Welcome everyone. My name is Sebastian. I am in the strategy team at InterWorks here responsible for everything in Europe. I've been doing webinars and trainings for half of my life and today is another opportunity for sharing knowledge, inspiration and everything else. And today I also have someone with me. Chris, would you care to quickly introduce yourself? Yes. Hi everyone and welcome to the webinar. So I'm Chris Goodman and I work for Sigma. I'm the Partner Enablement Lead here at Sigma and I work with our partners such as InterWorks in technical upskilling and showing them the best of our products. Looking forward to doing a demo later in the session and showing you the best of Sigma. Thanks for that, Sebastian. Awesome. Thank you so much. And also the voice you heard previously, our lovely Vicky here on the lower right in case you want to get in contact with her. So today we are talking about the next generation of AI in analytics, including AI apps and everything around that. But before we start talking about the really cool stuff, we also have to do a little bit of a recap of what happened last time in our BI three point o webinar a month earlier because we need a little bit of basic for or base for what we are trying to talk about today. You might remember the guy in green who stands in front of a big lake or ocean of data and the different stages we went through when it came to analytics and BI. So very early on, like the nineties, we had tools like Excel emerging democratizing the data world, but we also had Lotus in here and VCCalc, a very early one actually. But as great the disruption was that all brought, there were also a few limitations with that. You might remember we talked about performance limitations, there weren't any visualizations, everything was just tables or just data. There wasn't that much of advanced analytics and everything took a long time. Then we went into the BI one point o, where the first bigger tools started to arrive when it comes to analytics. Our Cognos of course, SAP BusinessObjects and quite a few more, again with quite a few limitations attached to them. No agility in this case, we had very long use, only batch data, only structured data, there was no streaming involved of any kind and high costs, high barriers. You had to learn everything to get into the tool and this was usually part of an IT department and nothing outside of that one. Then we went one step further into the BI two point zero with a few of the analytics tools emerging. We're talking about twenty six, twenty ten, twenty fourteen, around that area with Tableau, with Power BI, with QlikView and Qlik Sense later on with Looker or Data Studio, whatever they call themselves right now. And, of course, this was then the democratization of visualizing data. With that also for each and every business user being able to see and work with data to get to their insights later on. But, of course, when we talked about BI three point o, we didn't really get around another topic, which was our beautiful boat here and the one who will that doesn't really want to die. Spreadsheets are here to stay. And I think we made the point the last time that this is completely okay, at least for the foreseeable time. This year is going to play a big role today as well, extent. So stay tuned for that. Also, last time we talked about AI, naturally, and how it has emerged and what bigger tools currently, at least in the professional space, are used to get to all the insights, but also processes and flows and everything else that people want. This is such a big topic that we could probably study that by now. It's definitely too much for its own webinar. But you might have noticed the last time when we talked also about the limitations of everything around AI, mainly that everything is changing all the time. It's not really that resilient, yes, in terms of, yet in terms of change. That there was one word missing on that slide, Analytics. So we didn't really in BI three point zero have a slide in here. I have some content around analytics in our BI space. And with this precursor of what we talked about last time, we're also getting into our first poll after our icebreaker poll from before. Vicky, if you would be so kind to pop that up again onto our screens. How would you describe your organization's data infrastructure today? We have quite a few options in there. Again, four of them. Centralized, self-service, platform led and AI augmented. And I let you quickly click through that. Vicky is going to keep that open for a for how long are you keeping that open, Vicky? I think we'll allow another thirty seconds. That's fine. That's fine. While that is open and while you are clicking and answering, thank you for that very much, I am already going into the summary of that because, hey, we don't want to dwell in the past. How did we go or got to where we are today? So we started with BIS reporting. So we had IT owned dashboards. This was basically our BI o point o and one point o. And I'm going to pause again. Vicky, do you want to summarize anything in terms of our poll before I could just glance over that? Absolutely. So we're seeing the vast majority are still self-service analysts, which is fantastic. Or the second one, which are not so fantastic, so it's a lot of centralized reporting. But good to see that a few people are AI augmented. So yeah, all in all, thank you very much for people who participated, and we'll see you in the next poll. Awesome. So in our BIS reporting stage early on we the whole thing suddenly let me start another way we suddenly had reportings. It was not buried in one or two tables. No. We could