Build an AI‑driven Dashboard

Hello everyone and welcome to today's session. My name is Rhys and I'll be your moderator for today. We are a couple of minutes out from from starting. We're just waiting so everyone has a chance to join before we properly go live. Uh, as you will have seen in your email inbox, if you'd like to join in live with this session, there is a little bit of setup required so that you can get access to the Omni platform. The link to do so is in the session resources. So if you have a look in the video description and, uh, the pinned link in the chat, we've got a little, uh, 3-step process that you'll need to complete before, uh, you're able to get access and code along with us live. However, it's pretty straightforward. So yeah, you will have time to do it at the start of the session. Uh, we've also got the slides that we're gonna cover in there as well. So please do check out the pinned link in the chat or the session resources link in the video description. They're both the same link. So yeah, go check those out now and please do, uh, continue with the setup if you haven't done so. Already. For anything else in terms of registering for the session, if you haven't done so already, you can scan the QR code that's on screen now. You can also check out the link that's in the chat as well. You can also head over to datacamp.com/webinars where you'll find this session as well as all of our future sessions as well. This is, we're pretty much bang in the middle of our AI Agents in the Enterprise series. We've got plenty more sessions coming up next week, so please do check those out. I'll drop a link so that you can check out the series later on. But for the moment, the main link that you wanna be clicking is the session resources link that is pinned in the chat and at the top of the video description. So please do check that out. If you have any questions or notes throughout the session, let us know in the chat. We're gonna be answering your questions throughout the session as well as at the end of the session as well. We will a couple folks from Omni to help you out with any technical questions that you might have as we go. So please, uh, yeah, don't hesitate and get involved in the chat. For anyone that's new, thank you for joining. We are just about to get started with today's session. We're just waiting so everyone has a chance to join before we get going. There is a little bit of setup if you want to join in with us, uh, live today. Everything that you need is in the session resources link and the pinned link in the chat. So yeah, those are your best place to go if you'd like to join in live with us. However, if you do just want to watch along and catch up with the recording, make sure you register for this session. You can scan the QR code on screen now. You can find the link that I've put in the chat, and you can also head over to datacom.com/webinars as well. If you have any questions or notes, let us know in the chat. And yeah, please do like the video if you enjoyed the session. It'll help recommend it to more people like yourself. Brilliant. Uh, I think I will hand over to Richie now. Uh, but yeah, please do make sure that you check out the, uh, link pinned in the chat and in the video description for the setup info, the slides, and anything else.

Brilliant.

With that, I'll hand over to Richie and yeah, we'll get started.

Hi there, data scamps and data champs. This is Richie. Welcome back to the Agents in the Enterprise series. So we've had a lot of fun so far this week. Uh, we've been messing about with OpenClaw, building things with NHN and Pinecone. Yesterday, uh, we were on about AI use cases in learning and development. We've got more sessions for you next week, but today the focus is on business intelligence. So, uh, we're turning our attention to using AI for dashboards. So AI is, uh, both useful for helping you design and create dashboards. It can also help users understand the dashboards you've created, and I feel this is perhaps the biggest shift in business intelligence in recent memory. And today we're going to make use of the Omni BI platform to build our AI-driven dashboard. So Omni's one of the kind of the perhaps the newer generation of business intelligence platforms. So some very exciting stuff to see. Our guest is Peter Whitehead, who is a solution engineer at Omni. So we go straight to the source. Welcome, Peter.

Hello. Thank you for having me.

Yeah, great to have you here. And what I must say, what a great mustache you have.

Likewise. You know, always love to meet a fellow mustacher.

Absolutely. I guess anyone in the audience as well, if you have a mustache, please do let us know in the chat. So Peter works closely with Omni's customers to help them build out BI solutions and got over a decade of experience in both analytics and data engineering. So he's had stints in Google's Looker team and also at Astronoma. So with that, please take it away, Peter.

