Okay. So let's go ahead and begin. Hi, everyone. Thanks so much for joining. My name is Carrie Ann Beach. I'm a senior product marketing manager at Coveo. I am really excited to be joined today by my colleague, solution engineer, Paul Sheridan. Hi, everyone. Hi, Carrie. I hope everyone's doing well today. Hi. Well, I'm definitely doing great. Super excited to get into today's webinar. So today we're gonna be sharing the five best practices for AI site search and how it can help drive an end to end digital customer experience. Of course, this is a demo webinar. So we will be diving deep into the platform to showcase each best practice in action, including GenAI, which I'm sure everyone's really excited to see in practice. So, of course, feel free to ask questions in the chat. We will have a q and a at the end of the presentation. If we don't have time for any reason to address it, No worries because we will follow-up with you afterwards. So let's dive right in, shall we? So site search is a term widely used in websites, but it's really not always understood in terms of its impact on the overall customer experience and your team's ability to convert visitors. Yes. Of course, site search helps customers find and discover your content, but it's really so much more than that. It's the foundation of your website and really your digital transformation strategy, and all of the other features are really built off of this first experience, including cutting edge technology like generative answering. So it's really important that you nail this site search experience. Second, it's also a window directly into the customer. They're signaling to you what it is that they're looking for, and this represents an opportunity for you to capture these high intent inbound leads by delivering what they want when they want it. It's really a one to one conversation at scale. And these are crucial insights for you to use to optimize your site. And then finally, it's the glue that connects your sites to drive an end to end digital customer experience. Especially if you're able to surface data that lives outside of your DXP. Right? Maybe it's in your CRM, your SharePoint, even your intranet, These are all pieces of content that you can unlock. And, of course, there are stats to prove it. We know forty three percent of visitors will go immediately to the search box. So they're signaling to you their intent, and these searches are two to three times more likely to convert So you really need to be ready to capture this opportunity. We also know that seventy three percent of users will go elsewhere after just three unsatisfactory search experiences So it really is kind of a make or break experience. It could be the difference between a conversion or a slight abandonment. But even with all of this, ninety nine percent of companies still struggle to deliver relevant search results. And we believe that part of this could be due to the lack of understanding around the opportunity that is in front of you though. Right? It's not just discovering content. It's the quality and the relevance of this content. Findability while crucial is just not enough We know that visitors expect relevant and personalized experiences that are tailored to them. Now this might sound like a lengthy and quite costly type of task unless, of course, you have AI and machine learning models that are working for you that are learning off every interaction to understand user intent and surface the right content at the right time to really create those personalized interactions at scale. And ultimately, we actually want you to go beyond site search. Of course, it's a great starting point. We often say start with search. And from here, you can expand into other features to really drive that end to end experience, one that is cohesive and connected across all of your sites And this port is really important because your web strategy is not just about your dot com. It's really every site your customer is interacting with from dot com to service, and even community portals. So how does Coveo help you do this you might be asking? While we have a composable AI search and generative experience platform that leverages the power of AI to create dynamic and intelligent sites that convert. And this is done through our eleven machine learning models that enhance content discovery and personalization throughout the entire customer journey. So you can start by amplifying your site search. Maybe you're already using one to two models like query suggestions or content recommendations And then from here, you can unlock new use cases and with features like GenAI, which, of course, we're gonna be getting into shortly in the demo. Now a really unique thing about Coveo is that we're able to bring in content that lives in different systems. Of course, your CMS, whether that's Adobe, your site car, we are an agnostic solution. You can also surface information across other sources like your CRM, Internet, compreluence any system that's really powering your organization. And all of this is going to allow you to really drive that end to end digital custom experience for your users and personalize this experience depending on their real time usage and historic information, but also where they are in their funnel with you so that you can support them from prospect to customer stage. So what we see here on this slide are different Girl sites that you may have that can benefit from machine learning models. So, of course, like we mentioned, you have your your dot com, your support sites, your partner portals, But you can also use it for your internal sites, your intranet, your employee portals, even slack, because your employees are your customers too. And, ultimately, our goal is to really help you capture and convert these high intent