Good morning, everyone. And I'd like to also say good afternoon to those of, you joining us from Europe today. This is the new in Coveo for commerce spring release session. My name is Shareen Reed. I work on the product marketing team here at Coveo, and I'm joined by my colleagues from the product management team, Olivier Titsu and Anthony Delash. And they will be the ones walking you through some of our newest commerce functionality today during the next forty minutes or so. So first, before we get started, I, of course, have a few housekeeping items to cover. Everyone is in listen only mode. However, we do wanna make sure that, we can answer any questions you may have, so you can submit those by accessing the q and a chat feature on your screen. And we do have a few minutes set aside at the end of the presentation to answer any of the questions that do come in. A very popular question of is, of course, will today's session be recorded? And it will be. And you should receive the recording and related slide where about twenty four hours after the conclusion of the event. Alright. So with that being said, before we jump into our agenda for today, I wanted to just simply highlight that although this session is dedicated to commerce, the Coveo platform goes beyond commerce, and it's a composable AI platform that supports website as well as service and workplace experiences. So at the core of the platform are those capabilities listed on the left that you see there. So semantic search, AI recommendations, a Gen AI answering, and personalized product discovery. So why is that important, and why are we highlighting it? Well, we really believe fundamentally at Coveo that AI is a game changer, and it really has the potential to change every point of experience for customers as well as employees. So it can create those, you know, remarkable interactions that also drive the bottom line. So whether that be improving efficiency and reducing costs or creating experiences that really differentiate your brand and drive up revenue. Alright. So, in terms of our agenda today, it's fairly straightforward. First, we're going to cover the latest and greatest in AI and machine learning. Then we will move on to some new search and dicing functionalities really designed to further empower our business users. And then we are going to finish it off by talking about some important data reporting and connectivity updates. Alright. So AI and machine learning. As we dive into this section, I just wanted to set the scene a little bit. Olivier is gonna really focus on the newest enhancements that we have introduced. But as a reminder that when we talk about AI at Cloveo, it's really a whole series of AI models that impact almost every interaction that customers have on your site. So it's from suggestions in the search box to the way products are retrieved and then ranked to the recommendations and the the way fastening is is driven. And, also, now, if a shopper lands on your site with a specific question in mind, maybe they wanna learn something about your products. Well, we have generative answering capability that can support that as well and guide them in the right direction. And speaking of Gen AI, the amount of hype and the mindset change that has gone over the last few months and last year is truly different than anything we've seen before. And just, you know, as a witness to that, I wanted to show you some stats from a Boston Consulting Group report, and share it. The numbers are really quite astounding. It says that eighty nine percent of executives rank AI and Gen AI as their top three priorities for twenty twenty four. So just pause and think about that for a moment. That's almost nine out of ten. And, you know, of that group, fifty one percent put it at the top of their list. So probably most of you on the webinar today are seeing the same thing in your own organizations. But more importantly, I think, is that question on the top right hand side here. And executives wanna know where the value is and how do they capture it. And with that as a back backdrop, I'm gonna hand it over to Olivier, who's going to walk you through some of the latest and greatest Gen AI functionality of Coveo and the associated road map. Thanks, Shareen. And I believe you were sharing your screen, Shareen. So, you know, for everybody who are listening, you know, here here you go, in terms of, the metrics from the Boston Boston Consulting Group, that Shareen has mentioned. In terms of our relevance generative answering now, as as some of you may know, we've been live in GA with this since last October. So we now have, you know, many customers leveraging this in production. Amongst them are some commerce customers that are using this to summarize their rich content. So here, you know, with relevant generative answering for commerce, we're able to leverage Coveo as a very smart and relevant rag system to be able to generate answers based on rich content. So here in this case, if we take an example from one of our customers, if you ask for something like tips to build an outside kitchen with a barbecue, this is where we're at now. So now, you know, using their rich content, we're able to return a summer a a great rich summary based on the snippets that we've retrieved and use the GenAI summarization functionality to actually explain and answer this question. As you can see here now, we've added some rich answer formatting to allow us to essentially mark up the prompt automatically and make sure that we're always displaying it in the right context. And I will show a bit more of that later in the demo. And we're