Using the COPILOT() function

- [Instructor] In this video, we'll look at how to use Excel's new Copilot function to pull structured insights out of messy text. The exercise file for this demo is called Copilot function. In it, I've got a table of short text messages. Each of these contain snippets of customer feedback or brief internal notes. We're going to look to pull some insights from this table and we'll use the Copilot function. The beauty of this function is that for something so powerful, it's surprisingly simple to use, the formula will alternate between a prompt and a context. So let's pull up our Copilot function and you'll see prompt context. That's it. Those are the two pieces. So the prompt is going to be what we want our function to do. In this case, I want to say extract names of people, places, I'm going to say organizations and dates. And then our context is the actual data we want to pull that from. So that'll be our message text column right there. I'll close parentheses and we will get the list right here. This is a dynamic array. Right now, it is a little messy. It just goes row by row and pulls out any of those entities that it finds in each record. So if I wanted to make this a little less messy to look at, I could just clarify our prompt further. So right here in our formula box, I'm going to just add more detail to what I want here. If I said something like, separate into labeled columns, you'll see that now these are broken into people, organizations, places, and so forth. And this is a dynamic array, like I mentioned. So this will expand if I add another row here. (keys clacking) You will see that that gets added to our Copilot result. Let's look at one more use case, and there are so many here with the Copilot function, but sentiment analysis is a great one. Suppose you want to know whether each message sounds positive, neutral, or negative. Let's have Copilot do that for us. Label each message as positive, neutral, or negative. And then for our data, it's that message text column again, and you'll see these are all now classified. Most are positive or neutral. Looks like there is one negative here, had a frustrating meaning nothing got resolved, so that's a negative review. This demo shows some powerful examples about Copilot can structure unstructured text, whether you're pulling out facts or detecting emotions. Now, one important note, Copilot is a generative AI function. That means it's not guaranteed to give you the exact same result every time, even if the data doesn't change. So if you're working on something that requires strict consistency, like tax calculations, financial statements, or engineering formulas, this is not the right tool for that. But for tax heavy data where there's often no single right answer, this is perfect. You're not looking for an exact number, you're looking for meaning and structure, and that's where generative AI and Copilot can help.