inbound leads. So you can focus on your outbound campaigns and really drive a holistic marketing strategy that is not leaving any money on the table. So with that, hopefully you're now really excited about the power of site search and beyond. So let's jump right into our best practices. Starting with the best practice number one. And that is to design and experience driven UI So I'm about to hand it off to Paul, show you the demo of just some things that you wanna consider before I do is, the ease of navigation, does it align with user expectations, How quickly can viewers find what they're looking for, but also how quickly can you guide them to what you want to be showing them? So with that, Paul, can you show us some tips and tricks in the platform. I'll certainly do my best. Thanks, Carrie. Great introduction. Just as a sort of setting the stage, section to our or a demo here, what we'll be using for, for this session is really a set of websites that have been, that are that we use for demos. They're centered around a fictitious company called Barca, who are in the boating industry. They make sailboats, motor boats. They make accessories for, for these boats, GPS systems, and, and many more things like that as well. And, like, in almost any organization, they have multiple different user experiences that they put forward on, on the web or via applications. So they have their corporate or dot com site. There's a support site. There's a commerce site. And, we'll we'll dig into all of that with you here today to try and illustrate a few different, aspects to, good site search practices. And, yeah, let's just get started. As Carrie was mentioning, just starting from the very beginning, the placement of the search interface, the search experience. We we feel, you know, from a design point of view, this is pretty key to to success for your users. On this Barca demonstration site. We've got a nice big friendly search box, really giving the the user the opportunity to ask those ask those questions. Karianne was talking about this being the voice of the customer. So give them plenty of space and encourage them to make use of, of that to ask their questions to to express what they're looking for, whether it's a, you know, single word, multi word query or a a longer question here as well. Couple of things other things to point out about, you know, the search box itself is an ability to, start to suggest queries as you type. Everyone respects this. We're all well trained by a variety of, well, let's say, individual or commercial or oriented sites whether you're thinking of Google dot com or Netflix or any site that you use on a personal basis, and people expect this, in their interactions with organizations as well. So being able to type ahead, being able to, you know, misspell things like GPS, being able to get good quality suggestions as you type is super important. The way that Gavell, does this with, on search on sites that are powered by Gavell, is really leveraging a machine learning model that we call query suggestions where we're learning from user behavior on an ongoing basis, essentially to suggest good quality queries, queries that have worked for other people like you, queers that have been successful, essentially searches that have led to results, led to results that people have clicked on, and so on. So this helps you to scale really well. You you do not have to create, a predefined list of what's a good quality query. Instead, can learn from user behavior, which may be somewhat unpredictable. You never know what, folks are gonna look for on your site, or you don't know all of the possible queries people are gonna So, being able to leverage machine learning in this way, is super powerful and enables you to scale as Carrie Ann was saying. A couple of the things to think about here, and I'm gonna go off and execute one of these queries here. How do I use a GPS? Other aspects of the design of a, of a website, or site search tend to be filtered around. Well, how can I, now narrow down the scope of what I'm looking for, perhaps? Based on different kinds of filters or facets. These are pretty common, especially on commerce websites, but also on other kinds sites as well. You can see over here on the left hand side, these filters that are being displayed to help me as a user. I really narrow down and find the kinds of articles and information that I'm looking whether it's based on metadata tags that you've created on your content, whether it's potentially based on who the author is, of a given piece of content. In some cases, that's important if you're trying to push your organization as, as being an expert in a particular field or having certain ex, expertise in this field, But of course, these filters and facets need to be dynamic, depending on what I'm searching for, different information, different metadata, if you will. It's gonna be more use are important if, you know, if I'm looking for informational articles about how to use a geographical positioning system, sure. These tags are are useful. If on the other hand, I'm looking for, let's say, an event, Barca, as an organization has a number of events there. Different information is useful to me, such as, as you well, you can imagine, when does the event occur, what country or what location is it in? So those filters and facets need to be dynamic, responsive to really the results that are coming back from a search. And that's something that, again, you know, Aveo provides as, a machine learning model here. We call it dynamic navigation experience, but it's really a way to, enable automatically the ability to show up the right kinds of information to to a user at the right time depending on what they're looking for. Another aspect that, I find really important, and, you know, we're this is