Hi, everyone. Thanks for joining. I'll just leave you guys a bit of time to to connect, and we'll be starting in a few seconds. Alright. Lots of people joining. That's great to see. I'll start my housekeeping and then, we'll get right into it. So thanks for joining. Welcome, everyone, to our webinar called AI powered website search. My name is Clara Blanger, and I work on the marketing team here at Coveo, and I'm really thrilled to be a part of today's presentation. I'm, also thrilled to introduce our speaker today, Alyosha Asanovic, who's a, product manager at Coveo. He'll be giving you the demo, today. I have a couple of housekeeping items to cover quickly before we get started. First, everyone is in a listen only mode, but we do want to hear from you, during today's presentation. So please feel free to send your questions along using the Q and A, tool on your screen. Today's webinar is being recorded and we'll send the recording, within twenty four hours of the conclusion of the webinar. And now for those of you just joining us, welcome to the webinar, AI powered website search. Now, Ali, please take it away. Sure. Hi, everyone. Thank you for joining us. It's nice to see, quite a few attendees in the cloud. My name is Ali. I'm a product manager for Kaveo for websites at Kaveo. I've been with the company for five years, and I'm excited to show you today some of the things that we can do. I'm gonna try not to be awkward because, when you don't hear the audience, you can make jokes and they can fall really flat. So I hope you're laughing at home. I'm gonna presume that you are. Alright. Let's get started. So what we're gonna go over today is just quick introduction. So setting the tone here for the this presentation, but in general, the landscape work, things that you need to think of when looking to integrate AI powered search. I'm gonna talk about the virtual website, of course, features of the product, some customer use cases, which I'll show it to you directly, and we're gonna finish up with some of the business value that Conveio can bring. Now when someone visits your website, forty three percent of those users will use site search first when looking for anything, a product or a category. So that's almost half of visitors that are gonna interact with the search box on your site right away. And when we think about the different experiences that we use on a daily basis, it it can sound a bit generic like this, but Pinterest, Amazon, LinkedIn, Twitch, Airbnb, whatever you're using, there's always that big search box at the top. It's usually one of the main ways that we interact with it. Of course, there are recommendations on some of these items as well. But when you think about interacting with each of these sites, you don't think about a specific feature. You have a need. You have something that you wanna accomplish. You wanna be entertained. You wanna find, somewhere to go once the COVID, lockdowns are over, things like that. But all of these experiences are more than just the search box, more than the individual items that make up the search. It's everything brought together that really defines the experience here, and that keeps you coming back to these sites. What Caveo allows you to do is fight against the big guys in this way, replicate these experiences, and, give you the chance to have your own type of unique experiences that people keep coming back to because they feel, that it is a unified whole. I'm not someone who likes a lot of words and slides because people tend to to read them instead of listening, to the person talking. What I can tell you is that in the past year, more than ever, we've had customers coming to us, talking to us about their problems, letting us know that they are struggling to adapt. They have to make up for lost time, time that they didn't spend initially, perhaps because they thought they could evolve more gradually, and switch over to digital. We're forced to do so now and just know that their business might end if they don't have the ability to upgrade. So when we talk about Cberico websites, what we can bring you is the ability to unify your content first and foremost. So whether you have a website, only multiple websites, communities, whatever is in there, we can pull data from all across that and bring it into one source. We provide personalized experiences, so results recommendations tailored with AI and machine learning in various ways, which we're gonna go through, that are gonna make experiences relevant. And, of course, we add value by providing you also with access to the data that we collect on the visitors, on your site, on any of your properties, allowing you to make decisions, whether it's to improve the experience itself or to make other business decisions. Because when you add a search box, that collects data on your site, the search box becomes the voice of the customers. It allows customers to tell you exactly what they want, exactly what they need, and it's the most important part of what we're gonna start with. One thing that I realize is that people tend to take all of the parts of these experiences for granted. And my goal here today is to make sure that you realize how valuable the small things that each individual component here, does are really to the overall experiences. So when I land on Amazon and I wanna search for something, I may not know exactly what I'm searching for. I may have forgotten. For example, the other day, I was searching for, something to wrap my knee when I worked out because I've I've been having some pain. I've forgotten the word sleeve as dumb as it sounds, so I see it there now. I searched for a knee. It popped up right away. I can be guided towards what I want. Amazon is just, super addictive, especially during COVID. But for the