Awesome. Thank you, Richie Cotton. Well, I'm really happy to be here and introduce you all to Omni and help you walk through building out an AI-driven dashboard. Before we get started, please, if you haven't signed up, please take a look at the registration link pinned in the chat, also in the description. You'll have time to go through the setup. It is really easy and you can start coding along with me as we're building out this dashboard using AI and other interactions with data directly in Omni. So to get started, we are gonna today build an AI-driven dashboard. And as I said before, let me jump over to here. We're gonna go through a little setup. This is going to help you get set up so you can follow along today. It'll be really, really good to have everyone actually going along as we go through a pre-built dashboard on an Airbnb dataset, but then also creating this ourselves. You'll have plenty of time as I go through the intro with Omni. What we're going to cover today, we're going to cover Omni as a platform. I will teach you who Omni is, what we are, and how we approach analytics. We're going to learn how to interact with the dashboard, a workbook, and the underlying semantic model of Omni. We, we have a layered approach that I'll get to in a little bit in the intro. From there, we'll start the code along and we'll start actually building content together. So we're going to start on a pre-built dashboard and see how we can ask questions of the data. We'll move into our workbook environment and start from scratch recreate that dashboard and personalize it, make it our own. Then finally is we'll utilize AI directly to make AI summaries on the dashboard and see how we can put AI throughout this entire dashboard. Then finally, we'll also edit the model, all that good stuff as we're building this out together. What is Omni? Omni, we are an AI analytics and business intelligence platform. We're helping customers small and large, you know, get the most out of their data. If that's with leveraging our semantic model to have AI conversations, being able to interact with it in a traditional sense as well. We are built on from decades of experience within the data realm. You know, as Richie said, I came from Looker myself, and our founding team also came from Looker. Our CTO actually was the first commit to the dbt project. So we have a ton of experience within the modern data stack. And then we also are, are backed by some, um, large, large of those, uh, industries as well, with Snowflake, Databricks, Google all investing into us. We just raised our Series C funding, um, so we definitely have a place in the market and we're excited about our growth trajectory, uh, going forward. Now, Omni, how we approach data. As I said, you know, we are built on top of a semantic model. That semantic layer is where data teams can define their metric definitions, their business context, as well as security parameters of who has access to what rules. Now, this is very important in the age of AI because we've seen how text-to-SQL, um, is great for fast and, and loose questions, but if you ever need to make data-driven and business decisions you know, you want some governance backing that in so you can trust in the accurate results. And we've seen this in the market and we've seen how users are getting the most out of our data through conversational AI backed by that semantic layer. But we also trust that, you know, it's not just a one-way street with natural language. Users sometimes want to take over the wheel and actually start interacting with the data, you know, in a traditional BI sense or even a traditional Excel sense. So that's why we've brought all of these modalities of interacting with data into one platform. If I want to start my explore— exploration of data with conversational AI, I can do that. And then what I can do is pop open my workbook environment and I could start interacting from a traditional point-and-click sense. And then maybe I want to, you know, open up a spreadsheet and use Excel functions. I can do all of that within one window and then finish it off by modifying it, maybe creating a window function in SQL. That's really where Omni shines, is interacting with data in specific ways or all the ways for all the different personas across an organization. So how we approach data and how we've built, um, our product, we've built it with this layered approach. At the bottom is the SQL database. We connect directly to a SQL warehouse where you might have transformed your data from DBT, all that good stuff, and it is set up in a star schema or in the AB schema, however you want to model it out. Then we have our shared semantic model on top of it that allows the data team to define your business logic in metrics, in one central location, but not only is it defining your business logic, it's defining your business context because this is extremely important as we'll show in the code along today of like if I'm asking questions of the data, I need it to be able to answer it correctly for our business. As an example, different organizations have different definitions of what a customer is. If I ask me, show me my recent customers, Pretty vague question. Without that context baked into the shared model, the AI model doesn't know exactly your definition of customers and what it needs to present back to you with that SQL query. So that shared model is the metrics layer, the business context layer, security controls as well. And then on top of that is the workbook layer. And this is what we'll get into today when we're building our dashboard. It is built on top of the foundation of the shared layer, so everyone starts from the same governance standpoint. As you're— you all are logging in right now, you are given a dashboard. All those dashboards are baked on top of the shared model, and then within our workbook, we can modify that shared model as like the foundation and building blocks into it. And then finally is each of the dashboards, um, are going to be built on top of that workbook. So each dashboard has their own workbook attached to it. Great. Hopefully everyone is logged in right now, so I'll, I'll move into Omni and we'll kind of walk through what that is. And then Omni is very, very easy to understand, I believe, like especially from the conversational analytics standpoint, so you'll be able to catch up if you're still logging in right now. We'll start with the lay of the land. This is the Omni homepage. Right up front and center, you're going to be able to see our natural language entry point. We've seen many customers use this now as the entry point into any data questions that they can ask. We even have those bottom little quick questions that you can ask like, what can I ask? Show me some data, where it understands the context over what you have access to. 