You're good to, perfect. Oh, thank you so much. I have forgotten my love. Yeah. Great. We're we're recording now. Let me know when you want me to let people in. I'll start on Google then. Yeah. Let's start on Google now. We can we'll wait till three minutes past, I think. Hello, everyone. Good afternoon. Good morning, maybe. Just gonna wait a few minutes to let other people join in, and then we'll begin. I see everyone. Hello. Hello, everyone. Thanks so much for taking some time out today for joining us. We're just gonna start in about one or two minutes, just let people arrive and make sure we don't we don't leave people behind, And then we'll do introductions as well, and walk you through how we're gonna spend our time. Okay. Great. Okay. We're gonna start, I think. Let's let's kick off. I'm sure people will come in as we go. Thank you again for coming. Great to see. Well, this is a wonderful clapping, wonderful crack crack cracking start. Yeah. This is very exciting for us. This is one of our probably one of our biggest feature launches so far this year. We like to ship things in count, as I'm sure many of you know who are already users. And today, we're gonna do this as a bit of a a bit of a show and tell. To start off with, I I wanted to do some housekeeping. We'll probably keep this to about twenty, twenty five minutes of content. I can't promise that exactly, but that's exactly what we're gonna do, not fill the full hour. And we want to maybe leave time for questions. If you have questions, either you could put them into the chat, in the usual bottom right corner. But there's also a if, yeah, it doesn't work for you. If you put to the activities button, you'll see a q and a, option, which is one of the things that Google Meet allows, which means you can ask questions in the q and a, and that will get flagged to us as well. We'll answer as we go. But if they're relevant, I'll if you don't mind, we might leave things to the end as well. But we'd love to hear what's going on, and, yeah, that's the way to interact with us during the session. To start off with, if you don't know who I am, I'm Ollie. I'm, CEO, one of the cofounders of Calc. You may know who I am. You certainly if you don't know who I am, you don't know who my my colleague here is. This is also Ollie or Pike, CTO, my cofounder. So today, we're it's right. Ready to get Ollie out onto a onto a Google Meet screen. But, it's good to see him. Thanks, Ollie. Bye. Yeah. No worries. Thanks. Thanks to be out. And, yes, it is as confusing as you as you imagine to have two Ollie's as cofounders of count, but we make it work. Yeah. Lots of similarities, lots of differences. But we will crack on. Let's go through the agenda and and just get moving. So the plan wise, agenda wise is, well, the instructions. You heard enough from us. We're gonna set the scene. So firstly, I I think I joked a few weeks ago at the blog post about Count AI that we're possibly the last SaaS product in the world to have AI added. And I thought I'd just set the scene for why and just talk you through a bit about our thinking about AI, how we're seeing it fitting in. We hope this is useful context, but it'll take a few minutes. And then it's the demo time. So then Ollie's gonna talk walk us through some examples of how we're using count in AI and why we're getting excited, show you how it works, get you ready to, start using it yourselves. Hopefully, we then talk through a bit of logistics. So how how can you get your hands on this? How we're gonna run AI because people obviously have lots of questions about it. Understandably, we'll talk you through how you can engage with it in a way that makes sense for you. We'll leave time for q and a, and then we will close. Hope that makes sense. Anything else you wanna cover, you can dump into the q and a or in message in the message chat feed, and I'll try and add it in. But that's hopefully gonna cover most of the bases for us all. Okay. Let's, let's continue. Let's start by looking at sort of AI and setting the scene. And to do that, I want to start off with by just showing this chart as, obviously, gonna start with data. This is the Google search interest, which is a kind of relative score of how frequently AI is searched for. You can see here, it's gone up by almost sort of six or five or six times since, when OpenAI launched Chat GPT in November twenty twenty two. So you can see that explosion, that of interest. So it is now a very topical thing. A number of people I speak to in data, data leaders, etcetera, are saying that they are constantly being barraged by their CEO to understand how data fits into their world. It's a thing you probably can't ignore now. And it's got it's hyped for a reason. Max hype cycle, but there's lots of opportunity around AI in which we're gonna we're excited by. To give you a relative score, we move on. You can see where data sits in relative. So now where data used to be the darling child of tech, now AI is apparently more it's just creeping up above it into more interest. But don't worry. To make you feel a bit more that you're in the right industry, here's how data compares to Taylor Swift. So data is still better than data with Taylor Swift, which is hard to believe, I know. So this is the kind of context we're sitting in. And so data and AI, I think, is the real sweet spot, and I hope you feel that. We've I said we've been really excited by what, we think AI could bring to count. And just to set the scene again, just think about this. You've probably seen AI in every single product that you've been using already. So just a bit of product thinking here. It's clear that people are just getting to grips with AI and how it fits. And I think if I step back and describe the different paradigms