actually report something to a layer in our professional life where decisions could be made upon that reporting. But it also came with a new problem, speed and access. Everything was super slow at the time because we had a very, very thin bottleneck. Then we had our self-service stage. Now analysts were in the speed driving reporting, driving dashboards. Everything was suddenly drag and drop tools. It wasn't that much coding anymore like in our IT owned dashboards before. But the thing came with a new problem again. Now governance and trust. Trust is the currency of a data. If there's no trust, people won't use the data. And governance, of course, most of us probably know the issues with not having something governed. Most of the tools when they arrived didn't really bring governance solutions with them. And I have to say to some extent that's still the case, unfortunately. Today we are in the modern data stack. We do have big data warehouses. We have big data leaks out there. The market has consolidated to a certain extent already. Tools like Snowflakes, like Databricks, but also transformation tools like a DVT nowadays the standard, which I'm really grateful for. So we have the pull full power of data warehouses by now, but also we have a new problem here. Business user access as I'd like to call it. So each and every stage here solved the problem from the stage before, except for the last one. And this is what we need to talk a little bit more about today. It's what I like to call the big gap between data and users. Now from our last webinar, the BI three point o, you are used to cool AI generated images. And I promise you there are one or two more in this one here. I really love the colors of those ones here. When I'm talking about the gap, I really want to see a gap in here. What does that mean? On the left side, we have data platforms. We have powerful by now. It's variable. It's scalable. It doesn't really have any hardware anymore that needs to be run. Everything is in the cloud by now. We have centralized logic. Oops. Click the far. We have centralized logic in there. We have governed solutions and governed workflows directly in those platforms already. And on the right side, have our business users. Now our business users usually want data to create something, to decide something, and very many different options. The problem is mostly that business users don't have access to the direct data coming from our data warehouses. The gap here needs to be bridged, of course. But there is a kind of bottleneck between both. So who is filling that gap usually? Now if the word analyst comes to mind, you are absolutely right. In the past, basically, historically, analysts were bridging that gap here. But there is a small problem with that or let's call it a challenge. I myself am an analyst. I have been working in the analytics space for many, many years. So let me give you one more sentence here before I go to my point here. AI is starting to compress this gap here, as you might imagine, Text to SQL, we have co pilots everywhere right now, but AI is still a little bit patchy. The analyst bottleneck, the bridge in between, isn't gone. Now my opinion here, it's shifting. The analyst's role is evolving, not disappearing. Since we have an AI app leader in our panelist team here today, Chris, can I quickly ask your opinion on my opinion here? Yeah. I think definitely we're seeing that gap or certainly time to value really shorten. We've seen in different waves, different tools coming out that really help democratize how business users can make use of that data. I still think what is common across all of that is improving the underlying sort of data literacy, data fluency of those business users. So while we're seeing really great advances with text to SQL capabilities where you don't necessarily need to know how to write SQL. I think there's still a gap around business users knowing essentially the right questions to ask of those copilot's assistance type things. So someone asking a question around, give me my top ten performing products this month. Yes. You can kind of take that natural language, convert it into SQL, run that query in the warehouse. But what really is the semantics around best performing? Is that best performing by total revenue? Is it profitability? Total quantity sold? The AI will just interpret it one way or the other. So prompt that multiple times, and you may see that actually the SQL generated will vary per iteration of that. So the business user really needs to be able to know how to prompt well to get the expected answers, but also have a level of, I'm gonna say, critical thinking in saying, okay. This number that has come back, it wasn't actually what I was expecting. When I was defining, I I said earlier, performance, I want to do that by profitability as opposed to total sales. And, yes, they could look into the SQL. And for us in our roles as data analysts, analytics engineers, we probably have the right level of fluency to be able to actually go and correct and sort of guide that, but actually the business users. So we're seeing the need there to actually with the self-service analytics being able to allow the analysts to, yes, use these copilots and assistants, but also have the ability to build it in their language as well. I have quite a few thoughts about that, but before I share one of those, actually we do have another poll. So if I could ask everyone on our call to quickly participate in the question what concerns you most about AI in analytics right now. And while you're answering, I will share one more thought or actually a question. Chris, would you concur that this