'cause we can permission this out on a user level. Say some users might not have access to certain tables or they might not have access to certain columns where the AI model can't actually show you those. Now our AI agent, we call 'em Blobby, is going to be able to answer us quick questions of the data so we can get insights into what's underneath there. Over here is always how you're gonna be able to navigate back home. So if we start going to a dashboard or workbook, you can pop this open and you can navigate back to the home and get back to this homepage right here. Within the hub and Shared with Me files, this is where you're going to be able to find, uh, your dashboard that we're going to be working on. So we'll see the, the first dashboard and then we're going to start recreating it. So in Shared with Me actually is where you'll find one once you've created, um, your user and logged in. Each of you should have your own dashboard from there. And then if you wanted to start from scratch yourself, you can always hit this New tab up here and you'll see an option to jump into a model and start creating, uh, a workbook from scratch. Great. So let's navigate to that first dashboard right here. So if you go to Shared with Me you will find a brand new dashboard built out. Everyone's going to have the same dashboard. We've replicated it for everyone as they're logging in. Now what we're going to do is we're going to navigate to this dashboard and we're going to start asking questions with AI. And then what we'll do is we're going to, we're going to navigate to the workbook and see what lives within that workbook. So let's do that right now. So here it is. You can go to Shared with Me and find that dashboard. I'm actually going to go to My Documents because that's where mine lives. I'm going to pop this open. So here's what the dashboard looks like. You know, it has some high-level KPIs right here, some, you know, histograms, line charts, bar charts, map— maps, all that good stuff. Now with Omni, each one of these is a standalone SQL query to the underlying database. It's currently constructed using that semantic layer, and that semantic layer then issues the SQL query down to that underlying database. You know, we are an Omni database agnostic. So if you're sitting on top of a Snowflake warehouse, Databricks warehouse, Postgres database, all of that's gonna be compliant and able to connect directly to Omni. So now that we, we're looking at this dashboard, what I might wanna do is ask a follow-up question. As of right now, I do not have a filter or anything on top of here, but you know, I have my AI agent down here at the bottom, so I can always ask a follow-up question on my dashboard, or I can even ask it to summarize my dashboard. Say if I was an executive at a company and I just wanted to say, hey, summarize this dashboard for me. Since our AI agent Blobby knows the context of your business, it understands, you know, what metrics are important to you and what you're looking for, so that they can actually give you a summary in a, you know, what, what matters to you. We're here, it's going through the dashboard and actually telling you what it's finding. You can read this like a newspaper. It can even present you follow-up questions you might want to ask, where I can ask it, okay, how does occupancy vary by room type? It can give me that. The AI Agent is a great way for us to get follow-up questions here, especially for those users who are onboarding at a company or who might not be as data savvy, they can come in here and just continue to follow up and iterate on this rather than only having a path of like in a traditional BI sense of drilling into the data. I can always get into finer-grained detail or exploration right here with the agent. Even doing things along the lines of, you know, a follow-up analysis, like a deeper dive. Like this chart right here, available trends— like, what is causing these big decreases in the availability right here? I want to know that. So I can click these three dots right here on the top right of this chart where it says tile options. I can open that up I can ask a follow-up question right here. Not only do I can select the AI assistant down here, but I can do it specifically for certain visualizations. Right here I can say, what is causing the large decreases in availability? And we can have our AI agent go off, do some research, pick apart, um, you know, what may be causing this. As it's building a query right here, looks like it's doing some analysis. And this can be going through an agentic flow where it's passing results between queries as it is right here and figuring out what might actually be causing Seasonal demand. Great. All right. Now what we want to do is see how this dashboard is actually constructed. What actually makes up this dashboard? Now that is the workbook. Each dashboard is comprised of a workbook and tabs within that workbook. So if I wanted to open up the workbook and see where these, these, uh, these, uh, SQL queries are created I can hit this Workbook button up here on the top right corner, and that will navigate us over into the workbook environment. Now, this workbook environment is very much the traditional BI that you might be used to. Over here on the left-hand side is the field picker. It's going to be comprised of measures and dimensions. Um, now I'm going to go to this listings by room date so I can see a dimension The dimension is, you know, a dimensional value in your database, like a column in your database, or it could be one that's derived or calculated. A measure is going to be an aggregation over those dimensional values. So think about sums, averages, counts, all of that good stuff. Now when you select one of these— so let me edit this right now. I can— you can see I'm now in a draft, so changes I'm making here live only in this draft, so I'm not affecting content for other users. Um, if I wanted to, instead of looking at listing count, maybe I wanted to look at average, um, revenue for that. I can simply just grab that average revenue and deselect listing count and my chart has now updated with those new measures. We can see the results here. It's now looking at average revenue count, and I can even see the SQL underneath the hood that is being issued by me doing this. So this is that governed metric. We've defined it as it's an average of this dimension right here. Okay, so let's start from scratch. So what I'm going to do is let's hit the Omni O on the side over here. And go back to home. Now what I'm gonna do is, as I said, we've seen a lot of the ways users are starting to interact with the data and enter into an analysis is via our chat interface. So we can just ask a question of Blobby, our AI agent right now, and say, Can you show me the average— average, let's see what I wanted to do here. Maybe the average nightly price, unique posts, average revenue. And total listings, something like that. Just, I want to see a, a smorgasbord of metrics. In here we can then actually just have a pure natural interface, or a natural language interface. So here we got a visualization, um, it's going to give me a summary of this. I can continue the conversation here, or what I can do is I can go up to this chart right here. I want to use this chart, but I want to modify it