now in early access with follow-up questions. So now when customers go on and ask you questions such as how to build, an outside kitchen with a barbecue, we're able to recommend some follow-up questions. And any follow-up question that's asked there is actually gonna be maintained in context with the prompt, that that's see that's shown above. So now, you know, I I figured the best way to show you would be to, show you a small demo of where we're at with relevant generative answering, and then we'll dive back in into where we're going in the next work. So here, this is a demo, environment that's actually built on our our documentation. The documentation is actually gonna be updated shortly with these enhancements that I'm about to show you. But you can imagine here if you were a commerce customer. Right? This is the Coveo documentation. So in this case, we're answering questions on Coveo specifically. But in the case of our commerce customers leveraging this, we're answering questions, of course, based on their rich content throughout in in their context. Right? So if I head over here, I'm a commerce customer, for instance, and I want to actually index my CMS content using a rest source. Right? So here, you know, I wanna know what the difference is between, a rebuild and a refresh. So if I type this question here, you'll see that now through rich answer formatting, we're actually able to summarize the prompt in a way that makes sense from a human perspective. So we first go through the rebuild, the refresh, and then highlight key differences. We always provide, of course, citations here that were used to actually generate the prompt. And you see here we have some follow-up questions. But now with your chance for formatting, what's really interesting is that you can actually use many other formatting types, to render your content. So if I just add in a table here, you'll see that now instead of highlighting key differences in a separate paragraph, we're actually able to dynamically format a comparative table here, to actually show comparison between a rebuild and a re refresh in this case, which are two source operations that you can do with Kuvet. In this case now, let's say I want to actually get more specific. And as a developer, I wanna know how to trigger a source a rest source rebuild. You can actually see here now that instead of using this as a comparative table, we're actually able to do markup with the code and show code snippets in, you know, a rich format to make it more, to make it much better, actually, to leverage this functionality for an end user. And you can see here, I have some, follow-up questions that are recommended. So if I click on one here, the whole context is gonna be, of course, maintained. But the context of the documents here that are retrieved at the bottom plus the context of the actual prompt here are gonna be adapted based on the follow-up question that I clicked on. If I were to type a question here, the same thing would apply. Now going back oops. Sorry about that. Going back to my deck for, what's coming up next in terms of generative AI. So now I've shown, you know, rich answer formatting and follow-up questions. We're now working on inline URLs. So being able to dynamically tag pieces of your site map essentially to the generated prompt. In this case, right, we're using the product categories found in the site map, and we're able to dynamically tag these onto the actual summary in the generated prompt to contextualize it and start to bring it closer to the actual commerce catalog. And bringing it closer to the commerce catalog is the key theme here because up next, we're also working on contextualized category recommendations, being able to, you know, bring users to navigation in the actual catalog from a generated prompt built on rich content. So making sure that really the products that we're talking about and the categories we're talking about are actually grounded in truth and can be navigated too easily for, from an end user perspective. And up next, which is a good segue into what we're gonna be talking about next, we're We're gonna be able to leverage our semantic product discovery to be able to retrieve relevant products based on the prompt that are contextualized within the summary, to to be able to go into use cases where we can actually recommend what to buy based on the questions that you're asking and what customers are looking for. As I said, this brings us to a good segue into in into our semantic encoder enhancements that we're bringing to production now. So, you know, we pre we presented this as early access in the fall, and now it's it's more in beta. Right? So we have a few customers testing this as we speak. But in this case, right, we're leveraging an advanced semantic encoder to truly bring hybrid search to Coveo. Right? Whereas we had many semantic functionalities before, this is using, you know, the the the greatest, vector based retrieval techniques alongside our regular lexical retrieval techniques to to actually bring the best of both worlds in the search. Right? Now lexical search is in-depth. Precision is still important. Right? When I type something that's accurate and true, I want to see those results more than things that are similar to what I'm typing. But in the case of queries that may not exactly match terms in the catalog or that mean more may may be more descriptive than precise, you want to actually be able to expand the breadth of results to make sure that you're always catering to the end user. Right? So I figured, you know, why not do another demo on our Barca Sports property