Do you know the song Drop it like it's hot? by Snoop Dogg?
I'm by no means a fan of Hip Hop, but there are some songs that get stuck in your head even after years and cause the famous and dangerously emerging so-called earworm from time to time. And when this mixes with my imagination and my favorite theme Copilot Studio, I end up humming Prompt it like a boss to the rhythm of Drop it like it's hot?. Well, is that line starting to get into your head? Danger of catchy tunes? Not yet? We'll get there in the course of the article! But basics first!
What is a Prompt?
Generative AI (or as Microsoft calls it, Copilot... long live marketing!) can generate you texts, imagines, audios, videos and 3D models. However, we will focus on text input, as Copilot Studio can only generate text with the simple on-board tools.
Writing a Prompt is the process of creating input (usually text) instructing the Generative AI to generate the desired response. In other words, prompting is the way we ask an AI to do something.1
Let’s have a simple look on an example.
PROMPT: What is 2*2?
RESPONSE: The multiplication of 2 and 2 is 4. In mathematical notation, this can be expressed as 2 * 2 = 4.
OK, that wasn't really a challenge although you can outsmart the Large Language Model (LLM), as Scott Durow impressively demonstrated in his video, see screenshot below:
Anyhow, the generation of text seems to be more useful in this case.
PROMPT: Act as a close colleague of an employee at an IT company named Microsoft. Generate a congratulatory email for this employee who has been working as developer for 10 years. The tone should be sincere but playful. Add in appropriate emojis.
RESPONSE:
Subject: 🎉 Celebrating 10 Amazing Years! 🎉
Hey [Employee's Name],
Can you believe it? You've hit the 10-year mark at Microsoft as a developer! 🥳 It feels like just yesterday we were tackling our first project together, and now look at all you've accomplished. 🚀
Your journey has been nothing short of extraordinary. From writing flawless code to solving the trickiest bugs, you’ve always been the wizard behind the curtain. 🧙♂️ Your skills, dedication, and innovative thinking have set the bar high for all of us.
Remember that time you pulled an all-nighter to meet the deadline, and still showed up the next day with a smile and coffee for everyone? ☕ Your commitment and team spirit are legendary! 🏅
Thank you for being such an amazing colleague and friend. Your positive vibes and humor make even the toughest days enjoyable. 😂 Here’s to many more years of coding adventures, groundbreaking projects, and office shenanigans. 🖥️🎈
Let's make sure to celebrate this milestone with some cake 🎂 and maybe a friendly coding challenge (that you'll probably win, as usual). 😜
Cheers to you, [Employee's Name]! 🥂 Here’s to another decade of brilliance and fun!
Best,
[Your Name]
Have you ever heard of Prompt Design and Prompt Engineering?
Prompt Design and Prompt Engineering are closely related concepts, both integral to effectively utilizing generative AI models, but they refer to slightly different aspects of the process.
Prompt Design primarily focuses on the creation and formulation of the input queries or instructions given to the generative AI model. It involves crafting clear, concise, and well-structured prompts that convey the user's intent and guide the model towards generating the desired output. Effective prompt design considers the model's capabilities, limitations, and the specific task at hand to elicit meaningful responses.
Prompt Engineering encompasses a broader set of activities. It not only involves Prompt Design but also includes the iterative process of refining, optimizing, and experimenting with prompts to achieve better results. Prompt Engineering involves analyzing the model's responses, understanding its behavior, and adjusting prompts accordingly to enhance the quality and relevance of generated content.
In essence, Prompt Design is about the initial creation of prompts, while Prompt Engineering involves the ongoing refinement and optimization of those prompts based on continuous feedback and observation of the model's performance. Both are crucial for effective communication with generative AI models and obtaining desired outcomes. 2
Prompts in Copilot Studio and where to find and use them
There are 2 ways to use prompts in Copilot Studio.
Custom Copilot: You can create prompts within a topic.
First-Party Copilot: You can extend them via a prompt.
Let's first take a look at how to create a prompt in a topic in a custom copilot.
Start Copilot Studio.
Create a new Copilot by clicking on Create and New copilot.
Click on button Skip to configure.
Add just the name. Leave the rest free.
Click on Create.
After your copilot is set up, click on Topics, System, and Conversation Start. This is the topic where your Copilot begins the conversation with the user.
Click on +, click on Call an action and Create a Prompt.
The following should appear on your screen.
Give your prompt a meaningful name, for example City Sights Prompt and add as Input the key word location.
Add the following prompt. Note: Replace the normal text location by the key word location created in the previous step.
PROMPT: Can you name a few sights for location?
Note: Unfortunately, you can’t use the usual prompting power like “Act as … Do this task… The tone is …”. You will run into Microsoft’s Responsible AI topic for AI Builder. Keep your Prompt in Copilot Studio simple, like in the example!
By adding a message to ask for the location or city, you can connect the result of this question with your prompt.
Don’t forget to set the Output variable.
Add a Send a message node at the end to display the results from the prompt. Select the variable PromptOutcome.text.
When you test your copilot, for example for the city Hannover, then you should these sightseeing tips:
RESPONSE: Certainly! In Hannover, Germany, some popular sights to visit include the Herrenhausen Gardens, the New Town Hall (Neues Rathaus), the Maschsee Lake, the Sprengel Museum, and the Hanover Zoo. These are just a few of the many attractions worth exploring in Hannover.
The way how to enrich prompts for First-Party Copilot is the same as for Custom Copilot. The only difference is, that you find the Prompt action here much faster.
Click on Copilot for Microsoft 365.
Click on Add action.
Select Prompt.
The rest of the procedure corresponds to the steps from the example with the Custom Copilot.
Conclusion
Well, do you still have the catchy tune Prompt it like a Boss in your ears? Yes, no, maybe? What you definitely know is what prompts are, how prompts work in Copilot Studio and where you can find and use them. Prompts also work pretty well in Copilot Studio, as they have been available for a long time. In addition, you can use Dataverse as a data source, which is still in preview, but works quite well (see my other article: AlohaCheck: Dataverse as Knowledge Source in Copilot Studio).
And in case I still haven't got you with the earworm, here's my final attempt!
Mahalo for reading! 🥳 #DropTheMic 🫳🏼🎤 If you enjoyed it, please subscribe so you don't miss the next article!
https://www.w3schools.com/gen_ai/gen_ai_prompt_intro.php#:~:text=What%20is%20AI%20Prompt%20Writing,an%20AI%20to%20do%20something.
https://nothans.com/prompt-design-vs-prompt-engineering