still only good practice or best practice number one here, but we're really really looking at the design of a search experience, holistically is, well, how how can how do I display these, from called result templates? How do I display these results? To contrast a, what I would think of as a bad practice rather than a best practice. What we often see on on people's websites is it really is just to show title of a document or an item, and maybe a little bit of a summary and a link to it. You really need to be able to take advantage of all that great content that creating on your website images that are associated to, to events or articles or products, useful information here. Again, you know, thinking of an event, such as that the date it's occurring, the time, the location, any kinds of tags that you can provide. This sort of stuff really helps a user to understand what is it I'm going to click on? Where is I'm gonna interact with? And is this worth my time to, to, to go and, and click on this item? Similarly, being able to sort appropriately, you know, sorting my results based on date makes a lot of sense. So I'm looking at news articles or events and so on, but doesn't necessarily make sense if I'm looking at a list of people. So all of this to say really pick and choose the, the tools, the attributes, the, the things that you play to somebody in your search result based on what you understand of the user's needs and take advantage of machine learning models to start to you know, automatically promote, useful filters, useful facets, take advantage of the information that you're creating to, to really play in a friendly, useful way, to a user here as well. I would say also it's useful to test and improve as needed. You know, learning from user behavior, looking at analytics, which we'll get into in a little bit more detail later, understanding what people are searching for, what filters they're interacting with, what clicking on. Super important to, to improve on an ongoing basis. Perfect. Thank you, Paul. It's great. And this actually leads really nicely into our best practice number two. Which is personalizing experiences with relevant results. Of course, this becomes more difficult, the more content that you have, especially if it's siloed across different systems. So some things that you'd want to consider here is are your results dynamic. Do they vary depending on the context and the history of the user, you know, kinda mentioning about changing the experience, depending on who's searching for it, where they've been. Can you go a step further and proactively recommend content before it's searched? And, of course, what is the experience like across sites? Can you follow the users through their journey? And on their end, are they expect are they experiencing one consistent brand no matter which site that they're on, they know that it's yours. So Paul, can you show us how we help our customers with relevancy? Absolutely. Happy to do so. Relevance is a a word that we use almost every twenty minutes or so in our in our daily lives. Really, it's at the core of what Coveo provides as a, you know, relevance as a service platform is often how we describe ourselves. So the, the, you know, we've looked at search results and such the order in which those results are presented is, of course, super important. Nobody, as as we sometimes like to say, goes to page two, of, of search results. So it's really important that you try your best to provide that the best content you can at the top of the list based on what a user searches for. There's a variety of ways that, that that we do that, see, you know, we're sorting by relevance here in, in this, example of a search result. Now, of course, you know, Coveo is a search technology at our core. We've been in the business for nineteen, twenty years or so at this point. There's a lot of different factors that can come into play, to influence the order of search results. What you might think of as traditional search techniques. Yes. I'm looking for the words the user types into the, their search box and promote in content that contains those words in the, in the idle, let's say, that's just sort of automatic core search technology, searching across the full text of articles plus their metadata and so on. In addition to that, many marketing organizations in particular and sometimes, commerce organizations, they really wanna boost or promote particular kinds of content featured results. So we've got a little tag just to be transparent here about why is this result at the top of the list. Somebody in the marketing organization has said, Hey, for this this month, perhaps this is an article that we're trying to promote. So let's pin that to the top of the list if anybody searches for a query that contains GPS. And that's certainly useful and good and and many organizations need to do that. But at scale, that can become challenging. Do, you know, do you know every single possible query that users are going to, are going to search for? Do you know what they actually want to interact with Well, possibly, but, it seems unlikely at scale. So this is an opportunity to take advantage of machine learning models that learn from user behavior. So, learning from what people search for, what they click on, what they do next, Coveo provides machine learning models that we call, automatic relevance tuning to start to promote, popular content, content that's clicked on frequently by a certain type of user and given a certain type of query. And we can see here some examples of that a little further down. Their flag is recommended. So these are being recommended to you by a machine learning model. Now, again, you have the option to deploy, you know, the right tools at the right time, for the right job. It's a a machine learning