here? Barca Sports, so is is a full demo property that's fully built using generative AI. So So all the products are fully AI generated. All the traffic's generated as well, and all the descriptions and everything are generated. So this is great for demo purposes, but, of course, this catalog is not the richest in in terms of content and description. Right? So this is where, you know, semantic encoding becomes important because when users type and search through the site, similarity may become more important or less important than lexical search. If I give you an example here and I search for blue kayak, right, you'll see here that our search returns some great results. And right now, I have semantic encoding on. But if I turn it off, you'll see we get the exact same results. Right? Because here, our regular search engine is already able to detect that blue is a color and that kayak is a product type. So here, we're already scoping users on blue kayaks, and it's working great. And the engine knows not to consider semantics here because semantics would bring too much fuzziness. Right? I turn it back on, you see the number of results maintains the same. But now if I add an objective here such as dark, that's not contained in any product in the catalog, right, this is where the semantic encoder is gonna shine. So you see here if I search for dark blue kayak, I'm now scoped down to the dark blue kayaks found in this property, and I get eleven results. If I now go and turn off semantic encoding, you see that now we find no results. Now do bear in mind here that I turned off other of Caveo's semantic functionalities to replace them with the semantic encoder here. But in the end, right, this does show the power that this feature has and why also it's important to consider lexical match as well as semantic matching in these cases because you don't wanna lose precision at the cost of returning results. Going back to my presentation now, there is one last thing I wanted to present to you today, which is our business aware product ranking advancements. So in commerce, you know, we're finding that all users want to cater to two different audiences simultaneously. They wanna give the best user experience possible, make sure that they're all make sure that vendors are always remaining relevant with their end shoppers, and then they can cater to their actual needs, while at the same time, wanna be optimizing for business outcomes such as revenue, profitability, conversion, etcetera. But now how do you balance these things within a relevant set of results? Right? So if I take this example here, taking again from our bucket sports. Right? This is the category t shirt in the listing page setting. We're already getting some pretty good results. Right? You see here, there is one t shirt which is actually tagged as a t shirt in the catalog, but due to our generated catalog here, Gen AI thought this was a t shirt. So, you know, some of the pitfalls of Gen AI, in action here. If I look at the top results here, which are already relevant, though, I can actually see that they all have different business scores. Right? And what we're doing now with business aware product ranking is that we're actually storing all of this metadata on products and more, and using these computed attributes to be able to rerank and dynamically balance the rest of the Coveo equation. So in this case, you can see here that those two products at the bottom actually have a higher predicted revenue and predictive profit, which are two predictive metrics that we're computing for each and every user. So here, if I go back to my results set by turning on business aware product ranking, I'm now reranking my top results and all the other results in the results set slightly to attempt to gain an optimization boost, in terms of profitability or revenue from my listing pages. Now do note we are bringing this to search page as well. But for now, we're starting with listings, and this is in early access currently, and we're currently testing this with a few customers with great, initial results. So with that being said, that's all I had for you from the AI standpoint today. And I'll be handing it over to my colleague, AB, who can take it away on the merchandises. Thank you, Dolly. Amazing demos as per usual. Alright. So I'm gonna jump right into merchandising here. So I'm Anthony for those of you who don't know me. Oftentimes referred to as AD. And, before jumping in into kind of demos and features, I just wanna touch on merchandising. You know, our our mission here in the commerce team from a product perspective is to democratize AI powered product discovery. We think that is a there's a big gap in the market where that's not really being satisfied. There's a lot of great AI out there that it just isn't democratic. It's not reaching to the the the right people. And so, you know, when we see kind of the the the full picture, we think that AI really does only become democratic once merchandisers and and people with business context to front lines can actually start using it. And so kind of corresponding with this, we're investing heavily in our merchandising hub, and we're getting to a place where basically quarter over quarter, this merchandiser experience with Coveo is materially changing and improving. And so last time, last fall, when we we had our last session, we talked about our listings manager. And what I wanna talk to you today about is our search manager. And so we're gonna kinda talk about kind of this this first kind of core release that we we've made. So we we call it our essential release. And, you know, I I