in a bit. I can see I have a couple of options up here on the top right of this visualization. One is if I want to see the query details. With our AI, you always have that human in the loop, in the way, the option to debug what's happening. I can see what's actually happening right here, what fields are actually being used in this query. I can ask it to explain itself. If I wanted our AI agent to explain why it chose this, we can do that. But what I might want to do is actually I want to take over the wheel and actually modify this. If you've ever tried to have AI do something like, hey, move that chart 2 pixels over that way, it's frustrating at times. Sometimes it'd just be faster if I was able to do it. So I just selected that and I'm going to modify this here. Now let's get this chat window out of the way. So now I am, uh, in the workbook environment and I'm going to see I have a visualization, but I have multiple results here. So I'm going to recreate that first visualization that we had on that chart. So let's go back to the chart and what I can do is hit this add button. This is what we call our KPI visualization. We have a lot of options here where I can have a number, a comparison chart, um, even a line chart baked right into this, or even if you had images or text, things like that. So let's add a brand new number. Um, this number, now that we're selected into this, I can scroll to the right over here and I can choose which, um, value or measure I want to be displayed right here. So now I have that unique host directly in there. I'm going to add another one. Let's add that average listing or average revenue. And let's add another one. And then we can have that listing count as well. I can change the style if I wanted to, maybe put this in the center, align it a little better. Then even if I wanted to change the text sizes, I can do that. Now that we have this visualization and we want to start recreating that dashboard that we saw, all I need to do is go up to the plus dashboard button up top here, and I can hit Create Dashboard. I'm going to say Peter's— it gives me an AI summary version. Well, let's just use the AI summary version because that looks good and even gives you a suggested description. So I'm going to save that. And now what we are is in the edit mode of the dashboard. You can see now I'm in the draft mode of this dashboard. You can see how I can add charts, or I can add filters, controls, all that good stuff that we will get into. Um, I can also see that we can modify this dashboard. I can move it around. So let's put it over here on the right as we did before. And there we go. Now I have a nice KPI visualization directly on my chart. Sweet. So if I wanted to continue to add charts or tiles to this, I can always just select this option right here to add it. Or I can go back to the workbook and add a chart by adding a brand new tab. Now you can notice that we have each tab is going to be named with an emoji and just query. This is how you, uh, differentiate them. You can always right-click on this and rename it. I'm going to rename this KPI chart. And save it. Now this one I'm going to rename to, um, I believe the next chart we can do, um, available days, something like that. Now in Omni, we can either start directly from SQL if we wanted to, to create a visualization. If you had a CSV locally on on your computer, you can always upload that into Omni and we can start querying that directly. Or we also have a spreadsheet UI, but I'm gonna jump to this topic. Now the topic is going to be a collection and curated dataset of joins that you can interact with. So each of these tables is comprised of those dimensions and measures, But we've predefined the join paths to these tables within our model already. So you can see I already have two joins to this inside Airbnb table, to the calendar table, and the review table. Great. So let's create that chart where we wanted to look at available days. And let's look at the date field. I'm gonna grab that date field. Here we are. So we can look at the chart. Maybe I want to filter this to be, um, on a different date. So let's add a filter. So to add a filter, you can go to the date field over here on the left side and select the kebab menu right here. From there, I'm going to hit Filter, and the dropdown menu is going to allow me to select in the past, between, before, on or after. So I'm going to say this is going to be on or after. Let's go back to maybe November of last year. And now we have a chart with those drops, everything like that. So what I want to do is— I like this chart, and this is what was used in the previous dashboard, but what if I think maybe with those drops I might want to think about having a 7-day moving average? Now there's a couple ways we can think about doing this. We can always add a trend line, um, in here where if I wanted to add a reference line I could, um, put it— or not this one, I believe it's in— nope, I'll figure that one out later. Um, or we can have, um, an Excel calculation. In Omni, we have a one-to-one syntax match with Excel. So if I wanted to, I could hit a column to the right. So let's just go up here, and you can see the kebab on this column right here, and I'm going to select, um, add a column to the right. Now this is a one-to-one syntax match with Excel, so if I wanted to, I just hit equals and I can start writing my Excel formulas, everything like that. Now if I wanted to do a rolling average, maybe a 7-day rolling average of availability, I can go down to the 7th column or 7th row. I'm going to double-click into the cell, and you can see with the cursor now in the cell, I can then start typing. So I can hit equals Average, and then I'm going to select B1 colon to B7. And then once I close that out, I can hit Enter. And now let's double-click into the title, and you can change the name of the column, and I'm going to call this Rolling Average. So now that we have that rolling average, I can move back to my chart and I can go down here on the options on the right-hand side, and I can scroll to the bottom and grab that rolling average, and I can drag it up into the y-axis with the available days as well. So now I have that rolling average alongside my available days. Awesome. Now the next thing I want to do is let's create that map visualization. So that was a really cool heat map, and we want to see that alongside, um, these visualizations as well. So let's grab that. So let's add a new tab, and we can change the name just so I have— know what's going on down here. Um, let's call this Heat Map, and we'll save this. And again, I will go into my topics. From here, what I can do is I can see I have a location field. Now Omni does support lat-long, uh, we also support, you know, um, zip codes, everything like that for our mapping types. So what we can do is hit latitude longitude, and this is a row-based value for listings. So now I can go to the chart I can just choose map. The map region I want though is going to be a point map because I want a heat map, and this should be projection. We go