gap here then is a kind of translation issue between data and humans. So my I'm coming from when people need to learn how to ask proper questions to the data, then they first need to know the data. They need to know how data works. So there's a kind of translation barrier in the needs. Yeah. There's definitely something like that in terms of I I know we kind of had a sort of pre meeting about this yesterday, and the the other point I forgot to sort of mention there is, like, as a business you as a business analyst or kind of engineer, I I have the right terminology. So if I've got some messy data and I'm saying to a tool, go and pass this out, I know the word passing, and obviously, the AI tools understand that and know what to do. So there's a there's a gap there in terms of the language. I I always recall back, I'm gonna say with, like, my parents, like, in the age of the Internet, and my dad would say, can you find this on the Internet? I was like, you you could probably find anything you want to find on the Internet. You just need to know how to search, how to query. Are you doing Boolean searches? All of this sort of thing. So there's a lot that you learn through experimentation, so we definitely encourage people to experiment. And the other point I'd say, not necessarily on this, but I think in this age with this comment you've got here, Sebastian, around, like, starting to compress the gap, I think how we build going forward is going to drastically change. I've got two children. I think kind of in a few years, if I talk to them about, oh, I used to go into this tool. I used to drag and drop elements onto a page. They'd be like, why would you do that? Like, would you not just chat? Like, would you not sort of, like, talk into a phone and say, go and build me this application? So but I think we'll start seeing this big shift in how people build as well. So overall, I think kinda key things are around, like, that data literacy, data fluency, and then thinking about how do you give access to these tools to the people so that they can get and drive value quicker because I think you probably got it on this slide around, like, bottlenecks. Like, if everyone's having to go to a data team, there's only a a certain amount of bandwidth. So actually giving that power sort of democratizing to the users to start asking the questions because they've essentially got their own domain knowledge. Yeah. Before I talk about the bottleneck here, Vicky. Absolutely. So I have shared the results with you. So it seems like accuracy and governance are the major concerns, which is something that we hear daily really, Sebastian. So thank you for participating and we'll cover a little bit more of that as we go on. I have to say hallucinations in AI are still my personal biggest blocker, not just professionally but also in my personal life. I've been disappointed so often and it's really or it has become an art to reduce hallucinations. Alrighty. So Chris mentioned bottlenecks already. So yes, there are quite a few of those, but I put a few buzzwords here on that slide and there will be more. When this gap exists between data and our business users, There are a few things that can emerge. Bottlenecks are one of those. We also have decision latency, so people need to wait longer for data to make decisions. And that, of course, erodes trust in the end. But it doesn't stop there. Actually, you will create shadow analytics with a gap like that. Because if we have belated decisions, if we have trust erosions, and if we have to wait for anything very too long, of course, we have bottlenecks, then shadow analytics emerges automatically. People make their own analytics in their own tools. And historically, this has usually been in Excel. But it's not limited to that. And, course, when you have shadow analytics, governance becomes one more necessary, but also way more complicated because you need to capture shadow analytics somehow. And usually in that process, you want to avoid shadow analytics and reduce that again. All of that together, of course, then leads to an unrealized return of investment of whatever you are currently using in terms of tool set, in terms of processes. And all of the those here are basically intertwined. Everything depends on everything or has consequences for everything in here. But we also have good news. So back to our image in here. I said before, currently, the analyst is filling that gap in here, but we have to talk a little bit more about what our business user actually want here on the right side. Because this is a paradigm that has shifted within recent, I would say months, maybe a year by now, but it's kind of fresh. So an AI, of course, is a natural driver of that. What do I mean? Let's say data somehow came over this gap and it is now available to our business users. Historically, they were then trying to get to insights. Right? I look at data. I want to answer some questions I have about the data or from the data about my business, about my department, about my strategy, about whatever I have. From that insight, something else emerges because it doesn't stop there. Sometimes I'm happy to just know something, but usually I want to do something with that knowledge. So then we have an action that is happening. What is that action? Well, quite a few things could be an action. Could be that I decide something with which is usually the case. Could be an operational decision. Could be a strategic decision. It could be that I am alerted about something or that or I want to alert someone or something about something that has changed in my data or might escalate that or I might just adjust something. So