Conversational data exploration in Power BI with Fabric and Copilot

- [Instructor] One of my keys to survival has been my ability to get data out to people quickly. And by using some semblance of best practices, like we can't create databases out of everything. I mean we can. But in the case of our ABC Construction Safety Site, we want to get better about our near-miss reporting. The theory we're chasing is that if we report more near misses, we will have less safety incidents on our job sites. And near-miss reporting can get really complicated or it could be simplified, it depends on the organization. So right now, we're testing this construction safety site list, which is available to you in your exercise files through CSV. And I want to be able to work with my safety team to show them some of the benefits of the tools they're adopting. Again, key to my survival, using tools that are commonly available in places like SharePoint. Okay, so earlier you saw me do an integration with the visual in a previous video. For this next exercise, I need to place this in a Fabric workspace. So I'm going to Export this out and I'm going to Export to Power BI. I want to be able to publish this to a Fabric workspace so that I can work with my Copilot license. So again, I have Copilot and enterprise licensing and I have Fabric all turned on. I'll choose Continue. I used the credentials I used last time, I'll choose Continue. Skipping a lot of details here just to show you how this works. Okay, let me go ahead and open it. I'd already built it, so let me open it here. Okay, and this is showing me this is my semantic model. I'm seeing my properties, my data, let me go out of this. Let me go view my dataset details. Let's go this way. Okay, it's showing it's been refreshed. This is my details, showing me my tables. Okay, I can go see my Lineage. This is wonderful, I can go open that semantic model again, takes me to add more tables. Okay, I'm going to choose this Explore option. Okay, so from the Explore option, I can do things like Explore this data. I can have it create an auto report. I can have it Create a blank report or a paginated report, just depends on what I'm doing. Right now I'm just exploring blank reports 'cause I want to see again what Copilot is going to offer me with just this simple dataset. Okay, so I'm going to go to my table here. I'm going to add ID, Notes, Date of the event. Okay, so it's just pre-building this visual. I want to actually make that a table, so let me do this. There we go. And I can see those notes in there. And again, this is just sample data. Let's see if I have City, nope. And ProjectID, so I need to bring in my project table, so that's a requirement I need to make note of. So I'm going to just size this here. Okay, so this is the basic data, right? It's the notes from the near miss, it was when it was reported. Oh, it's counting that ID, I don't want it to do that. Let's delete that. Trust nothing. Okay, I'm going to say Don't summarize, there we go. Okay, great. Let's move that ProjectID up a little bit. That way we can sort by that and then we can do reporters. Okay, so I have some of my information here. I have some other things I can do to this data model to improve it. Let me go to Copilot. Okay, now I do have a Copilot license. Again, I'm on an enabled Fabric workspace. I think I'm going to have it Answer a question about the data. Okay, so Count project states, okay, how many tables are there, those are not the questions I have. I want to know what seems to be the most frequent type of notes report. I'm going to use the word notes 'cause that's the field it's in. And let's see what this gives us. Okay, and this is all sample data right? So, but if this was real life data, this is actually taken from here and I bet if we were to go count up my sample data, which we need to do for testing, we could see how this list checks out. Okay, how about I go copy this? And this note won't update, but it will be nice when I go in to show leadership that this is the type of question you can ask. Let me go ahead and grab a copy of that prompt. Yeah, there we go. All right, so that's one insight, that's good. Let's ask it, do you see any insights to share? Okay, so the most frequent type of notes reports. Wow, this is a lot of good information. I would expect a notable pattern because it's sample data, right? And if you look at it like it's the same notes, right? The forklift, like it's just repeated to give us a pure sample. Okay, but all of this could be tested against to see if it's accurate. That can be very, very helpful. The way you can use this when you work with your leadership is you can copy things like this, and again, you could paste it in here and save the prompt to show them like how you got this information, right? And then, they can then come back and actually, you know, tell you, "Hey, I actually need confirmation that this thing is true or not true." I think it can really help allow us to communicate more effectively about the types of things that people can get from these basic datasets. Okay, this is really interesting. Again, now these text boxes are just that they're textbox. That means they're not really going to update, right? They're not going to update. I would have to do something different for that. Okay, this is really great and I need go copy that prompt in there. I do think that if you prompted it in a different order, it might give you a different result. So again, I had just asked it to go and look at all the notes. And so, I would be curious to see how much different it would be if I ask it a different way. Like if I ask it for insights first, would it have given me a different result? Now I also want to point out that again, I'm completely in a browser-based environment here. I'm not working in Power BI desktop, this is straight browser. That's a lot of power to share information and gain a lot of insight that even if the insight is more or less correct, it does give us a place to start immediately to communicate the things we need to see or we would like to see in our future dashboards. So hey, you may find this may be a key to your survival. As you learn to communicate these very complex datasets and what people can actually get out of them, it's way better to have these tools to support that process. This might be a key to your survival.