could talk about it a long time, but I could also just jump right into the demo. And so here's our listings manager. We've shown it before. The way the product works is basically we have our listing pages, and we break them down by listing page, and we have our integrated analytics and our integrated control. Well, search works exactly the same way or almost exactly the same way where we break down by queries. So this is more of a responsive system where you're gonna have, you know, tens of thousands, hundreds of thousands of queries, and we can kind of take all the analytics on them and and also allow you to act on them. So here, that first view here, we're we're we're letting you see immediately kind of what my top queries are. This table is sortable. I can play around with it. But I can also kinda go right into a query and and get a lot more detail. And so here on this Kaya title query, I already had a little bit of information, but now I have a lot more. I can see kind of specific metrics that that that help me kinda make decisions about this this query. And I can also go in and see which products are being kind of returned for this query to see if, you know, the the the ranking matches my expectations. And so this is this notion of integrated analytics. We don't want you to have to go to a third party system to go figure out how how are my listing and performing, how is my search performing. We want you to be able to do it right here. The next kind of big aspect that we have is this notion of integrated control, where I can go in and I can, actually change the ranking right here from the merchandising hub. And so here, what I'm gonna do is create a ranking rule, and I'm gonna affect some of my relevance here, based on kind of the insight I might have. So I know seasonally what's coming up is a big push on carbon, kayak paddles. So on this query, I don't wanna boost things with carbon in them. And so say any product whose name contains carbon is gonna get a boost. And right here, what you're gonna see on the right hand side is an immediate, effect in my visual preview, and we see things being pushed up. Now right away, something's triggering here. I'm seeing these carbon blue kayaks also getting pushed up. Right? So in my kind of recall here, these blue kayaks are kind of at the low end of the ranking. Because this boost is so strong, they're being pushed up. Now what I can do here quite easily, though, is exclude them from the rule. So I'm gonna say here anything that's category is not, a paddle is not gonna get pushed up. So here we go. Wait. Sorry. Click the wrong button there. So like I said, category is not a paddle. Now as you see in my visual preview, we're gonna or is sorry. Go back and edit that. This is what these Boolean operators are for. And here we go. That's the rule that I want. So anything that is a paddle and that does contain carbon is getting boosted up. Everything else isn't. So this is that kind of nice visual feedback. You've got your rule. And if you make mistakes like me, guess what? You can catch them right away and not on your production site. So review this when we publish it. We'll call it boost carbon paddles, and off we are to the races. And so that's how easy it is with the Merchandising Hub now to affect your relevance. So this is something that we've seen with our customers. It's just these are day to day activities that, you know, you want to be able to go in on a dime and change your relevance based on new insights, new information that's come from the business. And so this is an area where we're investing heavily to make this extremely tactile, extremely responsive, hard to mess up. And so that's kind of a quick view of the search manager and where we've gotten to today. You can see here that synonym rules and redirect rules are coming soon. So we got an aggressive road map here. What I wanna talk to, a bit more concretely is also what's coming kind of in that over, like, that three to six month horizon. So what we're focusing on now is what we call tools to scale. So there's a lot of operational tasks that that you would do merchandising day to day, for search listings, that we want to kind of bring into the product here, to to ultimately make a lot easier for our users. So first things, easily creating listings based on any product attribute in your catalog. You know, creating listings is a is a pretty common thing. You know, you you might have a promotion coming up. You might have a seasonal kind of, event coming up, and you just wanna create a collection of products that are gonna meet a certain criteria. So if I need to create a a list of products that, you know, are under fifty dollars that are on sale and they're in a set of categories and just easily create a listing there with a predefined URL that I can publish on my CMS, that's gonna become really, really easy here. Today, we can support the API, but we're gonna make it available in the merchandise dot com. Second thing, configuring the facet behavior kind of either globally or on specific queries or or listing pages. So being able to go into the detail and say, I want these facets here. I don't want these facets here. I wanna manage the values. I wanna change the sort orders. That's the kind of stuff that people wanna do. But let's also remember that Coveo AI has a lot of kind of useful value here that can that can really take away a lot of the lift here. So while, you know, in in the market today, there are a lot of ways to manage your your facets that are quite detailed and tactile and and effort consuming. We want people to get to a point is if the AI takes