over to Amsterdam. Zoom in. Do I need to— oh, I need to add the latitude and longitude. Latitude. Longitude. There we go. So now I have the latitude and longitude on here. You can always— it looks like we've hit a row limit here, so we have more than 1,000 results. So we do have a row display limit here. It's— we would still query the entire dataset, but we're only displaying a certain number of rows. What I can do is always go to this Limits option on the top right, and I can up this to, say, 20,000. Done. Now this is going to query this again and looks like we got all of our rows. Size, maybe we might want to adjust this size to something Along the lines of— they're pretty big, maybe 5. That looks a little better. As you can see, when we highlight over these, we're getting a tooltip and it's just the latitude and longitude. Maybe I want to learn a little bit something about each of these listings as I highlight over them. So what I can do is then just start grabbing dimensions as well. So maybe I want to know the, um, hosting name. Maybe I want to know, um, the listing or the rental type. Um, maybe I want to know nightly price. Um, and maybe I want to know their average review, right? Something like that. Now what we can do, now that we have those fields, um, in here, you can see we have this tooltip option down on the bottom in the op chart options over here. So I can scroll up and put these tooltips in here. So I'm going to get rental type in there as well. We can get nightly price in there as well. And average score. So when I hover over some of these, you can see that I'm seeing, you know, who the host is, the short-term rental, or the rental type, which is short-term, and the nightly price. So you get a little bit more detail about each of these, um, listings directly on this map. Great. So now let's see how our dashboard's turning out so far. Looking pretty good so far. So let's, uh, maybe modify this right here. We got our heatmap down here that I like. Um, yeah, I think we can do maybe a couple more. Um, let's add, you know, a listing by room type. Now, um, Let's go back to our workbook. So we've seen a couple modalities how we can create visualizations, where it was we used our point-and-click UI to create this visualization. We've used Excel functions as well. But you know, you're still going to be able to interact with AI directly in the workbook as well. So if we go over to these sparkles over here on the left-hand side, I can select that sparkle and I can ask our AI agent a question just like, show me total listings by room type. Now it's going to construct this query on my behalf. I have started using this more and more just to get me to a starting point. So I don't need to spend, you know, 15 clicks to get to the, the analysis I want. I can just have it give me that starting point and then I can start building on top of it. Maybe instead of a bar chart, let's, you know, just— well, let's grab the bar chart as we did before. Now we have room type here. Um, we can also do something like maybe pivot. So like add a stats bar chart here. So if I go to— we're on, uh, rental type now. I can then select the kebab menu on the side and I can pivot on this, and this should break it down. As you can see, long-term and short-term. So not too many long-terms, But what we can do is look at the chart, and we can even stack this a little bit. Oh, maybe we want to stack it on— stack it on room. No, there we go. No. And then we'll go rental type as the colors. There we go. So here we are, we have the small sliver of long-term up there as well. Great. Now the other thing we can do— oh, I'm gonna actually update this name before I forget. Um, listing by room count. Now the other thing we can do is we can start, um you know, as I said when I was talking about the, the layered approach of Omni, Omni's shared model, as we can see, this is exactly what this is right here. We have the shared model. This was blessed by the data team. They've defined these metrics for everyone to use. But me as an end user, I might need to, you know, make a subtle change or maybe create my own metric, um, So we can do that right here within the workbook. For example, maybe I wanted to look at, um, average review score, but I only wanted to look at the review score by a specific room type. Maybe I wanted to look at the review score of all the private rooms. One way is that I could filter on this here, but it might be a metric I want to use all the time. So I'm going to go over to the average review score metric over here. I'm not going to select it yet, but I'm going to grab the kebab menu and I'm going to go down to the modeling portion and I'm going to select duplicate. I now have a copy of that metric which I can modify and say average review score and change the name. Now the way I'm going to change it is, as I said, I want to look at only review scores for private rooms. Now to do that, I can add a filter within my measure. Now that measure then would be an aggregate over a CASE WHEN statement. So I'm gonna add that filter here and I'm gonna search for that room type field. So let's grab that room type field here and then I'm gonna select private room. There we go. I'm going to update. Now within this field, what we can do is do some modeling as well. All of the things we're doing right here is generating the semantic layer on the backend, which is great because some people prefer to have that GUI interface and some people prefer to be in the YAML code. Those different personas right here. And I like to think of this more of that build-as-you-go workflow. I want— I don't want to be switching context between codebase, exploring the data. You can do all of your— that modal or all those right here. So let's call this average review score, uh, for private rooms. Um, I now have that. I tell what view it wants to be. I can change the description if I want. So average overall review score across listings. It's going to be across private room listings. I can change the format if I wanted to. This is us being able to modify the drill fields. So whenever I drill into an aggregate value, it is a query that makes up what that aggregate value is. So you can modify what you want to actually see in that follow-up query as well. So let's add this. And now you'll see that this should only give me reviews for that average review, you know, nulls for everything else. So we've just created a metric on the fly that we can use. So I'm going to remove this. I have my average review score for private rooms. I can now have my average review score as well. So So you can look at these measures side by side, um, and maybe we want to look at this by, um, property type. Great. Maybe let's look at this in— cool, let's look at this maybe as a a table visualization. And what we can do is actually have this as a table visualization here. And you can set up some conditional formatting. So if I wanted to do something like, um, have it on a scale, or here we can either just like have it across rows, or what I can do is apply that color scale. Maybe I just want