there's a new data point, a new insight, and then I change a specific process, for example, or I change something else. And that leads to a different shift or not a different shift, a different mindset outside of just analytics. Because the analytics is taking part is happening with our within our data and insights apps in here. The action tab here is goes above that. So nowadays, we are thinking more in workflows and processes. Every person who looks at a dashboard, looks at a chart, looks at an insight from data, they will usually want to do something with that. And mostly, this is something we do more than once. This is, by the way, my one and only rule to define processes and workflows. When it happens more than once, you can consider to make a flow out of that. What do I mean with flow? So let's say you have data and you want to make one decision with that data. And you know already this is a decision that I will make once and never again. Then it's not worth any workflow, not worth any process or anything else. If I do it more than once, then I can definitely consider to build a workflow around that or with that. The reason for that is actually pretty simple. If I have something happening one time, then there are certain conditions under which that one decision is taking place on a certain number of considerations that I have to make. If the whole thing happens more than once, then those conditions become rules. And when I have rules, I can build flows around those rules so that my actions trigger automatically, which, of course, naturally brings us to the topics or topic of analytics apps. And, of course, I'm giving you another great picture here, at least I think it's an awesome picture. So it took me quite a while to prompt this one. Yeah. So we still have our guy in green. Oh, by the way, compared to our webinar a month ago, he is now a hiking person for some reason. Something I did not know, AI thought, yeah, this guy is now hiking. And our man in green here, the guy in green, is building apps. This is at least what this picture is trying to show to her, well, with a little bit of imagination. Imagine that each and every boat we have, each and every ship, sailboat, whatever we have, is a lightweight app in your company. This is something like a small workflow or a bigger workflow, something that happens regularly. That could be a decision that comes day after day after day, and I need then to do something after I made that decision. A few years back, for example, I worked at an Internet company that had user generated content there and we had some very nice users, but also some very bad users and we had to moderate that. And once a day, we got a dashboard with a bunch of users who were a little bit more naughty than us and we had to decide if we punished them or not. And this was happening day by day, so we had specific criteria there for those users, had to decide what to do with them based on those criteria, the action actually took place. This is something that was very manual at the time. Nowadays, we can put all of that into an app. So we build a workflow that triggers automatically or that helps us to trigger everything less manually. And this is now where Sigma comes into play for our second half hour for today. So what is SIGMA? It's a tool. SIGMA is a browser based tool. It has a spreadsheet mental model as I'd love to call it And you will see in a moment what I mean with that. But there are a few things you need to know upfront. So first, Sigma directly connects to a data warehouse. So there's no extracts, there's no middleware. It's always live queries directly to a data warehouse. So whatever you do right in Sigma, you always live querying the data. So you can't be more up to date. We're working a lot with spreadsheets or spread spreadsheet like artifacts in Sigma. And one thing that you can remember for the future, if you want to dive into that tool to get to know it, The spreadsheet in Sigma is actually a query builder. You don't really see SQL queries unless you want to, then you can definitely see them. But whatever you click, whatever you play around in the spreadsheet creates queries back to your data warehouse. By the way, Sigma has complete stack flexibility, so it doesn't have to be just, I don't know, Snowflake. It can also be Databricks, can also be BigQuery, can be any data warehouse you want. So there's no vendor lock in when it comes to Sigma. And also, and this is something I learned way too late in my path with Sigma, Sigma inherits governance from the data warehouses. Whatever rule set is set directly in the data warehouse, Sigma just adheres to that. Then there are, of course, also two big USPs that Sigma is rightly proud of, which is write back and app building. App building, we talked about a little bit and we'll see a little bit more of that too. Chris, can I quickly ask? You recently told me that it's now called governed write back. Yes. What does that mean? Governed write back, I guess it's combining sort of two key elements. So obviously our write back functionality. So within Sigma, as Sebastian has talked to, we connect directly to the warehouse and no extracts of that data. So all of the queries are pushed down or managed through Sigma, and that enables us to not necessarily change any of the underlying base data. We don't want Sigma users to be changing base tables in the warehouse, but you can really extend it with Writeback by being able to add in calculations, update values directly in the Writeback schema, so using our input tables. And the governed part of it is combining that with essentially the governance that we inherit from the