search box in general, here's what you need. You need to have query suggestions at all times. And, as Adamus' sounds, there's just not enough people who have them. I see so many sites that that don't have anything in place there. Dynamic ones as well. They need to be based on what users, are searching for, and they need to be able to change. It can't be just based on a keyword, for example, a static list of keywords. If your products evolve, if your site evolves as trends change, all of that needs to be reflected, and that's what a machine learning powered query suggestions can do. You need to have typo tolerance, especially for mobile, where people can miss a letter or completely misspell a word, and we figure out, what the word is. You have to have the ability to do category suggestions as well. So something like gaming laptops, in portable computers, allowing users to refine the search right away from the first, interaction. And coming back to that machine learning aspect, when we base our, query suggestions on user data, we know that we're dealing with natural language and terminology. We know that the people who are searching are gonna think in a certain way. They're gonna look for something in a way that you may not have thought and that other you other users have. Most importantly, they need to be based on success. So someone searching for something is not enough for us to suggest a specific keyword. You need to have a search result leading to a search page where the user actually interacts with items and clicks so that, we know that this entire chain of events is successful and that we can further influence other aspects of search. The other thing are facets. So So when you're on a search page, you're narrowing down the categories on Amazon, for example, on any kind of major sites for commerce or anything that you're searching for, support articles. You need to have the ability to, again, not take facets for granted. They are super important, but show only the facets that the user actually needs. So when I'm searching for a four k screen or something, I don't want a facet telling me what resolution do you want. I already told you. Right? I want to have machine learning figure out, have the ability to do automatic reordering so that the most relevant facets are on top. So if I'm searching for a laptop, maybe I want the processor, the display size. Those are the things that are most important. And then in order of decreasing importance, machine learning is gonna figure that out and sort both the facets and the values. Having the ability to do automatic selection as well. So we can narrow it down based on a query. So if someone is searching for a keyword and we figure out that a high percentage of people searching for this keyword tend to click on a certain facet, we can deduce that. We should probably narrow down the results immediately and go further. And we have the ability to do things like improve the entire scope of results without even interacting with facets. So knowing what users tend to scope down to, we can, slightly boost a variety of popular results, leading people to finding the best results possible within the first query. The results themselves, of course, are incredibly important. If you don't have those, you don't really have anything at all. They need to improve over time. You can't have static results. And one of the things that we see is that people who tend to build their own search, they tend to underestimate how much of an impact this has. When the table stakes are determined by players like, YouTube, Amazon, and all the big experiences that people face, getting to the point where you reach table stakes is really, really, difficult. And then you have to add on the machine learning on top of that. So your results need to be able to vary over time, respect trends as much as the facets, as much as the query suggestions. So we're seeing that all these things come and play together and results as they are, you have a user clicking on a keyword, being led here, clicking on a result, we're gonna start automatically being able to reorder those in a way that makes sense for the user for what we know about them. We have to have the ability to also have different search keywords lead to the same results because people are not necessarily gonna think about searching for something in a certain way. You can have also the ability to do things like, apologies, like, find results that don't even contain a certain keyword simply because we know that when user searched for something, they weren't able to find it. For example, they changed their keyword, came back to the search page, and then, our machine learning can figure out that a certain keyword, despite document, is pertinent to that document. It's gonna bring it up right in there, and it's going to reduce what we call content gaps. So having a keyword which doesn't lead to results and always finding something to present a user that makes sense. You wanna have the ability to have future results, of course. So have the ability to put something on the top if you need to promote, any kind of notice, any kind of product. It really important to have that. And, of course, a thesaurus. So having the ability to add any kind of natural language, understanding if you have a lot of acronyms, for example, in the medical field, something like ADHD. If you wanna expand that into the the full term, you can do that as well. And, of course, success based. So it needs to be always based on users having a successful outcome, taking into account the whole chain of interactions, and making sure that we improve results over time. Now the last component here and very important in terms of not finding but discovering our recommendations, Incredibly powerful, of course, whether it's