of how we're seeing AI emerge, there's sort of four ways we're seeing, AI entering people's workflows. The first one on the left here is features. Just people adding in AI as like a button click, kind of a wizard. That could be, you know, smart formatting, smart data cleanup maybe in sort of spreadsheets. That could be, yeah, just simply doing a kind of a a one step click of a of a process flow filling out a form. And that's sort of feature one. Then the second version of that is then might might imply more involved is sort of the AI copilot where AI is making suggestions as you're working, getting smart auto complete. That's a very common way we've seen that. And, actually, that use case has been around for way before, ChatGPT and LLMs, but it's now, cooking on gas. And then there's the two other ones that are, sort of end injections using AI to really replace replace the human, rather than augment the human and and speed the human up. It's then more about taking, like, using a like a well, like a human as a replacement for a human. So there's a kind of, sort of the chatbot, the agent version where you're asking it more complex sort of knowledge worker at tasks, and the AI tries to solve that. And then, obviously, on the right hand side, you can have the black box AI, which is doing things that you're knowing it, and it's just telling you what to do, where where you might set, a general artificial intelligence all towards the end. As you go from left to right, you're getting less about the human engagement and working with the AI and less and less about human being involved and the AI taking on more of the workload. I, the question I think is you you none of these four are are fully mature in the market right now, and the question really is that, certain products are adopting certain ones of these paradigms. What is clear to us, at least, is that I think all four of these types of AI have a place, in the workflow for certain use cases at certain times. And, what is interesting to us is often the way that a product is picking which of these paradigms they use for AI really depends on the UI they already have. So the interesting thing I want to just make make maybe interesting for you to think about this and the way we thought about it is that really the UI of a of the software dictates how the AI works. Let me say that again. The the the UI what the UI is able to enable, what the what the of a software product, be that a dashboarding tool, an IDE, a a chat interface, that really scopes down the way the a the amount of value that the AI can bring. So this is, this is you may does that make sense? The idea that an, an AI in a chatbot interface cannot show you, a lot of information in one go. So it's limited in that way. A an AI working in an, an IDE, like the top left corner here, can't show you any visualizations here or show you, anything more than just a text prompt. So UI really matters in the AI world, and the flexibility of the UI really matters. Now I won't pretend that we, saw this coming, but one of the one of the things that Ollie and I did see when we're working with, BayData tools before we started count was how AI was already hugely limiting. The UI was already hugely limiting the workflow data that the UIs that that we were using to building dashboards, building queries, building visuals, limited the workflow and slowed it down. So in that way, we've already built count with the canvas in mind to give that flexibility. And this is what makes us so excited about AI in count. It's that the with with AI, flexibility wins. And in this regard, the canvas is one of the most flexible in space for using data. We already know with as you as you as many of our customers here will know that that flexibility allows you to do a huge amount of your work in one place, allows you to iterate and collaborate faster faster and gain trust and deal with complexity. And that same benefits of a flexible UI allows the AI the the account AI to be also very, very powerful as we'll show you soon. So I just wanna share this with you because I I I think as you're looking at AI, maybe your own products or your own software, the the the theory that we've been having and we believe in is that flexibility makes AI win and gives maximizes AI's value. And that is also a really big part of how what AI UI is also what's really limiting, data workflows and why the Canvas is also a really great partner here. So this is just to give you a sense of why we're excited, and a bit of the product theory, really, about how we've been thinking about AI and why how AI can really, really help. So just to make this more tangible, if you know CALD, if you're a user of CALD, this is not new information, but if you're new to us at all, you'll you may be helpful to hear that. But the whole idea of CALD is to give you one collaborative environment in this kind of in this canvas to let you do the full analytical workflow in one place, going from scoping out your work, thinking through problems with stakeholders as a data team, doing your data modeling, dealing with the complexity of their business logic in a two dimensional space, but then switching to things like Python or more advanced analytics to help you get to answers and do sort of statistical modeling. Then you have visualization, and then, obviously, then if you've seen some of our recent webinars about metric trees, using the flexibility of the canvas to present numbers in a in a clearer, more actionable way and all in this kind of collaborative environment. So CAL already is, we believe, the most powerful data tool in the world insofar as it's so flexible. It's fully collaborative. You you see our customers speeding up their workflows up to three times or more. And most