warehouse, but also providing the ability to build out these applications, and Sigma will just manage this for you. So if you're using other tools, kind of vibe coding is very popular at the moment, you're not necessarily inheriting that governance, and you're having to maybe do some additional work over what the code generation tools are generating to make sure that only the users that you want to have access to that data or to be able to have the permissions to write values, etcetera, is there. So we manage that for you. So I guess take it as managed and governed write back capabilities as a sort of combination of our features. Nice. Okay. Got it. Thank you so much. Alright. And with that, I'm going to leave my beautiful blue slide deck because we go live. So for the remaining time, we'll do two things. You will get a five to seven minute crash course on how Sigma works for everyone to get a grasp of what it actually looks like when you start with it. And then you will get a slightly longer demo about AI capabilities and what you can actually do with the tool in the end. So here we are. As mentioned before, Sigma is a browser tool, so everything is happening directly in a browser window. If you open up Sigma for the very first time, it will look exactly like that, except for my recent workbooks and other things that I've opened up in here. And you will find a very inviting big blue button. It says create new and this is what you will click on right away. So in here, you also have a quite few quite a few things you can build, like a full data model if you need to. We don't do that today. We are just creating a new workbook in here. We stay in our browser, and we are already in the workbook on page one that we want to create later on. Now I can immediately make this a little bit more colorful, like blue. Also give it a small title up here. I'm going to analyze biking data. So I'll write down biking. Way too small, of course. So let's make that a tiny bit bigger and blue. Today I'm in a blue mode. Yes. Biking. Awesome. So now that I have a dashboard, also colored it already, I'll bring in data. There are many ways of getting data in. Also, our right back solution in here, input tables. But for now, I'm just getting the data from a data warehouse. Now in here, I have different warehouses that I can choose from whatever I've connected to. In my case here, I'm going to the Sigma sample database and I'm choosing a fun one, the bikes. So there are a few tables here with data about bike trips. I click on the trips and this table is now loaded into Sigma as a spreadsheet, which is very familiar. Now I can do with that spreadsheet whatever I am familiar with or that I want to do. For example, I have a start date in here, which has also given me a timestamp that I don't really need for my analysis. So I can just keep that to a day, for example, and the time is gone. I can filter everything if I wanted to. I have a filter button up here. I also have a filter option directly in each and every column. So when I click here, filter, I could filter for a specific postcode or zip code already, by the way, with a histogram in here, which is telling me how many rows we have for each and every zip code in here. If I don't want that, I'll switch it off or delete it completely. In this case, I don't want that one. But I actually do want a filter for my subscription type. Now here, I'm going to add a control. A control is or can be many things. In this case, it's a filter users can actually click into. Not just me as the developer here, but everyone. And I see the two options that are also part of my column here, my subscription type. Let's make that a little bit smaller. Also that table. Yeah. We need to do one or two more things with that table. So we do have conditional formatting in here. Also four specific columns in a spreadsheet, which was kind of flabbergasting to me when I heard the first time about that. Let's say in my duration column for my trips that I do on my bike, I want to highlight everything that is higher than one hundred minutes. All I need to do is right click conditional formatting and I am going to say 'great to them' one hundred minutes. And anything else that I want to do? Not really. That's cool. Yeah. Maybe the color here could be purple or something like that. That's cool. Now I have conditional formatting for this column alone. And also it works in this column, but it could also work for many columns at the same time. So there are quite a few options for that. I can also, of course, create new fields, create new columns. Again, very spreadsheet like. So up here, I have my formula tab that I can write something into. For example, if I want to have my duration here, not in minutes, but in hours, I'll take that one divided by sixty, Get a new field in here that I call duration hours. And I am shortening or truncating everything here by clicking into a number and now have my duration per trip in here. So far so cool. There's of course much more that we can do. I can aggregate everything if I wanted to. New table. Let's move that up here. You will notice I am on a free canvas here. I can move everything around however I like or actually let it occupy space automatically. And this one here, I want to get rid of everything and just work with my start station name and my just built a duration in hours and aggregate those onto the station name. So I can do that here on the right with a grouping, aggregate my duration, not as a sum maybe, but more as a average or let's go with a median. The original field here for my duration, I don't need anymore. I'm going to hide that one. And now I have an aggregated