in, any kind of support site or regular website for exploring content leading to conversions. If you have any kind of documentation that you want to, lead users to, which can lead to conversions no matter what your business is, Recommendations can be tailored to learn on, whatever behavior you want to encourage, and they're gonna be contextual. So when a user is searching for a certain keyword, we can have recommendations on directly on the page. When they're visiting it visiting a page that is not searched, of course, we can have, other types of recommendations that are relevant to the context of the article, to the user themselves, what we know about them, what other users like them will tend to do and want to do to explore more. So it can be based on a variety of events, page views, add to carts. We have different in commerce, quite a quite a few different, models for recommending popular items, a few together, all the the classics that make up experiences where conversion really is going to have an impact. But we can also do things like, explore products in a product space. So making associations, between a product in a way that I can explain by saying to you, for example, when I'm searching for, Victoria, the word Victoria can lead us to, for example, the concept of England. Right? We can have, the concept of queen. We can have all of these things that are closely tied together that don't necessarily make sense, when you just take the keywords, but that due to their proximity in documents, due to understanding, the the context, we can figure out the association between these things and make recommendations that are smart. And, of course, personalized. We can personalize based on the user's profile, their action history during a single visit, during multiple visits, or past purchases and other things. The most important thing is that we can do this on any platform. So when we talk about for websites, we do have deeper integrations into systems like AEM and Psycore, and others as well, but we do have the ability to connect in a more generic way. So we have generic rest connectivity. We have, web and site map connectors with powerful abilities to extract metadata and allow you to to take concepts out of a page, with very little configuration required. We also have an open source, actually, two open source frameworks for building your search UI. So a JavaScript search framework with, which captures analytics automatically, comes with a variety of components as well as a headless framework, meaning that we only handle the logic of interacting with our search, but you build the components. So you have a full control over every aspect of how you wanna build that out. And we do that across different use cases. So service, manufacturing, ecommerce, we really have the ability to put our technology to use in different use cases to fill different business needs, and that's something that you can discuss, if you want with us on a one on one conversation to see examples specifically of what we can do. But for the moment, I'm gonna just skip in quickly into a demo of a few of our customer signs. One of my favorites is, Faskin. So this is a a law firm, which, of course, you may know. What I love about them is that they have this, amazingly big, search box here. Search box here with, of course, query suggestions. I can, search, immediately here. They put a huge emphasis on it. There is a mix here of different results. So, yes, we have people, but we can also search for something, like, you know, whatever criminal. I'm not sure exactly here what I would I was just gonna sell in. We have the knowledge articles. We have, different layouts. We have a great people search where, not only can we find the people that we're looking for, very important in the law firm to have that kind of recognition. But when we land on this page, we have Coveo powered search components that allow you to do things like this here. So we're using Caveo to pull out related work for this lawyer. So we know that, they operate in different, practices here, what their areas of expertise are, the industries, and we can find articles relating to this lawyer and allow us to make the page more dynamic. So this is constantly going to evolve. As we add content, everything is going to be pulled up at today. Another really cool use case that we have here is for Myka. So selling, having the ability here to search, for, tiles, for example, by color. We have a nice layout here for facets, for things like that, where you can have this really vivid and creative way of exploring the the results directly here. In terms of recommendations, booze is my favorite one with Virginia ABC where we have, of course, the ability to search here for, products directly so we can land on a single page, Find a result, and then we have recommendations. These may seem a bit, easy, you know, superficial. It could be relate, related to the type of alcohol that we're looking at. But in reality, when we search for something like skull vodka, for example, what's interesting is that we we select, for example, here this crystal head vodka. What are we gonna get as recommended products? Yes. We do get some vodkas here. But in reality, when people are searching for a skull vodka, they're usually not gonna be looking for the best case possible. What they're most likely looking for is, the, a novelty gift, for example, for Christmas, for a birthday. And based on user behavior on their patterns, we figured out that these novelty items are gonna be related. So we're able to recommend those when landing on a page like this instead of recommending Grey Goose or any kind of other Vodka directly. In more complex ecommerce cases, you can have pages like this that where we power the Dell product search and, support