model that you can use, for certain kinds of use and it can be super useful and helps your organization to scale, understand, or to not have to know in advance what people are going to be searching for on your site. Again, looking at analytics and reports can help you understand what is going on on your site and what your machine learning models are in fact recommending that can be a really useful tool to your organization as well. You'll see also if I go in and do a slightly different query around Marie navigation, Here, we're, once again, re getting recommendations, different recommendations somewhat. There's a seminar people have tut for machine pardon me, marine navigation have have signed up for this seminar frequently. So let's start to recommend that and learn that automatically rather than having to predefined what, what your, what your top result necessarily would be. Another really interesting, approach to making recommendations is when you think about frequent last questions, now almost every site, almost every organization tends to create FAQ pages. Now these are often list of questions and answers that that you know or perhaps anticipate are going to be asked about the services you provide, the products you have, and so on. And they're super useful. Absolutely. But being able to, you know, let's say I go in and I do a search for something along the lines of, you know, I've made an order for a product from, from Berca now. And I'd like to to know, you know, what is the status of my order? What is my order status? I'm gonna of a query here. Now, typically, when somebody does this sort of search on an average website, you'll get a link to your FAQ page, which is useful and good still have to go and click on that page, find within the page what the order status FAQ is if there is one and so on. What you can provide with Coveo now is something that we call smart snippets. You can see I've got my little smart snippets, slider on here. What this enables you to do is to, leverage you know, another machine learning model, which in this case is learning more from your content, your your verbatim content of your FAQ page or or other pages to be able to start to surface an answer to a question as opposed to just a link to the FAQ page itself. So this, this, this model leverages a large language model, but not in a generative sense. We'll get into that in just a second. But, to really, perform a semantic search based on what I've, you know, what my query is. It's not an exact match by any means with what the title or the question is within this page. But it's leveraging kind of advanced somatic search capabilities here to start to say, well, hey, this section of this document. In fact, is pretty close to what you're looking for. So let's start to surface that answer reducing the friction, reducing the time it takes for a user to find, and the answer to their questions. This is a pretty powerful, capability that we've had in our for about three years now. It's not, again, it's not this is not a generated answer. This is verbatim from, by content that's on your page. It is an interesting step towards using generated answers within, within your application within your site as well, and we're gonna get onto that in a second. A bit of a splitter there for you. Yes. Thank you, Paul. Actually, you know, you can you can kind of think of it, the next iteration of this question answering that you're talking about with smart snippets as really being generative answering. So, of course, best practice number three has to do about, with doing more with less with GenAI. So we're really excited to show our Coveo relevance, generative answering. You likely have your own mandates around JNAI or at least are starting to into it. It can be really great to help optimize with your content output, as well as help your customers self serve and really seem favorable at a time and efficiency for your teams. Of course, some things to consider here is privacy and security of your information. That's really important. But also what type of outcomes are you looking for? Cause this is going to influence your implementation? Are you do you have a way to avoid hallucinations? And are you able to in content across different systems, for those generated answers. This is really important, especially because we know demand or content has really been exploding. You're producing a lot of really great content, and you wanna make sure that it's being optimized and it's easily findable and searchable. So, Paul, can you can you show us what this looks like? Absolutely. So, of course, you know, in the last year or so, the topic of generated, generative answering has been of great interest to, to everybody out there. Kavell implemented and released now into general availability, really an, an integrated, an add on to our core relevance platform such that we can allow you to leverage your content, your selected content, your best content to be used to generate an answer where appropriate. I've gone in here and, I've done an example of a query on Barca's support site. And this is an example here of a search for really, you know, a comparison kind query here, a longer form query, one that is more natural language in, in, in nature. And in this case, Barka actually have a couple of different articles on their, on their support site that talk about sonar, talk about single side sonar and multi side sonar and I'm not a voter, so I don't actually know too much about that. But you can notice that we can actually generate from several different articles, a response that actually, you know, answers the user's question in a sense. It's phrased in such a way that it, is a comparison kind of question. Peeling back the, the onion a little bit and talking about how this all works. What Coveo is doing