table here with all my station names and the duration of my trips from there. Of course, from here, but also from the other one, I can always build charts as many as I like. Let's go with my station name here on the x axis and the median on my y axis. Simple bar chart, vertical bar chart, can build a line chart, a donut chart, a pie chart, waterfalls. Of course, I can change colors wherever I want to. And one or two more things I can add summaries. For example, my duration field that I've built before. There we are. Also, that one here to a median. So I get the overall median of my durations and I can convert this one here to a KPI if I wanted to. Create KPI element. There we are. Sigma sizes that automatically. When I go smaller, also there we are. Also the text size gets smaller. I am going to format that slightly. Let's make round corners. Let's make that light blue. Also my title is going to be bluish and my value is going to be bluish, but different blue. There we are. And something that is really cool, I think, is if I want to format my chart here the same way, I don't need to do all that again. I can just click into this element here, go to copy and paste, copy format, and in this one here, I paste to this format. And there we are. Moving my subscription type filter down here, I can move everything onto a completely blank page if I wanted to. Page. And have a dashboard without the tables, but with a chart with something I can do from here in terms of filtering if I wanted to. Everything is a live connection to my data warehouse. Everything is queried right the moment I click. So this was a seven minutes super small crash course on what it looks like when you first go into that. And with that, we are making a really, really big jump into something that is finished already. So, Chris, if I can invite you to share your screen and drive with us through the example you have prepared here. And you're still muted. Let me come off mute. There we go. So you can see my screen, and I've come off mute. So we're going well. So thanks, Sebastian, for kind of, I guess, that quick sort of run through of kind of building out sort of dashboards as an application. What I've got on the screen here is a more built out demo. We were using this demo recently at both the Snowflake and Databrick summits, and it really landed well. And I've kind of made a couple of tweaks to it just to kind of show you some some of the extra features that we have in Sigma today. So if I kind of like just walk you through what we have on the screen here. So this is a demand planning application. So in this box here, we essentially have data connected into the warehouse, and we're showing our historic sales data against a plan. We've got some KPI elements coming through here, so like you would love with KPI elements showing sort of variation over time, showing a trend, slight sparklines, So imagine the bit on the left hand side is your current state of BI, type analytics. So this is, tell me what has happened, in the past. If we move over to the right hand side, what we have here now is this building out as a sort of full application for demand planning. So this section here that looks like a table or a spreadsheet, this is our input tables, and this gives you the ability to write values back using our write back schema and then governed write back to then do calculations on that historic data. So in this scenario for demand planning, imagine I'm a manager for, as I've got on the screen here, direct sales channel in the Midwest and selling Apple products, I might want to generate a forecast for this. What really starts to make Sigma different to traditional BI applications is how you can then build in that action framework. So as Sebastian talked about, having that action framework so you go from data to insight to now go into action and build in workflow directly off the back of that. So I'm just gonna walk you through some of this on the right hand side. So for the business users, there is always that disconnect. We saw the Chasm, that we saw that Sebastian had on the slides. How do you bring the technology technological skills of IT, your data science, your data engineering teams to those business users? So what we can have here is essentially a simple button click. So if I click here, populate forecast, this could be a time series model that has been built by our data science team. They've deployed it onto the cloud data platform, and I'm able to now call that machine learning model to now go and build out a time series model. As a business user, I don't need to know how that model works. That's all been tested and built by the data science team, but I really need access to that. So I can see in my historic sales data, there's been a little bit of variation. So I wanna see how much is that variation going to change our forecast going forward. So from a simple button click, I've been able to populate that forecast. So we can see here in our input tables, we have these values coming through. If I go to this next button here, we've got a robot on a brain, so this is our kind of AI capabilities. Again, for the business user, do I need to know which large language model to use for which use case? Do I need to understand things around total number of tokens or temperature of models? All of that to me is actually very foreign, but I can kind of obstruct that away from the business user. So here we've got two different modes. We've got our AI query. So within Sigma, we can go directly to the warehouse, make use of the large language models available in the warehouse, and query those models. So I will just have it toggled onto this side, click