articles in the same interface, giving you the ability to have complex product results, templates that surface the most important parts of the product here. Of course, query suggestion still, having things like, the the facets for the display size, memory, storage types, all the things that help you do a breakdown of what you need when you're searching in an ecommerce aspect. But we also have different use cases in, for example, health care here where the ability to find a doctor or a medical provider is really important. So MetroHealth are doing two really nice things with us here where we have, of course, the full search on the site as well. But we have the ability to do this people search where we can search for, by specialty, for example. We can narrow down things by, a category and refine your results for a doctor, allowing people to find exactly what they need in a really quick manner. Another thing that MetroHealth is doing is using Caveo for map search. So the ability to find a location where someone can be serviced, by doing a a ZIP code, by getting the location directly from the browser, things like that. And you have the ability to display results directly on a map. So this is Caveo powered here and bring up, results for people for, sorry, a distance search for location search as well. One last example, and I find myself talking really fast here, is, one of our demos, so called Habitat Home, that showcases a very simple way of doing personalization. So when I land here on this site, I'm gonna do a search for a speaker. And what I'm gonna get in the results are a variety of different speakers. So I have outdoor speakers, floor standing, car speakers, just a huge mix. But if I go and visit an article on setting up a perfect home theater here on the right hand side, What happens so this is a a Sitecore site, which you may not be necessarily familiar with, but we've identified here the user as being as someone who is into home cinema. And with that information so this is set up pretty aggressively, of course. We would want more data to figure out what a person really wants. And if I go back and do the same search, I have now a a completely different set of results. Right? I know something about the user. I can customize this here. I'm gonna see five dot one packages, sound Barca, amplifiers, things like that that are much more relevant to the user than seeing car speakers or outdoor other speakers. All of that is administered through the Coveo platform itself. So, whether you're dealing with a technical user or a business user who is in charge here of, analyzing usage usage analytics reports, tweaking results, getting an understanding of of the content, it's really simple to do things like add different sources of content. We have about fifty connectors, available, and generic connectivity as well. So giving you the ability to really integrate any kind of source into your search that you want. Let me close that Dropbox drop down here. But most importantly, when we look, for example, at the the result that we saw here for speakers, if I go into what we call query pipeline, so a place where you can manage, things like ranking rules and, thesaurus entries. Everything that relates to search on a certain interface can be controlled through here. So when I see, result ranking, I can add a rule that says, if I've discovered that a user is an aficionado of home theater, I'm going to boost, category of theaters. Now this is a manual rule. We're not even talking about machine learning in this aspect yet, but it's a very simple example to say, I know something about a user that can have an impact. I'm gonna decide to do a bit of manual boosting here if I wanted to, or machine learning is gonna figure it out, in the long run as well here. In the short run, I would say, I believe it doesn't take that much data for machine learning to figure out, what to do. So you have too many options to go through in a webinar that's as fast moving as this one. And, the most important part for me here is realizing that when you build search yourself, you end up being often limited in terms of what you wanna do in the future. So you, start by recreating all of these features individually or at least trying to. And then when you need to do something more, you end up having to, reach into, your pocket, invest in a linear fashion once again. So keep adding these features, keep maintaining them. But on the Caveo side, you can get started with something really simple that's gonna blow away what you have already. And then you have an arsenal of tools at your disposal, AB testing, hosted search pages, different, ways of influencing the search results themselves, reporting capabilities, of course, as well. So you have, different dashboards that you can use, to analyze data that, that your users, that gets logged when your users visit your site, for example, here. And now on this site, there's not a lot of data, but no one can see quite a a a variety of different things. So at the end of the day, I'm gonna end it on this here, and stop talking so fast. The most important part is knowing that, when you have one property, of course, you can integrate Caveo, here, but we have the ability to service an inquire incredibly wide array of different use cases. Commerce, of course, community and support, intranets, support agents, websites, all of that can be brought together, the content of it. We log usage analytics data. We collect data all across the board, and we can unify user experiences so that your users, what they do on your support community is used on your website, is used on your ecommerce platform. Everything comes together nicely. That would be it for my part. So I would I will, let Clara take over for the rest. Thanks so much, Adi. That was