away eighty percent of the work, and using our facet management, you can go into the last twenty percent that, like, really creates that extra value. Third thing, scheduling filter ranking rules, to align with the campaign activity. So this is something that, you know, we we get asked that about a lot. We can support it today with query query pipelines. We wanna make this really easy in the merchandising hub, not having to log in at midnight to take your rules off. And then the last thing, which is a bit more strategic, but it's about this idea of, you know, figuring out what works. All of these changes you can make, it it can sometimes be hard to to answer the question, am I driving value? Is this actually improving the customer experience? AB testing is a great tool to answer that. So we wanna bring AB testing directly into the merchandising hub such that any configuration change you make, you can decide to test and figure out is this a good idea or not. So this is a bit of a snapshot of where we're going with merchandising. Like I said, every quarter, the experience materially improves, and and we're really excited about what's coming next year. Next thing I wanna pop up is data and reporting. And so what I wanna talk about here is this new event tracking protocol. We call it event protocol, and this is a replacement for our our collect protocol. You know, our data tracking protocol is how we would figure out, you know, on a commerce website, who is buying? Are they adding into carts? Are they seeing products? Are they clicking on products. These events are the lifeblood of AI. And all this stuff that Ollie was showing earlier about, you know, being, responsive to a user's predicted, propensity to convert on a product. Well well, that doesn't come out of nowhere. It comes out of clean, reliable data. And so we've invested all the way at the ground level at in our event tracking protocol, with the view that over the long run, having a great tracking protocol will have a lot of compounding, effect on the the the data quality in our systems and therefore the value we can drive. Alright? And so there's a couple kind of big things about event protocol that make it great. First is it's really simplified. There are fewer events to track, and there are smaller payloads. We're gonna kinda touch on that in just a little bit. But kind of high level, fewer events means that we're only tracking what we absolutely cannot capture elsewhere. And so if you hit our APIs to request, a search or a listing or a recommendation, we'll automatically track off of that. You won't have to kind of emit a secondary event off of that. And then using smaller payloads, so rather than transfer passing as a bunch of product metadata, you can just pass us a product ID, and we will enrich off of that. And so these two ideas mean that a lot less can break because you're doing a lot less with your tracking. You're tracking fewer things, and you're putting fewer attributes in those events. Alright? And so that creates kind of reliability robustness over time. It's integrated with our headless framework and atomic library. So, you know, same kind of tooling available to developers to accelerate that adoption. And, you know, beyond that, we'll we'll talk in about this next point. We've actually invested in new validation tools. And so we have this new thing called the Explorer, which allows you to actually validate events in real time. A lot of time as a developer, when you're setting up the data tracking protocol, it can be tough to answer the question, did I do that right? And so what we wanna do is not make you go to our dashboard, wait ten minutes for something to pass through the data pipeline, and figure out if it's working. What we want you to be able to do is know right away. And so we have here kind of this this playground where we can play around with our protocol. As you can see here, my commerce events, so a card action, a product deck, a product view. And so we'll pick a product view here. Nice, simple, light payload. You know, it's a it's a ski that costs ten euros. And so if you look at this new, Kavehto Explorer Chrome extension here, what this is gonna do is it's gonna validate events that are going through in real time. So I send this template event. It is in fact valid. We get the the purple valid. We feel good about that. That's fantastic. If I, remove a key, for example, something that's required and set an event, we're gonna catch that. We're gonna tell you right away we're missing something. Alright? If I reset this and make a more subtle mistake, maybe I send euros, this will also fail and we'll track it, and we'll tell you why. And so what we're trying to do is really support our customers in having great data tracking implementations and make this as easy as possible with tooling. We think that this is, you know, a great way to make sure we're all kinda working together and getting kind of good system good data into the system. So now it's gonna have a bit of a view on on Explore. The last big topic I wanna cover here is, connectivity and integration. So this is something that we continuously invest in, especially with some of our strategic partners. One of those strategic partners is SAP. And so what we've released now is this SAP catalog push integration. And so for our customers or or prospects working with SAP commerce, we created this native connector that makes it very easy to get your catalog from SAP into Coveo. As a result, we're kind of tying into SAP's investments in, modern and composable commerce. Right? And so this kind of flows in through our APIs, and