importantly, it just makes the whole workflow more transparent and gives greater clarity in terms of, the operating model of of the businesses because you can lay things out more clearly or just you get more trust and engagement with the businesses you're working. So this is the the core of what the Canvas brings already. And when it comes to how an Account AI fits in, the power of the AI, as you can see here, is that it can work across all of these stages, making each of them far more productive, but crucially that you're working across the different paradigms as well. This is the Canvas allows to make the whole workflow more make make the whole workflow faster. So we just believe that Count AI is going to enable what is already great about the Canvas to just be, even more even more productive, even faster, even more of a a support for your thinking and problem solving, and allow more users to do more with the product as well. So hope that's an interesting introduction to where we are and what we're thinking about it, and it's very high level. But let's make this now a bit more tangible, and I'll hand over to Ollie who can start walking through how AI fits into into account. Great. Thank you. So I've just come out here. Here's the presentation we've been looking at. Down here, we've got a couple of CSVs. Now, the first thing to say is that this is in beta. So at this stage, AI can generate misleading, inaccurate things, and it needs to be used carefully. But we wanted to share it with you because we think it's already, in a place where it can give you significant productivity, improvements. Just to give you a sense of that, I've got two, CSPs here. One of orders from a fictional store and one of returned order IDs, that can be joined to this. You can connect Count AI to any data source. I'm just doing it off these for simplicity. And you may notice if you're familiar with count, there's a new button, down here. This is one of the places you can access AI. If you click on that, then you can see that, AI allows you to generate things in the canvas. So let's start with a relatively straightforward example. Let's just do a bit of, data cleaning. So for example, let's take, the orders cell, remove the column, the row ID, and create a new column called margin, defined as profit oversells. So the AI has gone through you look here that, you can see that it's got a, it's removed the the row ID column, creates a new column called margin, and, that's a simple example. We can go further with this. Once you've created things, you can then edit, SQL cells, with AI as well. Let's do that. Select it and go to the AI icon here, select that, and click edit. And we could do more here however we want. So let's say we want to filter for I know region is east, orders since, twenty twenty. So, Oli, you you've clicked on the the cell there, so it's like contextualized this to come on to the cell. Yep. Exactly. So there's kind of two states, of this. There's there's one which is you're adding new things to Canvas and another where you're editing what's already in the Canvas. So in this in this state, I'm editing. So that's just simple example of creating a SQL cell, and then editing it. Now LLMs, the we're using, like, off the shelf models for this. They have been trained on a vast quantity of of SQLs, I'm sure you're aware. So they are off the shelf, they perform incredibly well at manipulation SQL. The context we give it is, and we will go to this a little bit later, but we give it it's aware of all the cells, in your canvas. And if appropriate, if you're attached to the database, it will also know, the tables there. It will know all of the tables or cell names and all the column names and types, but it doesn't know, what is in the data itself. So it doesn't know what's in this result cell here, like all this stuff It doesn't know that. In the output. Yeah. Exactly. So that's a really simple example. I mean, it it can do a huge amount more from that than that, as I'm sure you'll you'll find out when you when you get to play with it. It can also generate other things. So that's a SQL cell. It can generate visuals. So I don't know. Let's say, for example, we want total sales by customer as a bar chart. Again, I'm gonna say I want it from orders because there's now two cells with similar columns. So if you want it from a specific one, you need to give it some context. If you don't give it a context, it will make a it would it would it will make a guess. But if you want to be sure that's gonna give you a a certain thing, you need to be as specific as possible. There you go. Yes. So it says per customer. And, again, for visuals, you can do the same thing we just did for SQL. You can select it, come to edit it, and then you can manipulate, visuals just as you can. SQL. Now this is, we think, an enormous time saver, both for existing users, but also for beginner users that, basically, you can do pretty much everything you could do in the sidebar here without ever having to touch the sidebar. So, for example, you know, flip the visual, sort by total sales standing, make it a pastel green. And that is definitely pastel. He couldn't do that that kind of stuff, that pick a click, drag, and drop, could you? No. Exactly. So, yeah, I think the way we see this is that if you're already, an intermediate or advanced user, this is going to speed you up. If you're a beginner user, it means you can get going without having to worry about, what's in the UI and how to find things. You can just tell that what you want to see. You get a sense of the kind of things you can do, down here as well. It it's pretty much everything you can do through the UI. There's a a few small exceptions like like this. It's always going to be slightly easier