this button, and just walk So when I say that, you mean something like Snowflake Cortex, I guess? Yeah. So within the cloud based platforms, they host and serve up a large number of large language models. So Snowflake, Databricks, other cloud based platforms will make available the latest topic, latest OpenAI models directly in the warehouse. So they have been able to build on their side SQL functions. So some of you may be familiar with SQL functions such as AI complete, AI extract, things around like sentiment. You can write those as SQL functions within your code to be able to ask these questions. So if I just come out of this just quickly, I'll just apply this and be able to see this one go through. What I can see from Sigma, when this one applies, is we're actually running those queries directly on the warehouse. So if I come to the top right hand side here, we can see the queries that Sigma is running. So if I find the right query on this list, copy in this one here, we can see, and if I highlight on the screen here, essentially, we've run an AI complete. We've defined the model, and we've got a system prompt and an overview prompt. So it's taking this prompt, sending that off to the warehouse, generating the result, and then rendering that back into Sigma. So like I said earlier, business users don't need to know the models, but also we can see within Sigma the ability to switch your models out. So if you're in production, you might want to use the higher cost model, you might want to increase the number of tokens, but if you're in development stage, you might want to use a lower cost model and reduce the number of tokens. So we can sort of set that in the administration pane so that you don't unnecessarily drive up cost. Obviously, with tooling, that is going directly to a large language model, but we can also go via a chat interface now as well. So it's a very similar approach. Instead of just going direct to the model and that previous set of prompts would have said, take our current date, so it would take the second of July, go and build out a forecast for the next twelve months, look at events that we have either stored in the warehouse or are available to you if you have access to the Internet, and go and build out a plan for things that might vary our demand. I could do that exactly in this forecast agent. So I could type into here generate a twelve month forecast for direct Midwest Apple. I know and then because this is an agent, you can add more context. I know there are World Cup events happening. Incorporate these in the forecast. Obviously, I'm typing live. So here we go. So we're we're now using an agent. So instead of it just being a query being run against the warehouse, we can come into this and say, okay. Now use the reason and action capabilities that an agent has. So while this is running, I'll just give you a bit more context from agents within Sigma. So we have, first of all, Sigma agents. So you are able to use natural language, and once this runs through, I'll kind of show you how this gets built out. Natural language sets some instructions to tell it what you want it to do. It has the context and ability to know what Sigma is, so where we have actions like insert rows, navigation through pages in a workbook, go into apply filters. Sigma agents are able to interact because they have those skill sets there. I'll just click apply on this one. But we also were able to bring in any warehouse agents. So like that, populate forecast example, if a technical team has gone and built out a Genie agent, a Cortex agent, we're able to consume that directly within Sigma. So there's no we're building things here and we're building things here. We're able to inherit and bring those across. So, I'll just kind of walk through what this has now done. So this has now taken that prompt, looked at the impact that, the World Cup would have, looked at historic sales, and recommended now a new forecast for us. So we can see here's our baseline sales, here's seasonality adjustments all driven through the time series model that we run, and then we've got event effects. Everything to date that I've shown you is all driven by machine learning and AI, but it's really important to still have the business user in the loop. I've got the domain knowledge, so I might know that a particular store might be closing soon, and therefore, we need adjust that forecast for that bit of extra information or there's other events. So I could come into here and edit this. So I'm able to now update this. So I could come into this particular week, populate the baseline forecast in here, and then actually say we're going to minus fifty because actually we know that the store needs to close for refurbishment. Submit that, and then we can see that this is now, as those blue lines are running across the screen, running those queries against the warehouse, and we can see in this line that I've adjusted, we've gone minus fifty and store needs to close for refurbishment. So we're able to build out this action framework directly within Sigma, so I can now have a transaction log, so every time someone makes a change to these input tables, I can see who made the change, the date, the time they made the change, and the reasons for making those changes. So if I ever want to go back and edit that, I I can do that if I need to. So just to show you some of this under the hood and just so it's not all sort of smoke and mirrors, if I come into our edit mode, I can edit this workbook, apply the changes I've made, and we can see how things like our agents are being built out. So if I come into an agent, we can see, like I said