it kind of flows out into the the rest of what we can offer. And it accelerates kind of adoption, and it and it creates that level of stability over time to have good catalog data in Coveo. We've talked about how important it is to have great data in the system. This is another kind of way of achieving that. So quick kind of quick quick view on connectivity. I think the next step here is q and a. So I think, Shereen, you might have kinda collected some questions from the chat. I did. And I wanna apologize, sorry, for not sharing slides at the beginning, but there were only a couple of slides, and I think Olivier, caught up to me. So you did not miss much. One of the slides is right there. So let's start off by there were a number of questions that came in around GenAI. So let's take those first. One of the questions was, what happens if a user votes thumbs down on an answer? If you recall, Ollie, you know, on the Yeah. Response, there's this thumbs up and a thumbs down in terms of the Gen AI answers. Yeah. It's something we're currently exploring, whether to integrate an actual feedback loop, but, we're already offering reporting to merchandisers and to business users based on a thumbs up, thumbs down, allowing them to make informed decisions. And I actually see another question which I can piggyback on, which is, you know, how does it intersect? Right? How does the intersection between merchandising functionality that AD showed and Generative AI functionality that I showed, essentially, rules and keyword search, etcetera, come into play with Generative AI and more fuzzy queries. Right? And that sort of plays into the thumbs up, thumbs down because the whole Coveo solution is ahead of the generated prompt. That means that any merchandising rule you put in place, any, ranking rule you put in place, any model that you put in place will actually impact the snippets that are sent for generation to to, in this case, opening up. So, you know, once we get into a fully integrated solution between the product catalog and enriched content, any action you make on your products in the merchandising hub are gonna be directly translated to the generated prompt. So let's say you find in a report that users are thumbing down a lot of the prompts. You're able to actually go in as a business user, see which items were retrieved for that because for these questions for that prompt, and you're able to add rules and business logic to actually modify the outcome of the prompt, directly. So, I hope that answers the question. Right. So the merchandising is the final mile that will ground the prompt before it actually goes and retrieves the the information. Right. Exactly. Okay. Another one for Gen AI. Can it okay. I think I know what this person means, but can it recommend by creating bundle products based on matches? Sounds like the category steps to show, but I don't know. Yes and no. Right? I think what they mean is a bit more specific. So it's it's definitely the the direction we're going in, and these are things that are able to be done, by, you know, leveraging relevant standard answering and grounding it into more, you know, fine tuning approaches with LLMs and etcetera. This is not something we support today. It's something, however, that some of our customers are adding on to our use case by using our system as a rag and then integrating that additional piece of solution. But what we will support is actually to retrieve products that are related and relevant towards the prompt. So it could be seen as a bundle, but we we wanna be careful with bundle specifically because sometimes bundle need to be very specific. And in the case where you have very specific bundles, let's say you have compatibility and that sort of stuff with products, those can, of course, be retrieved. Right? But to automate it, we're gonna be, a bit more fuzzy at first. That's why we're starting with categories, to make sure that we remain relevant, no matter what the prompt is. Okay. Understood. So bundles typically in a b two b scenario are more relevant in terms of, the use case. Exactly. Okay. Great. Is there a limit to the amount of content we can use for generative AI? So there is. Today, you have a limit for one million content documents. We're actually coming out with an a limit of twenty million within the next month, and we're gonna increase that limit to over a hundred million vectors, by the end of the year. So so really continuously continuously working to to to upscale it. But right now, to be fair, we haven't come across customers at that scale yet, so we're just getting ready for for future needs. Okay. And we get this question a lot. Of course, we got it again. It's how do you make sure that the generative answers are accurate, do not hallucinate? So always the fear of the veracity that comes back in terms of the answer. Yeah. That and that's really what we're offering here with relevance generated answering. Right? Since we put the Coveo solution and search retrieval techniques ahead of the generated prompt, we're we're able to ensure that the prompt is always grounded in truth based on what's retrieved. If no snippets or documents are retrieved, then we don't generate a prompt to ensure we don't hallucinate. And and the prompt is solely grounded on the actual documents that are retrieved. So to the point around merchandising even earlier, right, any business rule you put in place, any entitlement is respected as well. Any filters that are applied are respected as well. Right? So, really, any traditional search functionality you would expect, traditionally from a search vendor are available