by just dragging it than getting an AI to to manipulate it for you, but pretty much everything else. Cool. So that's, kind of a simple introduction to to some of the functionality. So that's creating a SQL cell, creating a visual, and then editing editing the two of them. But you can actually do more than just manipulate single cells in in count. You can create as many things as you want. So let's go for creating, create two line charts. I want one, which is, average profit over sales, by month, another which is total sales by month in state of New York. And let's again tell it take these from the orders. So So you've is there any sort of syntactic that you've you've got things like brackets and you've got things like information marks? That's just to make sure you're just as explicit as possible, I'm assuming, rather than it being any sort of Exactly. Yeah. There's no, there's there's no sort of defined way you have to write things. I think you'll once you get to play with it, you'll get a feel for what it picks up and what it doesn't. I'm sure in this demo, there'll be there'll be things that get slightly wrong, slightly later on. But, yeah, the general rule is the more specific you can be, the more likely it is it's gonna get you what you want. You could do something like, you know, show me top customers from my orders table. It will probably give you something sensible back. It's unlikely to be that interesting. Cool. So that's, yeah, generating single SQL, single visual, and then, as I say, multiple things, or from a text prompt. Cool. So that's that's that's this button there. You can also access AI, through through other means. One of them is you can create objects now by using things that already exist in the canvas. And we think this can be really helpful if you perform scoping in accounts. You can scope out the analysis you want to do, select it, and then generate the actual analysis from that scoping. Let me give you a flavor of of how that works. I'll just move this stuff out of my first. For this, I'm gonna take a slightly more complex example, which will be like a multistage analysis, which you can do through a text prompt. It should work it should work okay. But often, it's it's quite good to to use the power of of the canvas and layouts. So let's do that with sticky notes. What I'm what I wanna look for is I wanna look for our our kind of top customers, and I wanna compare that against, everybody else and see how their profitability differs. I don't want the reason the returns table is here, I don't wanna look at any orders that have been returned. I wanna exclude those. I'm only interested in things since twenty twenty two. So there's quite a few things going on there. Now in order to break that down, we could do something like this. So title sorry. Table showing, unreturned orders made in twenty twenty two or later, call this cell, unreturned orders. You can see here I've been explicit about, I want you to make a cell. I want you to call it this. You don't again, you don't have to do that. It's useful if you want things to then later refer back to it. So that's the table on the ten, orders. Then we want sales by month and customer. Let's call this sales per month. And let's create another one here, which is then our top three customers, in each month. You're trying to spell it badly intentionally so it picks up? No. It's just how to spell. My fingers, it's quicker than my brain. All these top customers. Okay. And then finally, we wanna bring this together to show, just the orders from these top customers. So show the orders, from the top three customers. Oh, from well, I'll rephrase it. Sorry. From customers again, being being specific is is is the way to go. From customers who have ever been a top three customer. Let's do that. Okay. Now you could you could, collect all these and generate from there. We found through our use, it helps slightly to paint out what you think the DAG is going to look like. There's a couple of ways you can do that. You can, in each of these, say, I want to select from, you know, this cell. I want to select from this cell, etcetera, etcetera. You can be explicit. I'll just pull this result. Another way we found that seems to work quite well is you can just draw the dag lines yourself to explain to it where you think, sorry, the dependency should be. The AI is has been told to look out for these. You can see this is kind of the right thing. So we're looking for it to create a bunch of four cells that are related in this way. Okay. I'm just gonna select from the cell. Just, again, being because that's useful. Okay. Now in order to generate to the canvas, all you need to do is drag, select all those, click on the AI button, and click to generate. Suspense is killing me here. Okay. So let's have a look at this. That's awesome. We've got four cells called what we want. That's some returned orders, sales per month, top customers, results. You can go and look at the SQL in any of these to check it's what you want. You could this is obviously the final cell. You could look at the the complete compiled SQL of everything put together. They're doing a rank, under function ranks less than equals three. That's good. Another way to, like, check analysis really quick So, like, we we we we we all that that was it was not for watching right post it notes, but that must have taken, like, fifteen minute ten minutes, I reckon, if someone's gonna SQL to write all that and do it in a few minutes. That's the kind of the power of this. Right? That's the thing which is Yeah. It just removes a lot of the boilerplate. A lot of the stuff I mean, I'm a software engineer, so I've been using Copilot and and DS code for for a while now. And I think that that that helps in a sim similar ish way to this, which is it does all the stuff that isn't complex. It just saves you typing. And I think in this case, like, I still here have this This is the difficult bit. Right? What do I actually want to see? Let me let me break it up. This is the bit that actually is you know, needs to be a good analyst to to set this stuff up well. Actually, conversing that into SQL is, you know, an AI can do that, really quite well, which is exactly what why this is, I think could be so powerful. But as you say, can be quite time consuming. Like I say. And also then checking it is all audible for you just like it would be if you're working with two humans working two analysis together. Exactly. And then, you know, you can edit this, just obviously by changing the SQL if you want, or you can use AI to then go and edit the SQL, as you wish, or you can regenerate it. I mean, you can you can really do whatever you want. And what I was gonna say is that you've got the SQL audit trail, which is useful. Another kind of way to depth analysis and see what's going on is to use AI again. So go to AI here and click explain. And you can see it's here explaining the analysis, and it it gives you a summary of what's going on to get to the cell, and then it gives you each cell beneath that cell. So all the upstream cells that go to make up the analysis that leads to that cell there. You can then click on any one of these, to go and find it. Now that works for any cell in any canvas. It's not limited to AI generated cells. And particularly when you've got large complex canvases, we think it's gonna be a really useful way to go and understand what's going on, particularly if cells are in different places in the canvas and in cases where you're looking at some logic that you haven't written yourself. Again, it's just slightly quicker than reading SQL. And it and then, I guess, the the whole idea of the Canvas has been to enable two analysts, two people and data team to work together really efficiently or to help an analyst work with someone else in the business solve a problem, talk through methodology. And you can already I hope you can see how much we think this AI just speeds up both of those two ways of working. Either an individual analyst can go away and just build quicker, build multiple steps at once, and still have that lineage, that spatial layout to help understand what's going on and deal with the complexity. You also then have, the ability for someone to come in to someone else's code and understand what's happened by the click of a button. So just the auditability of some of this piece of work is done for you by the AI, doing all that heavy lifting of reading through all the code for you. But you can also see and, well, obviously, as it will develop more and more, you'll gain more gain more confidence in this, allow a user who is somewhat capable, it's all data literate, to understand the data that's already there. Using explain, but also build their own work without having to know all the perfect SQL syntax. And I think that's where things get really, really exciting. Doesn't remove the data team, but it does allow data teams, like, efficiency to just rock it. Yeah. Exactly. Great. Okay. So that's generating, from a cam from post it notes in the canvas that's generating multiple, SQL cells. You can also generate other things. One of the examples is visuals. And the way to do this, which is quite fun, is literally to draw the visual you want to see. So here we go. Let's draw, a line chart. That's squiggle. Let's go to something like, total sales, and let's go for month. And then here again, we can say we want this done by, from cell result, tell it where where to get it from. Select all of that and generate that. So I've seen loads of our customers doing this kind of wireframing of visuals already, building out metric trees just like this. And now you can basically build it if it's sufficiently well defined in one day. Yep. And and this extends just like this. You can you can jot down as many different visuals as you want, and you can mix up, SQL cells visuals, you know, do whatever you want. You could even you even connect control cells to to this so you can literally just sketch out a live dashboard, and everything should be connected and and work as you expect. Now that that's that's that. The final thing I wanted to know goes back to what I said in the beginning, which is the AI doesn't know anything about actually what's in your database. It just knows the metadata about the database. It doesn't actually have access, to database results. Let's get rid of those. Now but because the, AI can generate lots of different things in the canvas very quickly, it is sometimes useful to be able to give the AI a little bit more information, such that it can, for example, give you a version of a chart for different different values in a column. Unfortunately, by default, it doesn't have access to any of that that information. So a kind of workaround we've been enjoying using is to do something like this. So I'll just do this in SQL quickly. So I'm gonna take all the distinct, regions from there, of which there are four, and zoom in and screenshot it. In a color there. And I'm gonna explain from cell result. Right. Four bar charts. One for each region in the list to the right, showing total sales by subcategory, in a different pastel color. And then for this, you can just select that and, again, generate from there. And when you set that, it's sending a picture of that set of objects to your AI to read. That's it. So you see they've got four regions. So it's central, east, south, and then west. They're pretty pesterly. That's awesome. That is very pesterly. That's that. Now it's just so that's a very quick