a text using GPT.
Have you tried any of these Prompt templates? If you haven't find more information here: