Hello. Good morning, everyone, and also good afternoon for those of, our customer base that are joining us from Europe today. Welcome to New in Coveo for Commerce, and I'm happy to say this is the spring edition. For those of us in Montreal, we are very happy it's the spring edition. We've been looking for forward to this. My name is Sheerine Reid. I'm part of the product marketing team here at Coveo, and I'm going to be your host today. I'm also joined by my colleagues from our amazing product team. Simoneau and Anthony Delage are here with you today, and they're gonna be doing most of the heavy lifting, spending the next forty minutes or so diving into the latest and greatest functionality for our commerce product and also providing a bit of a sneak peek into what is coming next. So before we get started, just a quick reminder that everyone is, of course, on listen only mode due to the large amount of attendees on the webinar today. However, we do wanna hear from you. And if you have any questions, please, you know, feel free, pop them in the chat, and we have some time reserved at the end of the session to answer those. And, also, and always, pervasive question, is this session being recorded? And, yes, it is. So you should receive a copy of the recording within your inbox if you've registered for the webinar about twenty four hours after the event. Alright. So with that being said, let's get started here. There we go. Slides are working. Before I hand it over to Simon, I just wanted to take a quick look at what's going, on out there in the market. And I'm sure most of you are aware that product discovery is really going through a fundamental shift. So for years now, digital commerce has been built around search boxes, filters, and navigation. Right? Users expect to type in some keywords related to products they're looking for and then see a a product grid, a bunch of tiles with relevant ranked products, and hopefully some just as relevant filters on the side so that they can further, refine and filter their results. But that model really assumes, you know, two fundamental things. The first is, of course, that the customers know exactly the products that they're looking for at that particular time. And the second is that the search system is intelligent enough to retrieve and rank those the right products matching what the customer is intending to using those two or three words that they use in the search box with the right items in the catalog. So intelligence systems, you know, using AI for retrieval and ranking like Coveo have really gotten have gotten really, really good at doing that. But what's changed now is the influence of LLMs like ChatGPT. So they've reset expectations and some behaviors in accordance with that. Customers now know it's possible to actually start with questions. They can express them in natural language, like they would be talking to a sales associate associate and get help on the discovery process itself. So, you know, guidance, answers, education type of information, not just a list of of results. So for those that are not clear yet on the products they actually wanna add to their cart, this is really super helpful. And the data we've compiled here, you can see on the slide makes that clear. Right? So nearly seventy percent of shoppers expect conversational experiences to shape how they shop. The majority are ready to engage with AI that can answer questions in real time. And importantly, I think this isn't just a novelty. So forty percent are more likely to purchase when guided by an AI assistant. So this isn't a feature trend. It's it's really a behavioral shift on how discovery happens. This data, by the way, it's it's just to note that it comes from our relevance report, and we're gonna have that published on our website in just a couple of days. However, I I wanna point out something that it's it's still early. Right? We we don't know exactly how far or how fast this will reshape experiences on retailer or brand sites themselves. But the shift in the user behavior is clear, and it's something you really need to start preparing for and actively experimenting with now. I was recently at Shop Talk a couple of weeks ago, and what I heard over and over again from, you know, the sessions and people on the show floor is that there's no wait and see mode at this point. People are really leaning in. Consumers are adapting quickly, rapidly, and you really need to be ready for that. Okay? So what's our point of view at Coveo? So at Coveo, our point of view is simple. We've been talking about the intent box for quite a while now, and we believe that the shift the shift that we see happening, it doesn't replace search. Instead, it elevates it. Right? So conversational experiences only work if they're grounded in the same systems that power your relevance, your ranking, and your product discovery today. What matters isn't just generating that answer. It's generating the answer that are grounded on your catalog. They're contextualized, you know, around real time availability, pricing, and the business rules that you've established. They're aligned with the same relevance and optimization strategies that are driving the rest of