with, with large language models and generative answering is really to ensure that any answers that our generator are coming just from your content. So we're, we're essentially leveraging our core relevance capabilities to return the most relevant documents to a user's query and the most relevant chunks or snippets or passages from those documents We then pass those chunks of documents and the user's prompt to an isolated large language model in order to just try and generate an answer. Now, you know, Carrie talked about a few of the challenges, around generative answering. One of them certainly is hallucinations. The the possibility of generating a wrong answer. The way that we're approaching generative answering is to, as I mentioned, ground any answers that are being provided, strictly on your content, in this case, of course, on BARka's content. And probably, actually, I would suggest curating that content, and, to say, let's only produce answers from really that up to date content, that content that's been validated that you know is up to date accurate source of information so that we're not going to be, generating potentially wrong answers. It's a major concern. And, of course, security being another aspect here as well. You don't want to be leveraging, a large language model that let's say is is completely public. You don't wanna be sharing potentially, secure information from your, your own organization with an external language model, and that's why we use one that's completely isolated that's, that's not saving any information that you're sending to it. It's just processing, trying to generate an answer here as well. A couple of other things to consider as well when you're looking at the possibility of using generative answering, and it's a very powerful tool, perhaps even more advanced, I guess, in a sense, than that smart snippets example we showed, because it's a longer answer. It's able to answer different kinds of if I look for example at something like, more of a how to question, like, how do I download updates from my GPS system that happens to be called Skipper? We can produce like a step by step process from, again, multiple, sources of information and we provide citations for these here at the bottom. So, you know, when I get an answer like this, it's something that I, tend to potentially trust a little bit more than just a list of links down here. It's It's something that, you know, a user can look at and go, okay, yeah, this is step by step. So it is really important to make sure that any of these kinds of generated answers are coming from again, your most trusted, your most secure, your, your, a knowledge base, let's say. So we feel that this is, an approach that is really going to help a lot of our customers is already helping a number of our customers to, help to produce good answers, help to reduce the friction for people to get the answers to their questions and potentially reduce, loads on their, on BARka's customer support organization. They don't mess someone doesn't necessarily have to call in to ask this question. They're able to get this answer really quite quickly and easily here. So that's just a brief introduction to what Cavet is doing with generative answering. Obviously, we'd be happy to, share more, in a a longer session with you. Thank you, Paul. That's really cool. Really excited about that one, which kinda leads us to our best practice number four because as Paul mentioned briefly, the the quality of your content is going to impact the performance of GenI and really all of your search, but it's especially important in this scenario. That's why, our fourth best practice is to create a content management strategy. Companies are spending so much time and money in creating content as we said. So you need to make sure, that it's pulling on the right content, and it's it's up to date. So some of the things you wanna consider here is what I like to call content hygiene. Like, do you have a way of ensuring that your content is relevant and up to date? This is especially important if you're using composable architecture. How are you able to ensure the quality of this data, especially if it's scattered across systems? So, Paul, can you share a little bit on how we're helping with this content management strategy? Sure. Gladly. Now, typically, Kavell is, of course, the, you know, not where you store your content. We're not a content management engine. We're more of a, a a layer of, of search and relevance on top of multiple content management platforms, whether these be, you know, your, your website necessarily, you know, Adobe experienced manager, sitecore optimizely, composable or headless, CMS's, your CRM systems, your, your commerce sites and so on. And we have connectors that enable you to index content, from all of these different sources and potentially surface them in one or more search interfaces. But like Carrie Ann was saying, it's really important to, to make sure that you're indexing that content in an appropriate way. So whether that's a matter of pulling, good metadata from these, from these sources of information through connectors or scraping the content that you index from websites to ensure you're really just getting the, the body of a document and not navigation elements or other aspects that are typically not as relevant to search per se. There's a lot of good practices that we would suggest and it's a bit of a, a bit of a favorite of mine of being able to, ensure that you, that you do that and that you map consistently across these multiple sources, any of those metadata fields that you're dealing with. You have this, you're having this opportunity here to create a knowledge layer on top of multiple different sources of information to kind of unify that, provide a consistent use case, a