Learning how to prompt correctly and effectively holds the key to unlocking the full potential of Microsoft Copilot.
We recently hosted a career day event for a group of students from a local Technical College. At the end of the event, we invited four engineers from LouGeek to do a Q&A with the students.
One of the students asked the panel what advice they could offer to an individual on the verge of entering the workforce and seeking a technical advantage.
Normally, when you ask the same question to four engineers, you expect to get four different answers. To everyone’s surprise, all four engineers stressed the importance of learning how to prompt clearly and effectively.
What are prompts?
Before beginning your quest to master this skill the experts refer to as “prompting,” it’s important to understand what prompts are. In large language models (LLMs) like Copilot and ChatGPT, the term “prompts” refers to the starting point for generating outputs such as text or images. A prompt can be as succinct or as detailed as you prefer, but the essential element is a clear goal. In Copilot, the more specific and detailed your instruction, the better it can align its response to your intended outcome.
What information should be added to prompts to optimize responses from Copilot?
Similarly to the methods we as humans use to communicate effectively, there are strategies you can use to improve the outcomes when crafting prompts for Copilot. Prompts should include four parts:
- Goal: What do you want Copilot to do?
- Context: Help Copilot understand why you are drafting the email or calling the meeting.
- Source: Help Copilot know where to look for specific references or information.
- Expectations: Help Copilot understand how you want the email to be written or the format for a meeting agenda.
Other Useful Copilot Prompting Tips
Provide clear and concise prompts is critical to utilizing the true potential of Copilot for Microsoft 365
While Copilot is designed to understand context, reducing ambiguity in your prompts can lead to more accurate results.
Provide positive instructions
Another way to improve Copilot’s response quality is to provide positive instructions. Like a Thoroughbred horse in Kentucky is built to run, Copilot is built to take action. Telling Copilot what “to do” rather than what “not to do” is more effective.
Try using “if-then” instructions.
Using if-then instructions in Copilot to specify which action to take based on whether a statement is true or false.
Give Copilot a POV (point of view) from which to answer
This helps Copilot understand the context of your request and generate more accurate responses.
