Understand the ingredients of a good prompt

- Anytime you interact with an AI, like Copilot, what you're typing, whether it's a question or request or an instruction, is called a prompt. You're literally prompting the AI to respond. With almost all modern AIs, you can think about prompting, like having a conversation, where you can use plain language and engage in a back and forth with the AI to get the results you want. Some examples of basic prompts are things like, "Summarize this presentation" to get the gist of a PowerPoint presentation, or "Check this document for errors" to edit text in Word, or "Give me three ideas for a blog post" to generate new content, or "Tell me what happened in this meeting" to catch you up in a recorded Teams meeting. So these are all valid prompts and they use plain, conversational language to ask for results, but the better your prompts are, the better your chances are of getting good results. So for example, while you could chat with Copilot in Microsoft Teams and say something simple like, "Tell me what happened in this meeting," you might not get the information you're looking for and you might find yourself having a longer conversation in order to get Copilot to tell you what you really want to know. Copilot works best if you take the time to craft your prompts to focus in on the key elements of your question or request. So let's take a look at the ingredients that make up a good prompt. So the first basic step in prompting is to know what you want, to have a specific goal or outcome in mind. This could be something like the "Tell me what happened in this meeting" prompt, but again, you'll get better results if you continue to fine tune the prompt. So next, try to provide some context. Tell Copilot why you need this information or who the information is for or who it involves. This can help Copilot better customize its response to suit your needs. Next, if possible, provide a source for a Copilot to reference. Maybe instead of telling you about the entire meeting, you want it to focus mainly on what one specific person said in the meeting. And lastly, let Copilot know your expectations on how it should best meet your needs. This could be something like asking it to use professional language or to include a short paragraph at the end that provides context for future meetings or just to be brief in its response. So let's take a look at this prompt we might write in Microsoft Teams. Tell me what happened in this meeting to prepare me for a follow-up meeting with my manager, Anna. Focus on the questions Anna asked and any action items she requested. Please use simple language so I can get up to speed quickly. So in this example, the goal is tell me what happened in this meeting. This is the basic thing we want from Copilot. Next, we provide context by telling Copilot the reason we need this information. In this case, to prepare me for a follow-up meeting with my manager, Anna. Then we follow up by letting Copilot know the source of the information we want, which helps it narrow down its focus. In this case, we're saying focus on the questions Anna asked and any action items she requested. This can potentially help Copilot place less emphasis on the other items discussed during the meeting. And finally, we finish our prompt with our expectations. I'm asking Copilot to use simple language in its response and adding some additional bonus context here so I can get up to speed quickly. So with a prompt like this, you're much more likely to get a useful response from Copilot than if you simply write, "Tell me what happened in this meeting." Now, this isn't to say that you have to write your prompts with goals, context, source, and expectations every single time. The way you compose your prompts will also vary depending on what you're asking or what you need. But keeping these key ingredients in mind will, more often than not, help you get better responses overall. Now, it's also important to bear in mind that in many cases you won't just be providing Copilot with a single prompt to get your answer. I mentioned earlier how prompting is really like a conversation. So it sometimes takes some back and forth to get the answers or responses you're looking for. For example, your first prompt might be a broader request like asking Copilot for a summary of a meeting. But once Copilot has responded, you might continue the conversation by asking it to give you more specific details about what was discussed between two people or ask it to provide additional context for a conversation that happened during the meeting. Or maybe you need help with the technical problem. You could start with a question about what options you have to solve the issue, and then ask for step-by-step details on the option you decide to pursue. But always keep in mind the four main ingredients of a good prompt: the goal, the context, the source, and the expectation. And with those in mind, you'll have much better and quicker results.