overview. It's it's difficult to give there's a there's a lot you could do with this. It's difficult to kind of go into each, individual, example that that that's kind of possible. And I'm sure you'll think of things that we haven't come across. But just to give you a slightly more flavor about the kinds of things that we've been doing, here we go. Let's get back to report. You can create backboards where you kind of put things in different places. I'm finished, actually. You can again, it's useful to tell it where you want to get the information from. You can add controls in. Everything should be wired up. You can reference if you're connected to a remote database, you can reference tables in that directly. It will pick them up. And, again, you can create multiple, I should say charts multiple line charts for different values of things. You can do data cleaning. And then as we saw, you can do multistage analysis just by telling it to do. That's generating. Editing code, you can you can really do anything here. You can kind of do you get it to fix your queries. You can get it to manipulate those queries. You can get it to change all occurrences or something, change the metric, add order order by limits, group by time functions. At time functions. Right. And then similarly with visuals, we're building out such that you really sort of need to use, the UI for very much tool, which kinda get as much functionality into AI so you could literally just tell it what you want to see. Cool. That's awesome. Cool. Let's now talk a bit more about I hope that there was interest useful. As you can see, it's we will keep sharing more and more examples as we go the next few weeks and months as we as we, continue. This is definitely in beta. As you may, as always been caveating, some of this is about we've waited because we want to make sure the models are mature enough to give a great experience, but it's not gonna be perfect. And we are working on ways, and you'll see as we our road map to make it increasingly robust so you can be using it again and again more and more, as you go. But it's a thing you'll learn into as you'll get used to it. The we're finding that way we write to it, a bit like Googling. You start to Google talk to it in a way which is much more applicable as you go, as you get used to it. So it's a bit of a learning curve. We love you to go on it with us, and we'll keep sharing more and more examples and hope that you will also too. So as users in the user community share where it's been really great, what what's also feeding that you have hasn't quite been what you wanted to as well. But we're gonna keep building into this. It's in beta when we launch it, but we'll keep building into this, over and over again as we go the next few months, towards the end of the year and beyond probably. Yes. Thank you. The the last thing I was gonna say, which is a really important thing for a lot of your organizations and our customers is that, you can use these AI with these AI features of confidence. All the account AI features are fully GDPR compliant. It is also, most importantly, opt in. I know that some organizations have just a flat no policy to any AI features. If that's the case, you can turn the features off in your workspace settings. It is not a thing which you're forced to use or have to have any concern over. But most importantly, we've worked really hard to make sure that the partner that the airlines we're using are within the sort of normal security requirements you've always been using with you. So any g d p any g d p r stuff we've signed with you as a customer or or that you've been looking for us to deliver for you, we there are still nothing has changed in those policies. The other thing to say as well is that we're not running queries under the hood of your database. That's why we don't, by default, have access to the old the results of of sales is because we are not primarily because we're trying to avoid an access database load without your approval. So then what what we've shown you there are some of the ways we've dealt with that. We'll work on how we can you know, we'll keep working on the experience, but the goal is to make sure that we're not driving more compute to you unnecessarily. I think that's it. Is there anything else worth mentioning about security or what to expect? Well, the only other thing, which is the third point now, is we're not creating models on your data. So don't worry that the things you do in that is gonna find its way into a bigger LAN somewhere that you have no control of. It's we're not doing that. Cool. Let's talk about launch then, because I guess people wanna be able to get their hands on this. So we have a bit of a, apology, I guess, that we will be launching these features you've just seen. Plus, as you may have noticed in the as you eagle eyed a view, a number of other features that you can probably see into the fonts you've already got in this document is not the default font. So there's a little sneak peek there. But the this that's gonna launch soon, fonts wise, but this, AI features are gonna be launched for all paid accounts on the eighth of July. And the reason for that delay is to make sure it's GDPR compliant. I don't wanna go through the details of it, but this part of that is just to make sure that we are doing things in a GDPR compliant way. So apologies that we didn't quite time it perfectly for this today, but we don't want to move this meeting whenever we can make it. But the launch is gonna be eighth of July. We'll let you know when that goes out. And then for free accounts, we'll be launching this later this summer. I'll be candidly with you. We we don't quite know the the additional cost it's gonna pass on to us from using this. We