consistent, set of filters, a consistent set of, result templates, in your user search experience. Also again, you know, curating the content that is being actually delivered to your customers. You wanna make sure that these are valid articles up to date and so on. Are cases where you need to be able to search everything, but there are cases where you just want to recommend, let's say, community posts that have been, you know, upvoted, let's say, or knowledge articles that are, that are vetted. So that that concept of, of, content hygiene as carry on was describing a super key to what, what we provide here. That's it. Let's, we're we're getting towards the end of our session. Let's switch back to, the last of our good practices for the day. Thanks, Paul. Yes. So, of course, best practice number five. Finally, you wanna be able to learn from customer insights to enhance your offering. Paul mentioned this at the beginning. This is super important. It's not really set and forget. You wanna be continuously learning and improving. So some things that you wanna think about is if you're able to visualize this data on customer usage and identify trends. Are you leveraging machine learning models that are trained on every interaction to further personalize the experiences? You have an easy way to identify content gaps. So when someone is searching for something and there's no, responses that are coming up, that identifies a gap that people want that you are now able to go in and and add back into your experience. So Paul, can you show us a little bit about the reports and metrics available at Coveo? Sure. Why not? Let's do it. So, just as we wrap up here today, Kavail by default is logging a lot of different kinds of information but user interactions with your site, certainly what people can query for, what they click on, also potential other events depending on the kinds of use cases you're dealing with on commerce website, does somebody add up, add an item to a cart? In a service website, do they, create a support case after having clicked on a document on kind of a that would be kind of a negative indication. And you can, based on a set of templates we provide, create and configure quite a wide variety of kinds of reports to understand what's happening on your site. As Carrie Ann was mentioned, is a bit of a voice of the customer. Default kinds of reports you might look at, it would include things like, of course, what are the most common user queries, the top click documents, how these interact, you might start to also look into the pardon me, the context of these users, where are they coming from perhaps if they're authenticated what else do you know about them earlier customer, for example. Excuse me. But also, of course, as Carrie was mentioning an ability here to look at what's not working well, or I've got potential gaps in my content people are searching for keywords and phrases, I'm not perhaps not getting any results. Does this indicate that we need to create new content? Or perhaps, create synonyms in a Coveo thesaurus, perhaps to, really point people towards the content you do have that, that can answer their question. And also, getting in a little deeper into the, analytic side of things. What are, what kinds of, average click rank and click through a percentage, are you getting given particular queries? This helps you to understand how well is, how well a search working on your site from the user's point of view, and that's super important. And I I don't know that I've emphasized that enough. It's really important to look at how things look from your user's point of view. Experiment with the, the kinds of queries and suggestions of machine learning models that that provided, but always look at this from the use case of the user. That's really who's important on your site all through all that, that customer journey that, that Anne was talking about earlier. People who are new to your site, customers, employees, and more. With that, I think that's what we have in terms of a good practices, for you today. We'd love to, to share more, ideas and learn more about your challenges as well. But to Gary and I'll head back over to you. Awesome. Thank you so much, Paul. That was a great demo. Thank you to everyone for joining just to kind of bring it all home as marketers, you know, optimizing site search will really help you convert your high intent inbound leads and drive that end to end digital for journey that we've been talking about so you can provide personalized experiences to all of your users, wherever they are in their journey with you. Just to, leave you with a few resources before we get into some questions. We definitely invite you to sign up for relevance three sixty, which is our yearly event to hear from industry leaders. You can also download, our ebook to learn more about Jenny I and our ten AI models. Of course, if you want to go a little bit more in-depth, you can definitely, click the link that was just shared to request a demo. So let's see. I know we are a little bit Coveo time. So for those of you who are you know, for me. Thank you for staying, and we'll get into some questions. Okay. So we have a question here about, okay, how long it takes to get started? That's a great question. A bit open ended perhaps, but I'll do my best to to give a quick summary. Generally speaking, when I hear that kind of question, people are asking about, in particular, AI and machine learning models. But I'll back up just a little bit and say, how long does it get take to get started with Coveo? That's really quite straightforward. Although, it does depend on the kinds of sources of information you're dealing with, you know, which connectors do you want to to use to index content and so on. Tends to be pretty straightforward, honestly to set