to actually control the row, the the generated prompt as well. Mhmm. And it's only based on your content. It's not taking content from the Internet, which is another thing? Exactly. Yeah. So it's only using the summarization function and the natural language understanding piece of the generative AI model, not its general knowledge piece, right, which we're we're expressively asking not to not to leverage. And we also work under zero retention, etcetera. Right? So none of the our customer data is retained in any generative AI models. Right. And if there's not enough content to provide a relevant answer, it will come back with that response. Exactly. Yeah. Which we feel is the the right approach to make sure there's no rules in terms of the amount of content needed in order to generate an answer. Exactly. Alright. Switching to merchandising, there's one that came in on the chat channel. Does Coveo merchandising tool allow to configure custom know, you know, any any listing page that's created in the Merchandising Hub is based off of your catalog. And so anything that is indexed. So, you know, if you wanna create a sales page based off of a combination of, you know, category, price, sales status, you know, accessibility to certain kinds of users. Whatever you're indexing, that can flow through to how that listing page works and what is is not included. Same thing works for for for search rules, for example. So if you wanted to filter search rules or filter a response on a given term, you could do the same thing. You know, we have this constant approach of we have this great catalog. We have this great investment in content ingestion. We wanna make sure that that flows through the merchandising hub, not that we're we're hamstringing people with these arbitrary limits on what data they can or cannot use. So if it's indexed in that way, it can be merchandised in that way because it's the survey. Exactly. Great summary. Right. For Coveo relevance generative answering model, is there a page limit as well for single document to consider for answer generation? So that's like if you have, let's say, a document that's, you know, two hundred page long PDF that, you know, is deep into specs maybe. Is there a limit there? So yes and no. There's a limit in terms of the number of chunks from documents that we'll actually consider for the first stage retrieval. So we actually works our RGA actually works under a two stage retrieval system. So at first, we actually rank the top hundred chunks for each document, and we we rank those to actually use for the first retrieval stage to know which documents to look at. But then in the second retrieval from that two hundred page document to send for, you know, generation, but only the top hundred chunks from that document will be considered to retrieve the actual document. So so far, we found that this is a good trade off between, you know, performance to ensure good latency and, and precision, but we're continually exploring, of course, to to improve that. Alright. Perfect. I don't see any other specific questions that came oh, hold on. Are my company synonyms shared with the LLM when a user asks a question? Like, does Coveo supply additional information to the LLM, tells them how to interpret a question given my data? So this is around data security again. I think but I think that's the merchant to to how we merchandise Yeah. Your synonyms. Right? And and to all these early point, you know, our our our way of interpreting generative answers is still gonna leverage our core search, you know, infrastructure. And so if you are configuring synonyms, I'll need to pick up on that. We care about those too. Yeah. So they're actually shared for generation. But once again, I just wanna make sure to to share this. Right? We use we use LMs with zero retention policies. So even though we we do structure the prompt in a way that considers any synonyms and any, know, regular thesaurus or regular rules that any user can put in place. Those are not used by the LLM for training. Right? They're just sent within the prompt that's not that that's not retained by the LLM, essentially, to answer. So I hope that that answers the question. Okay. Yeah. Jared in the moment and deleted after. Exactly. Yeah. Okay. And to follow-up on the merchandising, question, AD, does it support geolocation based relevancy? Or I don't know if it's Or underlying models, to support geolocation based relevancy. So that's already kind of baked in. In addition, I think kind of further out, you know, road map wise, we we'll probably need to do some more manual controls around that where you might be able to affect users based on their geolocation. Today, I think the the the core AI approach is already driving a lot of value, and that's kind of where we're focusing for now. Okay. Perfect. So and to answer some of the other questions that came in that were basically the same, how do you get access to, some of this functionality? So if you're a Coveo customer understand more about the features and what access you can get into them, then reach out to your customer success manager. If If you're not a Coveo customer and you wanna find out more just about Coveo and then you can, head to our website and fill out a request form. And that is, I think, it in terms of the questions that came in. So I just wanted to take a moment to thank both Adi and, Olivier today who gave an amazing presentation, demos, and also answered all of the questions. And thank you everyone else for joining us, and have a good rest of your day. Thank you very much, and thanks for the great host. Everyone. Thank you.