average review score And I wanted to see that scale across this. Or here, maybe the max is 5, the min is 0, or min is this— they're all pretty high, so maybe the min is 3, which it should show values, average score, maybe the min should be 4. Give us a little bit more. There we go. Great. We'll call this average review. Now, the next thing I want to do is create an AI summary for my dashboard. Now, OMI has an AI summary visualization. So let's say maybe I just want to have an AI summary over just hosts and like their, their properties. This is a great opportunity to let the AI do its thing. It's very great at taking in results and presenting back, uh, you know, a summary whatever context you provide for it and explaining the dataset. It's really not as great as looking at a raw schema and then coming out with a query on top. That's why we bake context into the model. But we can also do things like if I wanted to just grab a host, let's grab over here on the left-hand side, maybe a hostname, a host neighborhood, if they are a Superhost, Let's look at their, um, nightly price, maybe property type, rental type. And then let's look at, you know, um, the total estimated revenue, listing counts, and then average review score. We have a lot, a lot of details from these hosts. Now I can go to Chart and I can select this option on the right-hand side that has the sparkles on it. This is our AI summary visualization. Now it's going to take in those results that we just created and actually provide us with an AI summary of those results. Now I'm gonna actually up the limit here so we can have, um, all of our hosts in there. Um, but right now it's just taking them in and giving you just a plain summary. It has no additional context over what it is, so it is a very stock, just like, here's what I'm seeing. If I wanted to add additional context, I can then add that additional context right here. I could have something like, explain like a pirate, you know, have the AI explain who's the best host like a pirate if I wanted to. Or here it should give me a "Ahoy matey" or something along the lines of that. Or I can even have it do some fun stuff as well to make our dashboard more presentable. Maybe instead I'm like, show me, or Tell me about the top 3 posts. Also highlight the worst re— reviewed post. Add some fun emojis and markdown visual And then here it's going to take that context I'm providing it and actually give me hopefully a much better AI summary that I think would fit perfectly into my dashboard. So here we go, top 3 hosts, worst reviewed, and it gives me a fun visualization right here. So let's go now to our dashboard and finalize some of this. So we have a few things here. Maybe I want to move this over a little bit. Um, let's add— we'll see if we want this or not. Um, we have our AI summary, which I think would be great to have up top, like as maybe a headline or just a quick overview that we can have. So I'm going to move things around to make sure we get that. And then I'm going to move this over. Perfect. And then finally move this one over here, and maybe this one's over here. So an easy way for us to make this look a little bit better There we go, starting to come along. Now, not quite the exact same as the last dashboard, but I think this is, this is great and it gives us really a rich, you know, visualizations in like just like appeal as we're looking at this where I can see different types of charts. If I wanted to look at this at a glance, I can understand the dashboard right here and always I can. Once I publish this, I can then see that AI summary as well. Now the last thing I might want to do is maybe add a filter on top of this where I can go up to the Add button up top here. Let's add a filter. Maybe I wanted to look at just like a neighborhood filter so we can focus in on certain neighborhoods specifically. I can grab that neighborhood filter, I can add it. Now you can see that's added to the top here. And you can actually control which visualizations listen to that filter. Even say if this was a date filter, you can make it more of a generic date filter where maybe one visualization was filtered on booking date while the other one was filtered on listing date, so you can change it across the board. Cool. Now let's add that filter there. And we have this good to go. I think this is a perfect dashboard. Um, let's publish it. So let's publish this dashboard here, and now everyone can see this. And now really, to take it one step further, is that we can then, um, you know, make this dashboard even more of our own. In Omni, we can theme our dashboards as well. So what I wanted to do was I wanted to make this even more AI-driven. You can import JSON theme copies, and so I took the DataCamp website and I decided, hey, let's make a DataCamp website theme. So I go back to my dashboard, and that's what I did. I made this in 1 minute where if I go to File I go to theming. I now have this DataCamp, uh, dashboard theme where if I hit edit, you can see if I wanted to import an object to import it in with all of the colors, everything like that. We can see here that I have the generated theme from it, and I can save this and apply it to my dashboard so we can make it even a little bit more of our own as well. So here we are, brand new dashboard. Maybe I need to go into dark mode with those, um, some of those. There we are. All right, that looks great. So, um, we've covered a lot today, uh, In the slide deck, I did, and we kind of missed out on a couple of these slides because I was so excited to get into the product, but we are sharing the slide deck out with you. Um, you should have it, but also we— I am going to leave some leave-behind exercises if you want to, you know, take another dive into here, learn more about Omni, and try to find some answers to some questions that we have on here. Um, I've, I've put the answers directly in here as well, so you can see if you can hit those, hit those questions right. But then we also left some follow-up resources for you as well. This has links to our documentation, our training guides, so you can follow along with other videos on how to create dashboards, how to work in the workbook environment, and some of— a lot of the functionality I did not cover today because there is quite a bit I did not cover today. So I'm excited for you to learn more there. And then also I left behind our weekly product demos and blogs. Omnia as a company, we, we love to develop out in the front with everyone. So even right after this call, I'm going to our, our team's all-hand call where we have engineering demo every week, all the new features that they're building. So we want to showcase everything new and exciting that we're building out to all of our customers and potential prospects. So we'd love to for you all to log in there and see what we're doing, all the fun stuff. Great, so we covered a lot today. I wanted to just check to see if there's any more questions in the chat that I can touch on today, like I can go back into the workbook, the dashboard, anything like that.