earlier, this is all built out with natural language, but we're also able to connect it to warehouse agents or give it a set of actions. So in this example, I've said when someone says go and generate a forecast, do the following and apply these actions. So as Sebastian was talking about, we have a whole action framework. So there's a range of actions that we can do, whether that's things like navigating within the workbook, whether it's updating values or clearing values from controls, if it's inserting, deleting, updating rows in an input table, or going out to external services. So I might want to trigger an external API, like I might want to send a notification to someone that the job has finished or that the activity has updated. Taking all of this as a whole and sort of like breaking it down, I've got if you can see on the camera, I've got lots of Lego behind me. I I like to use an analogy of actually if you break down a lot of our different applications, so whether that's a point solution SaaS tool or something that you're building today where you're extracting data out of a system, loading it in Excel, and building out something like a scenario planning forecast in Excel, if you break it down into different points, so things like our input tables, if you're doing scenario planning, that is really giving you the ability to update assumptions. If it's in an approval workflow, say someone has built out a project management tool, then our input tables help you enable the raising of new requests. Go into our action sequence, it'd be I want to click a button to go and create a new scenario, or if it's in approvals, I want to approve or reject that. So I'll just come out of this and go back into my demo. If I go back to the published version, we can see this sort of action sequence flowing through. So what I've shown you to date is essentially taking that warehouse data, building out a scenario, a forecast, making use of AI machine learning that is built by other teams, I can now move this through to an approval flow. So I can submit this plan to a manager. Will get All those buttons you are clicking here, they are all built for this workbook. Right? Yeah. Yeah. So they're all drag and drop elements. You can put them on the canvas. We have code representation of our workbooks, so you can actually start today by prompting and saying, go and build me a demand planning application. And using coding assistance or doing that with Sigma assistance, you can go and start building. So it will generate the tables. It will generate the buttons, generate the action sequence. So it will get you that sort of, like, eighty, ninety percent of the way there, and then you, as the business user, with your domain knowledge, can come in and modify that. I just finished this part. So I can come in as a manager. I would have received a notification to say, okay. There's a plan for you to review. I can come into this one, say, I'm the manager. I've looked at this plan, and I now want to approve this. That will then send a notification back to me, upstream, downstream, let people know that there's a new plan that they can use as their baseline forecast. So I think just to kind of make sure that we've got some time for q and a, I'll just pass back to you, Sebastian, but hopefully that shows you sort of going end to end from essentially that sort of spreadsheet type interface through to that action sequence and workflow that you can build in Sigma. Makes sense to me. Thank you so much for that. I'm going to share my screen one last time before we wrap up for today because of course there's one more slide I want you to enjoy. There we are, a lighthouse. So two questions, not questions, two talking points here. Number one: Yes, we are in the Q and A part. If you have any questions right now, please let us know. Write them into the chat, write them into the Q and A section directly in the webinar. If you don't want to ask them right now, of course, feel free to get in contact with us. We are happy to help. It can be something completely trivial or if you have a super complicated strategic question, also happy to help with that. Vicki, can I quickly ask anything in the pipeline yet or not yet? No questions but I do have a final poll that I would love to share which is: What is your top priority for the next twelve months? I'm afraid I haven't got remortgage house or buy a boat as an option, but just keen to get your thoughts. I'm also going to be a little bit naughty, if I may. And can I ask you to go to the next slide, where I've got a couple of events that I wanted to highlight with the wonderful people on this call today? So in the next week or next couple of weeks we've got a couple of events coming up with our wonderful partner Sigma. First of all is a hands on lab. It's a bit of an opportunity for you to get hands on and learn a little bit more about Sigma and the data apps that you can build within. And then in London on the fifteenth, we have the Sigma workflow conference. So quite a few of us are going to be there, along with all of the team from Sigma. We'd love to talk to you and demonstrate a little bit more the power of app building within Sigma. But that was it for me. I mean, I don't have any questions. We might end up giving a little bit of people some time back unless there's anything pressing that you wanted to cover. No, that works in that case. Wonderful, in which case thank you so much everyone. We've really appreciated your time here today. The music is back on. Stay if you want to. If you have any questions, reach out. Otherwise, it was a pleasure and thank you Chris for the demo and all your thoughts today.