good. Just before we move to questions, we wanted to show you the kind of results that, some of our customers have been seeing with, AI and relevant, relevant search. So as you can see, we mentioned FASTQN, twenty two percent increase in time, and we have some other really great numbers there. And if you're curious as to how you too can get to, such great numbers, we do offer website, assessments free. So basically, when you register, we'll go look at your website, and, our experts will do an assessment of all of the capabilities of your website and provide custom and actionable recommendations on, how to improve the search and personalization of, your website. We'll send a report, and you're you're good to go and get started on, on your project. So on the next slide, Ali, if you can switch, you have the link, and we'll make sure to send it by, email also afterwards along with the recording. If you haven't, posted your questions yet, now is the time. We'll get started with those that we have. So, Ali, how do you handle multilingual sites, and are you able to show an example? Yeah. I mean, multilingual sites are pretty easy to handle. So we have support for, I wanna say seventy three languages, but it's somewhere in that range. So maybe forty three. So a high number of languages that can coexist and index. We have the ability to, take the context that we know of the page. So if the site is in French, of course, display the UI in French, display UI in in different languages, search for things like, Chinese or Japanese characters, for example. All of that can work together seamlessly in the index. That's awesome. Thanks so much. And, how much administration is required to keep the machine learning accurate? I would say none is required. Where the machine learning models will continually learn, you can initially set how much data you want the model to use. So that's a very simple parameter. We're talking about a query suggestion model, automatic relevance tuning to improve the results, dynamic navigation for facets, things like that. These are all different use cases for models. You can tell it to use a month of data, three months of data, six months of data, and you can tell it to refresh at a certain rate. So if your data changes often, you can have the model relearn, more frequently. So that's very minimal tuning. Now if you want to have a very specific use case where you use, only certain data to have the model learn. You can customize that pretty easily. That still doesn't represent maintenance, more of an initial setup. And then you can do things like model testing. We have a section on a platform allowing you to compare the results of different machine learning models and see what one keyword, for example, is going to yield as a result with one model and with another that you've tweaked. So you can do these kind of previews, use AB testing, and have different ways to influence your, your machine learning models into testing. Thanks, Ali. We have a follow-up question to the languages question. Are multilingual sites set up a separate accounts with Maccoveo to help maximize machine learning dictionaries, etcetera? So you you're gonna have a a few things. In integrations like Sitecore, we have the ability to, so we index multi language sites by default. For example, we're gonna have multiple versions of the item, that are tagged with a certain language. So if you're in an interface where, it's a certain language, you can filter out the, the the different, other languages. And sorry. The question disappeared because there was, like, a I think I'm gonna into the answered slide section here. Okay. So it's not, per se, a separate account, more as different items in the same index tagged as a certain language. We can decide to integrate those, work with them separately. You can set conditions where you say, if an item is of a certain language, I'm going to apply a certain rule to it. I'm gonna do boosting. I'm gonna add add the source entries that only apply when it's a certain language, for instance. So everything is fully customizable, and you can create segments. But at the end of the day, all the items live in the same index. It's up to you on how you wanna structure the separation. K. Thanks for the clarification. Another question. Can Coveo power different search boxes within one site and do scope search? Yes. So if you have, different search Barca, we have a concept called a search hub, which is the origin of a query. Machine learning models, for instance, will learn based on that origin, first and foremost. So if you have a website where you have a generic search where everything is mixed together, you have a new search and you have a product search, those three things can be treated as distinct entities where you redirect to different search pages. The machine learning models are gonna create subcategories, submodels where, the query suggestions for the new sections are not gonna be the same as your products or as your generic search where everything is combined together. So you can have multiple search boxes, a single search page or multiple search pages, but, usually, you're gonna split that up if you have, like, a new section or something like that. So, yeah, you can definitely have multiple search boxes and everything can be, scaled. Very good answer. This is unfortunately all the time that we had for today, but if we didn't get to your question, we'll make sure to follow-up with you, after the webinar. I wanna thank you for attending on behalf of Ali, myself, and the Coveo team. And as I mentioned before, we'll be sending the recording as well as the link for the website assessment as a follow-up email in the next, twenty four hours. So thank you again for attending, and have a great rest of day. Thank you. Thank you.