Brilliant, thank you so much, Peter. That was a lot of fun. I love the, the DataCamp theming of the dashboard at the end. That was remarkably easy because I think that's often a sticking point is, you know, you make a dashboard and then someone in marketing's like, oh, you know, you've got to do the proper official style, whatever, and you spend ages changing all the colors. So that was instant. Um, and I also love the chickens in the background. Uh, it's very relaxing having, uh, some animal noises going on behind you.

Oh yeah, they're, uh, uh, they go crazy in the morning. Nice.

Um, okay, so, uh, we've got time for a couple of audience questions. Before we get to that, I've couple of questions for you. So I think maybe the secret sauce here was around the semantic layer. So just to make sure everyone's understood this, you just want to go through like what the semantic layer is and why it's important for building dashboards on top of it. Totally.

So that's very important. So the semantic layer, like as I said, when I, when I covered where we came from, a lot of us came from Looker and Looker was a tool that was built directly on top of LookML. We believed in governance around not just— just governance around business logic. Now in the age of AI, like, business logic is important. It is, it is extremely important for everyone to be reporting on the same metrics. Like, if two people come to the, uh, a meeting and each of them have defined, you know, sales revenue in slightly different ways, there's going to be an argument. But having trusted metrics on the back end for every user to have the same, same metric definitions is extremely important. But now, even when AI is involved, is extremely important for AI to choose the same metric. We don't want to risk AI hallucinating the wrong metric definition or joining in the wrong table, so things are slightly off. And who's making the right decision? Like, that is really the, the importance of the semantic layer in, in today's day and age.