your search experience. And that's why we see a conversational product discovery as search native, and it's not separate from it. And that's a real big distinction. We're extending the core of what already works, so our indexing, ranking, merchandising, all of that those features into a new interface that can interpret intent and dynamically assemble the right experience. So if there people are looking for products, they're going to see products right away. If people have a question, then it's able to understand that question and then act as a guided discovery mechanism, dynamically structure a UI experience. So, and it's all powered by the engine driving your search experience today. So with that, I'm gonna pass it over to Simon, who is going to show you some of this, really cool new Agentic, discovery features that we are launching today. Take it away, Simon. Thank you, Sheerine. And as is tradition in this company, we do demos, hence the transfer is important here. So let's dive right into, I would say, our main investment and and really tied to what Sheerine just presented, which is conversational product discovery. It is using an agentic layer, obviously, in order to orchestrate the entire journey. And I will go straight into the demo and come back to that slide to explain a little bit the intricacies of it. So here, I'm, you know, on a a typical b to c shop, surf shop. You know, if I'm searching for for anything like, you know, RashGuard, for example, I will get products. But if I go with the I would say, you know, a more education query. So, you know, I'm going to a y, let's say, next week. Let's add the fifth temporality here and want to get into surfing. What should I look out to or look out for, actually? So let's try here. So what's happening here is I have kind of a multi ended query, mostly. I'm going to, you know, to a y. It's into surfing. I you know, I'm I'm not entirely sure what what I should start with. There's kind of multi intent here. There's probably, you know, lots of different products that could be shown. So you can see here what's happening is the agent is thinking right now. So, obviously, you know, for the matter of demo, we're showing this thinking happening here. But it's it will not necessarily you know, it's not necessarily something you need to show to your to your users, but it shows that the agent itself does more than just, for example, retrieval and and relevance. It it looks at disambiguation, rewrite the queries, extract the content, validate the content. So you can see here, for example, you know, it starts doing me a few beginner surfboard. So I will just, you know, keep that conversation going. So now that I've seen, you know, this first board and, you know, I I have the the wave writer, so let's, you know, take a look at what this board's all about. And let's see here. So you can see what's happening is the agent itself does more than retrieval, as I said. It will also dynamically create a layout that makes the most sense. So what's happening here, it's currently rendering the layout, and you can see what it does is that it looked at the product and generated a dynamic PDP, so a product detail page that was generated. It extracted the data from the catalog as well as did a AI summary here. Now keep in mind that this is, this can be grounded truth such as things that are part of the catalog as well as ungrounded truth, meaning that, you know, if there's any kind of missing pieces as part of your catalog, it will understand and extract it from overall common knowledge. And here, you know, I'm following up just, you know, trying to compare a few boards, and we'll see here what's happening again is that, you know, it look it's looking at recommendation. It's starting to look at products. And, again, here, the dynamic layout taking effect. So what's happening here is we're creating automatically that comparison, layout, and it says, you know, the way the WaveWriter offers the easiest learning curve, blah blah blah. That's the that's the, the AI summary here to help you, make a decision. The important thing here is that I got to that conversational experience through the search box. And, obviously, I had a very long, because I wanted show that that multi intent search. Right? But I had kind of a a longer query, a more like education query, something that definitely would have returned kind of weird results on a normal search even with advanced semantics enabled and whatnot. But, you know, it could be much more subtle than this. So for example, you know, if I'm searching, if I'm searching for a GIF for my mother, for example, it has enough ambiguity that automatically this system will detect, that there's an intent of, a GIF, which, you know, would require additional conversation rather than just retrieving products and have, you know, kind of random products that, yes, could fit a mother gift, but, you know, might not necessarily fit what you're actually looking for. So it will blend that search listing discovery and conversational experience together rather than being, for example, just a, a side chatbot, which forces you as the, you know, as the owner of your digital property or even your visitor to select what kind of experience they're looking for. Here, Coveo makes sure that this selection is done automatically. So it is it includes the orchestrated agentic solution. It's grounded in catalog data, like I said, unified shop ex experience with dynamic layout for different intent. So, again, here, you know, if I'm more into a larger discovery, it will show me, you know, different type of category, will respond to multi intent. While if I'm looking for something very precise, it will look at a more more focused layout, for example, something similar to a product detail page. Comparison also will be one of them. Just in here to show a little bit the type of layouts that you could expect. So, you know, standard product search that would include filters, multi intent product search with multiple categories, conversational refinement, which is kind of your typical, you know, multi turn conversation, product education, so PDP look alike, product comparison, product bundles. And, obviously, we're expanding this to a more b to b more b to b distribution, b to b manufacturing system as well, for example, to help with buyer guide, fitment system, which will automatically detect the the attributes within the catalog and make sure that the step by step is grounded within these attributes so that the fitment can work. So this was for conversational product discovery. It is an early access right now. If you wanna join, whether you're in retail, you're in b two b, you're in manufacturing, it's all open. You can you can join. The core layer of Coveo is still there, meaning that even if you have complex entitlement, specific pricing, and all that, you can join the early access. Another, sorry, another feature that is in early access is a search page optimizer, which is a meta model that will take all of the current ranking models that Coveo offers, so popularity, one to one personalization, attribute attribute boosting, and we'll put them all together to optimize for one specific outcome. Right now, the the the targeted outcome is revenue data. But at the same time, it will make sure that it preserves engagement, conversion, AOV. So it will not, for example, increase revenue while decreasing, for example, click through rate or AOV or whatnot. So this one is also in early access if you ever wanna join. We're also increasing our semantic encoder model to include multimodal embeddings, meaning that if you are searching with the description of an image, you'll be able to. So it is really an augmentation of the semantic layer, meaning that, you know, more traditional approach to multimodal search has been you, you know, that you augment the attributes of a certain item using the image. Meaning that, you know, if you have this first coffee table here and you pass it into a catalog enrichment system, you will have new attributes. So if, for example, the brass the brass material was not part of your dataset, it will be added to an attribute. The vintage, will also be added to an attribute. It sounds good at first because you end up with these, products that now have a lot more attributes on them. The drawback of this, though, in the last few years when we've tested more and more of that multimodal approach, is that it creates a lot of noise. And now if I search for brass, for example, there's a good chance that, you know, lots of different products that include brass will return. So it still kind of fall into that a that idea of lexical matching, but you now have more data to, you know, to power the lexical element of it. The way that we do it is that we augment the semantics later on. It's not something that would necessarily be v v visible by the human eye, but will be understood by the machine when you search for things such as vintage graph table or when you upload image to do comparison. So this is also available in early access, if you want to join. Again, here, it is not an attribute enrichment system. This is something that third party tools will do, greatly. It's it's meant to really increase the semantic layer so that natural language search will return even if you describe a product while at the same time not adding the noise of additional attributes. And one last thing on the on the AI front, we're also opening the MCP server in order for you to host Coveo search into into, off-site agents such as, ChatGPT or Gemini or other. So, mostly, you have a recipe now available in the Coveo documentation if you ever want to make sure that you front the power of Coveo search when these agent reach your own property in at the same time, increasing hit rate and success, which would again, here, there's kind of a blurry line here, but which should increase also your visibility on these different platforms. And on this, moving to merchandising with AD. Thank you, Simon. I'll take over here. Alright. Well, thank you really for your time this morning. It's a pleasure to be talking to you about, merchandising and everything that's doing in the Merchandising Hub. This is obviously a huge area of investment for us. So for us to be able to so you'll you'll see a lot of new things coming, and and you can expect this pace to continue. So, the first thing I wanna talk about is audiences. Audiences are a kind of new tool we've reduced to the merchandising hub that allow you to scope rules to specific users. And I'll give you kind of a canonical example here of how I might use this. And so we're back in our merchandising hub here on our product listing manager, kind of the the first thing we ever launched in in in CMH. And what I'm gonna do here is show you how I can, support a campaign mirroring use case. So it's very common in merchandising. I have a digital marketing strategy. It might have, like, an off-site ad or an email that, is going out to kind of a specific cohort of users. It might be featuring specific products, and I might do something where I want to, on say, my, women's shorts page, I might want to pin a specific product and specific position. But I don't wanna do this for all of my visitors, only for the people landing from that off-site channel. So now I have the ability to use these predefined audiences, based on technology, visitor type. I can create custom audiences that, target URLs or refer URLs. And what I'll do here is use this kinda summer campaign audience I've already created, and I'm gonna use that to create this specific pin. So maybe in this specific position, this pink short is what I featured in my ad. I wanna make sure when people land on the PLP, it's the first thing they see. And so I can review this, publish it, We'll schedule it, obviously, because these things are time bound. Right? Ended at one AM, and we'll call this, like, pin for summer campaign, and off we go. If I go to my site, you'll see here, this is the regular PLP. You can see there's no query parameter here. If I reload the page, nothing's happened. But here, you've seen the query parameter. Right? This is what would have happened had I landed from the campaign externally, and now the pin's gonna apply in first position as you would expect. So we're giving merchandisers more and more tooling here to kind of align their strategies, be more complete, and, again, reflect kind of the the the more holistic view of merchandising and not just purely, on the site. Next thing I wanna cover is spotlight content. So this has been a off requested feature, and it's all about, in your your listings, your your search, being able to inject content, to to kinda help a a user kind of cross sell, get more information. There's so many use cases here where you'd wanna leverage your base of content into, something that aligns with a product listing page or a search result. And so we'll go back to our, our product listing manager. And here, I'll switch over to this sports page. And, in this situation, again, we have kind of all of our our our our same, capabilities. And so if I, for example, go to my sports kayaks page, I might want to create this rule. So here, same concept. I have all users available. I have all that kind of notion of flexible targeting. I can come in here, and I could say in position four, for example, we're gonna put in some actual content. So I might say that this is my image. I'm gonna reuse it for mobile. I'm gonna send users to this link, and I'm gonna give them this heading. So I'm gonna send I'm gonna inject this surfing guide if and when, you know, people are on this page. Super simple, super easy. Right? Name my rule, and off we go. Again, if I go back to my site here, if I go to my sports kayak, you'll see in position four, the surfing guide injected. Super easy. For customers who are using our, who are already kinda using our product listings or search, it's just a headless version bump to be able to kinda start accepting this data. If you're using the commerce API, it's a slight schema shift. So it's very easy to adopt this. We've made we've tried to extend a lot of tooling to make sure that you our customers can can get up and running with this. So that's Spotlight content. Next thing I wanna talk about is retail media. This is a huge topic, obviously, kind of across the industry. Seeing a clear shift in ad spend also as well from amongst our customers away from kinda Google search as you see here, towards more retail media, and it makes sense. Right? Retail media is a great point if you are a supplier, a manufacturer, a a brand. It's a great way to get users at the bottom of the funnel, rather than, you know, high up the funnel with a Google ad. And so if you can kinda leverage retail media, it is a clear and obvious way to generate revenue. And so one of the the challenges that we've seen is, like, well, how does how does product discovery and or how do product discovery and retail media coexist? Right? We want them to we want our customers to be able to benefit from this channel while also maintaining relevance, and a clean product discovery experience. And that's exactly what we've done. And so we've created kind of this this notion of relevance aware sponsored products. I'll show you how that works. So here, again, I can create a new rule here. When we create a rule on this page, when I create a pin rule, I have this new kind of sponsored variation. And what I could say here is, let's say, my position five in the results is going to be, right here defined by the retail media engine. And this is very simple. Right? It's just the merchandiser exposing a slot. And the the mechanics of it is that we expect on where, you know, wherever you're rendering this listing page, the first, we're gonna get the retail media networks answer. Right? So in this position, maybe we wanted to show, this carbon blue kayak, for example. That might be the highest bidder, or some retail media networks are being more sophisticated. It is the most relevant high bidder. Right? Whatever the retail media engine, can spit out, we'll push that in, and we'll make sure that the pagination stays clean, the analytics stay clean, that we still have good signals into our machine learning models. So all of this, is is kind of handled by Coveo. You no longer have these two distinct experiences. And I think when you go out throughout the Internet, you're gonna see a lot of cases where people have kind of taken their first step into retail media. The consideration for how it interacts with product discovery has been more or less made, and you'll see these odd cases where, you know, either retail media network products are really irrelevant or maybe even that they're taking a product from up here and pulling it down so that that the advertiser is actually losing money or position by by doing this. So all of this is stuff that we can extract away, and we can keep these two parties doing their best jobs. We handle relevance. We handle the whole product discovery experience. We understand that retail media networks, you know, abstract away a huge amount of supplier relations and and complexity there. And by communicating what they wanna put in, we play nicely together. So really excited about this. We've seen kind of early returns that are really positive with customers, especially in terms of the cleanness of the implementation. Next thing I wanna talk about is something really light, but it's pretty it's it's still important. It's just our overview page. And so I'll click out of here. So our overview page is a place where we can, just kinda start in the Merchandising Hub. You come in, you're gonna see, kind of all of your kinda key metrics, your attribution, whether it's kind of the the the revenue we're attributing to Coveo solutions or the revenue that goes unattributed without Coveo. Right? You might have customers landing on a product detail page and converting directly because of your digital marketing strategy. That's still relevant revenue. It's important to kinda understand the relative prevalence of each. And we can do this for a variety of metrics here that we think are just a good place to start. All of this is date filterable, comparable over time. This is a page that we expect to grow over time. So today, it's focused on analytics. It kinda caters to where most of our customers are. They're well implemented, and they're they're merchandising actively. But over time, you'll start seeing kind of places to start, or if you've if you're a new customer, we'll help you get up and running. We'll you'll see things like health checks. So if your catalog or event tracking have issues, we'll be able to tell you that. So this is just a start, but we think the foundation has been laid, and it's just a it's a good place to to start your journey and get your mind right. Next thing, again, in the tone in the in the theme of analytics is facet analytics. So, when we launched our facet manager, this year, we wanna make sure that people didn't just have control over facets. They also had an ability to understand their performance. So that's exactly what we did. So facet analytics gives you kind of three layers of information. I can understand generally what is my facet engagement rate per page. So it could be per listing or it could be per search query. I can also understand given, I I can filter down to a specific query. So if I wanna go back down to, kayak paddle I want that kayak paddle beginner. There we go. I can break it down. I can see which facets are getting engagement. In this case, it's kinda boring because the bots are boring. But at scale, I can also look at it. So, again, brand is getting the most of the usage with a little bit of engagement on the other facets. And then I can even filter down to the values that are being, selected. So here in my brand facet, let's see, Adidas, Barca Sports, and AquaBound are the main ones. We've done this because people need to understand what's happening. And, again, I think one of the the the questions we've gotten a lot of is, like, what is good, fast, and engagement? What is what kind of benchmark should I I be striving for? And here, my answer is actually kind of, maybe unsatisfactory, but it's it's all in it depends. Right? If you, have a strategy where you might have listing pages that are very broad, right, you might have a thousand products in the listing, you should expect high fast engagement rate, and they're kind of how you place the facets is gonna be really important. And then the values themselves are gonna tell you about your users. Right? They need to engage. They're telling you something about themselves. Conversely, some of our customers go for a much more restrained approach. Right? They might not even show l one or l two facets. They kinda force you down the tree before you actually see product. If you only see fifty products or maybe twelve products on a listing page, facets are far less useful. And so there, they might be used for kinda slight refinement. But they're less they're they're less relevant. And that's okay. It's not a problem. It's just a question of strategy. So these rates are all relative. Our goal is to give you a clear picture here of what's going on and should I be making changes. Last thing I wanna talk about, which is, probably the last but not least, is the merchandising co pilot. This is a huge area of investment for us. This is something that we think is going to, meaningfully change how people interact with our products. The copilot is effectively a natural language, companion that's gonna help you, interact with the Merchandising Hub in natural language. The Merchandising Hub, or the Merchandising Copilot, operates basically on top of what you already have access to. So all of the analytics in the Merchandising Hub, are available. All of your configurations are available, and all of the actions that you can perform as a user are also available. So this is really interesting. Right? Because it means that you can talk in natural language as if you were, as if you were gonna use the product, and it allows you to kinda go from insight to action. Give you an example here. So you can see here already that I moved over to the listing page. I'll start a new chat, and you could see how my kinda context has changed. So the Copilot is context aware. I'll start from an insight perspective. Let's think high level. How are my listing pages performing? And we'll drill this down all the way down to a plan, and and a way of improving this. So we see this as a as a huge win over time for our customers. First, I think we're gonna see people become much more efficient in kind of how they they use the the products. We've invested heavily in, obviously, in usable flows to do things like use an audience, create a rule, change your facet collections, and that's not going away. However, we do see a case where, power users, and let's see. Let's look at that. We see case where power users today kind of they know the product really well and the the the need to, to click through can be can be a a a constraint more than a help for them at some point. And so what we want to do is give a place where if you really know what you need to do, you can go really fast, and you can kinda perform actions. Think of bulk actions or or more complicated merchandising strategies. We also wanna create a place where kind of the best users can kind of stress test their ideas or or even kinda go from a an incomplete insight to a complete insight, take people from you know, everyone has a feel for their their business and their catalog. I've never met any merchandisers that don't have opinions on what's going on. If they didn't, they probably wouldn't have their job for very long. But there's always kind of nuance on the edge questions that you might ask, and the Coveo can be there to help you. And over time, what we what we can do, is turn this into kind of a a power tool. Right? So think about, you know, autonomous agents that, you know, if you if you're gonna run the same report every morning or at nine AM to see what's happened, Just have an agent run that at eight forty five and tell you what the main insights are. If you do content gap analysis, every every week or every month and you kind of do some small, low leverage things like introduce synonyms or or or or review your catalog in certain ways as a result, you can tell an agent to do that eventually, and it will just tell you what it did. Right? And so all of these are tools where if we can delegate, if we can take, you know, tactical work off of people's hands, we can make the merchandising profession more strategic. We can give people more time to create even more value for the business. So that's what we're really excited about. It's it's it's the potential here to to to change what what people are doing with software. And so let's try this. Let's try this priced tiered merchandising strategy, and we'll actually implement some rules. Now the essential thing about the Copilot back to what it kind of its principles is that when we actually, implement rules, we don't run without permission. And so any kind of action requires user, confirmation. And so this is super essential. Right? It means that you never feel out of control. It means that you're gonna be, it means that you're always in the hotel agent or the copilot. Nope. I don't want that. And so there's never a case where it just kinda runs off. So here's our plan, and we can apply it. And you'll see the confirmation here. Right? So I can deny it or apply it, and that's what's gonna take me end to end. So this is kind of a kind of a quick view of the Copilot. If you wanna get involved, we are onboarding people left and right. So we're really excited about it, and we hope you are too. So that's here we go. We're creating the rules. It kinda self corrects as well. And now I'll hand back over to Simon. I'll keep presenting, but we got a few more topics we wanna share with you guys. Indeed. Thank you, There we go. So, yeah, just a a few, I would say, closing closing points. Content deduplication, a very important feature, not necessarily the one that shines the most, but extremely important since it reduced drastically the number of item you need to maintain in the index for multi multimarket, multilingual setup. So mostly, you have one source that is global, and every single item can have multiple versions by country, by language, as well as translation directly in the item itself. So if we are looking at catalog, let's say, catalog of a million products times fifty languages, this was, you know, an index of fifty million plus items, which complicates management indexing and whatnot. Although possible, it's not necessarily the the easiest to to manage. So with this, we're able to combine everything into one document, one source, and one catalog. It is an early access. If you if you are a kind of customer with a very large catalog, we probably reach out to you already by that point. And as the last feature, we are improving on our Shopify connector. So we already offer the Shopify native app. It was connected to a GraphQL API source. So mostly the app itself was kind of using a source that was created on the other side. We've kind of blended the two together by creating a source that is created automatically from the Shopify app and will be updated automatically also from the Shopify app using Webhooks. So anytime that you have an item, a product that changes in Shopify, it will be updated directly into the native source without you having to actually trigger a rebuild or refresh manually. This is also in early access if you are a Shopify customer. Hey. Oops. Shoot. Sorry. My video is not going on. There I am. I'm back. Questions on. Thank you, Simon. Thank you, AD, for covering all of that lovely material. There are a couple of questions that came in that I would like to cover with you. The first, of course, is around conversational. I guess I'll combine a couple of questions together around getting access to this new feature that we've introduced and around whether their current licensing, would cover it, or is this priced separately? So, Simon, I I guess, pass that one to you. Absolutely. Yes. You can have access right now. Just reach out to your customer success manager or account manager, and we will enable it for you. There is a license change, however, since it's using, you know, a different type of queries. It's using conversational type queries. Obviously, it's it's queries that require a bit more processing, so there is an additional cost. But we can definitely try you know, put you on on some form of trial for free, at least to get started to see how it reacts, how it performs before diving into the licensing part. Excellent. Okay. And I have one for you, Adi. I think you covered part of this, but people are wandering around the guardrails in place for Copilot. So I know that you said that merchandisers need to approve something before it's actually executed, but I think there are a few more guardrails that you might not have mentioned around deleting stuff on your site and stuff there. That is a great question, and thank you for asking. Yep. So we've also sandboxed the the Copilot a bit. So I mentioned actions. It can do about, like, ninety five percent of what a merchandiser can do, but we have kind of removed, its ability to do certain things that we deem to be, more dangerous than helpful. So the Copilot cannot delete things. And it cannot create filter rules. Deletion is a bit self evident. Right? We don't want the Copilot going off and deleting something accidentally. Even if you approve, that's a like I said, that's an annoying process to have to undo. And we'd rather people kind of delete manually. We're open to revising that in the future, but for now, that, creates a layer of safety. For filter rules, this is a bit subtler, but it's worth understanding. Filter rules can be a bit pernicious. If, for example, a customer has two price fields, like price and price two. And for some reason, price two doesn't really have a good fill rate, but still in your catalog. We wouldn't want someone accidentally accepting some you know, creating some rule that says, hey. A product has to be below price two, and suddenly you've just nuked your search because price two doesn't exist anywhere. Right? So these there's certain kinds of rules that, even though they could look good in the Copilot, you could can you could potentially create, some errors. So, we're we're open to revising these in the future. We've got very kind of, detailed telemetry on how customers are using the Copilot. If this becomes a point of friction, we might look for other solutions. But for now, we've kinda sandboxed in such a way that people feel safe picking it up. Okay. And a little bit back to the conversational product discovery one. Asking about how how it would handle questions that go, I guess, beyond the product catalog, like, information that might not be in your product catalog? That's a really good question, and that's why I said it's grounded truth and ungrounded truth. Right? So mostly, you know, the LLM itself has a knowledge of the world, which we've grown accustomed to, Truth, ChatGPT, other. So it will answer with ungrounded truth. For example, if there are things that are not part of the catalog, you pretty much decide what is the balance there. And the I would say the more complex your environment is so for example, you know, if you're into b two b health care distribution or whatnot, we want to reduce as much as possible the ungrounded truth and really fall ungrounded. But if you're into, I would say, retail, more traditional retail where there's lots of information and context out there outside of the catalog, then we will use ungrounded. So it it even goes through trends and whatnot. So one thing that I didn't necessarily do in the demo, but you can ask questions such as, you know, what kind of rash guard will Taylor Swift be interested in. Right? And it will understand that there is a certain concept here. You know, there's a there's a gender. There is a type here with I think it was something like elegant and and sporty or whatnot. I don't necessarily remember, but it will understand, you know, at least these concepts from outside of the catalog. K. Definitely not information you would include in the catalog for from my experience. Alright. And then one last one for you, AD, around, you know, what you're seeing. I guess, you know, we talked a little bit about the trends that we're seeing in the market around product discovery, but we didn't touch a lot on the the AI adoption across merchandisers and business users in general. So maybe if you could give a little bit of color around that. Yeah. I think it's a it's a good question. I've, so, you know, the the copilot is something that's been kinda suddenly in the works for for for you know, especially just kind of wrapping our head around how and what we would do here. We're kind of over a year now, and we've kind of been going kind of faster in the last quarter or two. And so for a long time, I've been asking users these questions. If you go back a year to even, like, nine months ago, I would ask our customers, hey. Like, what kind of LMs do you use? What what is your kind of experience? And so many people were like, yeah. Like, I've u I've obviously used ChatGPT in my personal life. I used to plan my daughter's birthday party. It was awesome. And at work, I use it to rewrite emails. Right? And it was and and then the penetration in the enterprise, I think, was blocked up by, I think, a bit of fear with regard from from, you know, central teams as to, you know, where we're handing a shotgun to a monkey. But then also, just, kind of maybe the mall is not being there yet. Over the last quarter, I've seen that completely change. I think all of our customers have some, at some point, some kind of official enterprise licensing. Usually, it goes beyond, like, just, we have Office three sixty five, now we have Copilot. So there's a Microsoft Copilot. So we've seen, like, a lot of Anthropic. I've seen stuff, a lot of OpenAI, obviously, some Copilot enterprise, even kind of, other players that kind of federate across, different kind of model types. So really kinda interesting adoption. What I've seen as well is that, a lot of it is is still kind of generic tooling that you have to assemble. Right? And so it'll be good at, like, communications. It'll be good at creating a slide deck, maybe even jumping into Excel. Right? These are all kind of the the tools of the trade for white collar workers. We all know how to use Excel to some degree, and so this is kind of the the standard. What I haven't yet seen is really penetration into, like, detailed tools, and this is what the Copilot is is solving for merchandising. We have, obviously, an expertise in this area. We know how our product works. We can build agents that are extremely helpful to our users, and we can build a system that helps all of our customers really kind of go faster as a result. So it's still an area they think we're we're ahead of the curve going. We're really excited about that. But seeing so much AI adoption just makes us so much more comfortable. People jump right in, and they know how to prompt. They have a good feeling for what's gonna work and not work, how to correct the agent. That's That's really exciting to see that converge. Okay. And nobody asked the question, but I'm gonna clarify that for the Copilot, there actually is no licensing implications. Correct? Something at the end. I just wanna make sure that was the yeah. This Yeah. From the conversational. So we're including this as part of the merchandising hub. We think that, Agentic merchandising is the future. This is part and parcel of working with Coveo. Yeah. And and it opens the door for experimentation. They can get in there, try some things out, see how it works, and and see how it applies to their specific business context in their organization. Right? Absolutely. Alright. And with that, thank you for joining us today. Just two minutes past the forty minute mark, but, hopefully, you received a whole bunch of new ideas and features that you're looking to try out in in Coveo for Commerce, and we'll see you again next time in New in Coveo. Thank you, everyone. Have a good rest of your day. Thank you. Bye. Thank you so much.