that up to build a search interface as well. We provide you with a couple of different options, either a very quick, and easy builder to, to drag and drop and create, a search interface there. And, of course, more developer oriented, APIs and frameworks to let you build, a really rich interface like you saw today and a few different examples. That fits into your CMS into your CRM system or wherever you need to put it. So that's sort of, you know, core Coveo, getting started. From a machine learning and AI point of view, there's really two different kinds of models that we deal with. One, is learning from user behavior, as I described, so query suggestions and automatic relevance tuning. In those cases, we're learning from user behavior on your site. So it does depend, on how many, let's say, queries and clicks your website gets. So for a large, busy, website, those models will start to produce good results typically, you know, in days or hours. We're talking hundreds of events, small numbers of hundreds, let's say. So depending on how busy your site is, you'll start to get some good results and good suggestions and such within hours or or a small number of days. If on the other hand, you know, your site doesn't get nearly as much traffic as that, keep in mind that we're still, leveraging additional search techniques and, any rules that you create to really provide a good experience to the users even before machine learning starts to kick in. On the generative side, when we're talking about generative question answering, an interesting thing to keep in mind is that's not learning from user behavior. It's learning from your content. So it will start providing the potential for for good answers immediately. As soon as you've, created a machine learning model and pointed it at your your suggested content or your qualified content. Perfect. Thanks, Paul. There's also a question around, or if you need to have a lot of data for AI to work. That's an interesting question too. Again, you know, we talk about AI in two different ways. There's some ML models that learn from user behavior. Those will get broader and better. I I would suggest with more interactions with your site. On the generative AI side, I would say in fact, you can get really good results. I've I've been working with a couple of POC customers who are able to produce some really good results with a relatively small number of documents, say, you know, couple of thousand documents or even even fewer than that. Now clearly, we're not gonna, we're not gonna be able to generate an answer to every possible question. If you've only got a couple of hundred documents that contain knowledge and information and so on. But if the questions that people are answering do have answers, let's say, in those documents, you can get really good results, with a relatively small number of documents, I would say. And again, I would emphasize that focused and qualified information is probably more important than a really large amount of information, but there is a balance to be struck there. It'd be it'd be really interesting to to talk with you more. I did see one more question here, Carrie Ann, from an attendee I've got up on the screen here about, is there training for the different modules, reporting and analytic so on once we've been implemented. There's there's lots of training. Yes. We provide, on our website, this is, a training site called LinkedIn. And there's training on absolutely everything that Kavail provides. So we'd be happy to to share that with you and to discuss further. Awesome. And then, actually, we just got another question. Will GenAI help for internal site search results? That is Absolutely it will. Paul, you wanted to get that. Sitting there nodding. Yeah. It it certainly can. And in fact, that's a really interesting use case that we're seeing quite a lot of. So being able to generate an answer in a, you know, typical use cases that come to my mind are, let's say, you know, a new hire within your organization who's looking for information, you know, either about HR practices or IT or simply knowledge about how to, how to function within your organization. If you're passing to your, or essentially, pointing your GenAI model at you know, let's say a set of internal knowledge based articles, yes, it's gonna learn and make good suggestions internally as well. An important aspect of this is security in permissions, and I haven't really emphasized this in enough detail, I don't think. But when Coveo was indexing content, we're not just indexing full text of the document and the metadata and so on, but also the permissions associated to it. So let's say you're pointing Coveo at your SharePoint internal, SharePoint Online, internal site, because I was able to index that yes, but also to limit based on user permissions what they can see from in their search results. That implies also that any content that's sent to the GNAI model when somebody asks a longer question is also going to be limited by their permissions. So they're not gonna get a generated answer that contains content that they shouldn't see. It's just not going to happen. So we think that our our Genay solution is especially useful for internal site search results and would love to talk to you more about it, at your convenience. Yeah. Exactly. It's really, you know, we talked a lot about the customer journey today, but it's all there it also helps with the employee experience and journey as well. So with that, I would like to thank you so much for joining our webinar. We will be following up, with additional assets and resources. And as Paul mentioned, yeah, please feel free to reach out if you'd like to see anything in more depth. And have a great rest of your day. Thanks, everyone. Great to talk to you today. Have a good day.