hope you can use it lots and lots and lots. So don't worry worry about it, but we therefore wanna hold off of free accounts for a bit. And all this is gonna be in there and certainly no cost to you while in beta or if you have an annual contract with us. So that's it. We are that is, we've gone to thirty seven minutes, so thank you for sticking with us. I hope that was worth it. As you can see, there's a there's a lot here. There's a lot to come. This is just the start of a big journey, but I wanted to paint for you both the features that you can see and how how how much you can improve your productivity and change your workflow already, but also just where we're gonna be going with this, which is it's gonna be part of every every, feature that you already use and loving CALV will be just enhanced more with AI over a little bit. As you know, CAT is already incredibly flexible, so what AI will let you do is just gonna hopefully blow all of our minds. So grateful you're here. Any questions, please shout now. There's certainly not the end of the or you'll be hearing all of us on on AI, and we'd love to get your thoughts or feedback or initial questions that we can feedback in as we before the eighth. It'd be wonderful. If not, I'm just gonna ask her to throw some random things on through AI as a random random request. I'm joking. Thanks. Thank you for that question, Anonymous. The question is, might have missed this, but is it possible to use the AI feature from the feature bar to summarize an analysis? I love being able to refine the prompt to summarize according to my preference. It's a powerful value add and for rubber and for rubber duck also for rubber ducking. Yes. Then we have, thank you for asking that question. I hope you all missed it as well. One of the core capabilities of count is this explain function where you can go to any cell, any output, and ask it to give you the the to verbally, in words, give you the full blow by blow how it's got to the answer. So making that lineage that's in the canvas, even easier, to understand and see. It's a great way of doing code reviews as well, if you're doing checking fuels work or getting someone to come and look through some analysis that you've done. Hope that helps answer that question. Anyone else? Can it help with Python Python sales, like It can. It can. Yeah. A hundred percent. It's similar to SQL. Like, these these, models come off the shelf knowing SQL and Python very well. I just personally don't write that much Python, which is why I didn't demo it. But it can it can it can change Even you've been the ideal ideal demo candidate. Probably doesn't like Python. I didn't say I didn't like it. I should use it. But, but, yes, so they guess the way you are stating, you can you preferentially also do in Python versus SQL if you put it in the in the prompt. Is that sort of how you differentiate? K. Great. Thanks, Andrew. Anyone else looking for the q and a section? You go to the weird selection of shapes in the bottom right corner, and there's a q and a button there. We'll happily stick around. If there are any other questions or less, feel free to to to leave. We are certainly, kind of sticking around, and we can also chat more in our Slack community, if you have not really part of that. For users, it's a really great place to learn more about calendar share best practice. The link is in the chat if you haven't joined yet. Wonderful. Okay. I think there's no more questions right now. But if there are oh, so I would yeah. Could you directly summarize a trend and it takes output from just the data's the the data cell? Could it could it read a visual and explain the narrative of a visual? Is that what you're asking, Andrew? Ollie, does that does that question make sense to you? Well, if if it means kinda read a visual, the answer is, no. Because, so why what we didn't say is it it doesn't have access to any of your database data, which means that when you select things from the canvas, to generate things from, any cells, so any any any tables or any visuals, they are all the data was removed. It's replaced with dummy data. So, just just so that, I mean, often, we found that if you send they obviously include a huge amount of information, and it often can overwhelm the model. We haven't yet looked at, you know there was a workaround, which is you copy it to PNG as I did for the those few guys region, and then you Yes. You can take a you can take a you take a screenshot of the of the of the image and then send it. You can take a take a screenshot. Exactly. But at the moment, there's no it's not hooked up to explain things in the canvas. You need to do so. I'm sure there's some kind of weird workaround where you get it to explain that. It can it can return text, but it's looking to generate what what is there, if that makes sense. It's not looking to explain. So if that's of interest to people, we could, before. I mean, as I say, this is this is already, I hope, a a a good start, but there's many more things we wanna try and do with AI, of which that could do one. Thank you, Andrew. Great questions. We'll keep talking to you. Thank you. Fantastic. We're grateful for you turning up. I hope that was a really interesting session. I bet if anything like we were can't wait to get our hands our hands on it and experiment. Certainly, we will I said, there's a journey here to go on with you, so we'd love to get more questions and feedback on it. As you say, the UI UI is flexible to do a huge amount here, so it's gonna get building out. It's gonna build out more and more and more as we go. I think for now, I'm gonna sign off. But any other questions, you know who we are. We always love speaking to you all, and we we love working with you. Thank you. Thanks.