Absolutely. I mean, uh, just even counting how many customers you have, it's like different teams want to define it in different ways. And like, simple things you think would be simple, uh, can often be a point of confusion. So I love that idea that you've got the one definition and that's going to help the AI get the right answer.

Yeah. And, and as I was showing how we created the, the, uh, metrics in the workbook there, we can also add context directly in there as well. Um, so like, you start thinking about democratizing the context building into the, the semantic layer as well. It's It's not just relying on solely the data team. It could be the marketers who are sitting closest to the business, who understand what customer acquisition costs means to them and why it's important to them. So then everyone across the business can share in that same importance as they're asking questions of the data.

Okay, nice. I also have a question about the, the chat, the AI chat you have to query what's going on in the dashboard. Board. Now I know some of the answers said, uh, this is generated by AI, you need to verify the accuracy. How does that work? How do you verify that you've got the right answer?

And of course, and that's really where I think about like the take over the wheel, like always being able to pop open into the workbook so you can either, you know, explore the data a little more, verify it's giving you the right answers, Or you can take a look, and most of that is the auto-generated summaries are what you're like verify with AI. It's fun that even here on this title, if I edit this, the summary one, you can hide or replace the disclaimer to verify. Let viewers know. We want to be very transparent about what is AI-generated and what is not. And I mean, I think we just need— knew it's AI is non-deterministic as, as, as like it's generating summaries as well. Like we want to make sure people know we have the tools to verify this within the workbook environment. So we're going to let you do that as well.

Nice. Yeah, because if it gives you the wrong answer, then that's a bit of a disaster. So I like that there are ways to go back and check things. All right. And one more question on that. So if you're getting lots of questions from your users, that seems like it's gonna provide good information on how you need to update the, the dashboard. Like if you've got 50 people asking the same question, you might be like, oh well, I'll have that feature in the dashboard. Is there a way of collecting all that information, those chats and for the dashboard developers to, to get feedback?

Yep. So our, our admins also have access to the analytics page over over here where we can actually see, you know, the AI query responses, where I can see who's asking what. We can export this and you can then actually— we've started seeing customers actually take this and use it as, um, uh, model refining and context building. We can see what users are asking, what's important to them, and actually push this back into our model as well to improve the actual AI responses.

Okay, that's cool that you got that feedback loop in there.

Yeah, so I think we don't have the ETL process running just yet for the— what everyone was asking today, but yes, it'll be directly in here.

All right, maybe like administrating dashboards, that's a whole separate webinar then.

Yes, yes, yes.

All right, so we've got a question from Don. So Don says, do you have to define all your metrics before you could use them, or can you just use columns of a table as metrics by default?

Great question, and yes, you'll be able to just use columns from the table, uh, by default. So if I go over to, um, our, um, view here, or our workbook here, like, we can use the topics— like, that's the curated approach, and we think, like, what is going to be best in the long term. But by all means, you can also just go browse all views and fields, and this is just the raw columns if you want to. If I wanted to start from here, I can just start grabbing fields, and this is just going to be the raw database columns. And then you can even start just using this area to start then building on top of the raw database columns. And that's how you start building that shared model directly here. Because if I made a change to this, maybe if I wanted to add an aggregate, maybe a count distinct here, you can see how this works. We have a Git workflow where now I have model changes staged in here so I can just start from scratch with nothing and then just start creating them right here on the fly and then promote them to the shared model and Omni will generate that model for you.

Okay, that seems pretty straightforward. Nice stuff. All right, so with that, we are at time. Before everyone dashes off, I want to say next week we've got 4 webinars for you. to continue the Agents in the Enterprise series. So we've got a session on, creating GPTs on Monday. On Tuesday, the session on, agentic data science with Zerve. Wednesday, we've got a session on, oh, I can't remember what it is. What's Wednesday? Oh yeah. Uh, so Wednesday we're doing, uh, putting agents into production. And Thursday, the session on how AI can improve enterprise search. So, uh, please do, uh, sign up for those. And if you're not already, come back then. All right. Uh, thank you once again, Peter. That was, uh, a very fun session.

Awesome, thank you all, um, and thanks for all the questions, um, and hope you enjoyed the demo along.

Wonderful, yeah, uh, great stuff. Uh, it was a pretty fancy dashboard that was created in the end, so, uh, excellent stuff. And, uh, yeah, thank you to everyone in the audience who asked a question. Thank you to everyone who showed up today. I will see you all again