Hello, everyone. Welcome. Welcome to our webinar, AI powered search for customer support. My name is Maggie Bliss, and I'm your host and moderator today. I'm really excited about today's session. We are going to be doing a live demo of our service offerings, and Kevin Club is here. He's our sales engineer manager, and he will be walking you through all of the wonderful things that Coveo can do. So I'm sure you're all seasoned webinar goers at this point, so you know that you're in listen only mode, but you can ask questions via chat or the q and a functionality, and we will answer the questions at the end of the session. And this will be recorded, and I'll send it out to you either later today or tomorrow. And with that, I'll pass it off to you, Kevin. Alright. Thanks, Maggie. Good afternoon. Good morning. Good evening, every everybody, depending upon where you're, you're joining us. I'm gonna go ahead and start sharing my screen here. And today, we're gonna be talking about Coveo for service, realizing that a lot of your customers will come out to your website for a couple reasons. One, they're existing customers of yours, and they have questions about the products that they currently have. Two, that they are interesting in purchasing additional products or licenses, whatever it is that you may be selling, and they're starting to do their their research. Or they're coming out just because they may not have a question, but they just have curiosity about the products or about your your your company as a whole to start in and and to investigate and learn something on their end. So being able to be relevant to your customers makes for a much better experience. It's things we've been all accustomed to from an Amazon search, Netflix, Hulu, insert whatever streaming service where they always seem to know the next product or the next video or movie or even music with Spotify that you're wanting to watch, listen to, or buy. And it's always a nice experience when I go out to a customer's, to a a corporate website or support portal, and they know what I'm going to ask sometimes before I even start asking the question. And so relevance is huge for us. And if you think about it from a support perspective, let's allow your customers to find the answers to their questions at the level zero support, meaning that they're self serving themselves versus having to open up a ticket, being able to reach out on a phone to speak to to an agent, or sending an email to generate and create a ticket that then consumes resources on your end. And then you get into whole conversation of how quickly can we respond to these questions and how quickly can we service the customer and get the right answer at their hands in an efficient and timely manner while also not overburdening your agents on the support side of things. So we've come out to we're gonna well, first off, we're gonna look at our demo environment from a Coveo perspective. But as you can see at the top, these different, tabs that I have open here, these are some of our clients that are using Coveo for service. We wanna spend some time looking at how they've used it and, you know, some of the functionalities that you'll be able to go out and visit on your own after this session. So from a visitor perspective, when we come out to the community here and I have questions about this new fitness watch tracker that I just purchased, you know, these little things that we all wear and track our steps and our calories and everything else, within our lives. I'm unauthenticated. I have not logged into the session yet. But we talk about from a relevance perspective. We wanna be able to take the content that is most relevant to me as a user. Even though they don't know who I am, Coveo still knows a little bit about who I may be. And what do I mean by that? Well and this is not part of the an implementation. This is just for demo purposes only that we have this. We're extracting information to make for a much more relevant experience for the user. Meaning that we know that they are speaking English because that's a language pack that they have installed within their browser. We know the browser that they're using. We know geographically about where they may be. So we're able to push and modify content based upon what we're able to know about this anonymous user. And then the hopes are as the your customers start, even as an anonymous user, start to be able to perform search, click on things, make adjustments to the search experience, we start gathering that information as well, and that starts to alter what future, search experiences will look for that individual user. Now as they authenticate in, and if they're authenticating in through a system of a CRM systems, yeah, we can certainly use that information that we have. If you know about what products that they own or when their support expires, what level of support that they may have. You can certainly leverage that information that you have in the outside systems to be able to impact what content gets displayed and how it gets displayed in the messaging behind it. So during our time today, we're gonna talk a lot about machine learning. And it's that machine learning that allows the your customers to have a unique experience, that Netflix, Amazon, Spotify experience of saying, boy, they really are in tune with what is important to me. So the first type of machine learning that we're gonna be talking about is event recommendations, and we've we've all seen it. It's the people like you have also viewed. It's the community post that may interest you. Even as an anonymous user, we're still able to ascertain what may be of interest or what has been popular in the past and present it to the end user. So unbeknownst to the end user on this or the your your visitor here, this is machine learning that's at play here. It's just not a you know, that we hard Coveo these specific articles or content that you may have in your content library and your search results that we're pushing out. This is relevant based upon, you know, maybe geographically where I am. If we're selling snowblowers, well, maybe we'll see snowblower information about here. But if I lived out in Arizona, the AI will say, you know what? We know where he is. It's not relevant to push snowblowers to somebody that's never gonna experience snow in their neighborhood or in their state. So from a machine learning perspective, that certainly makes for a much much more relevant experience. But then also as they start to perform searches now as I come up here and I click into the search box, we'll see some of the queries. And this is not a dictionary dump of just different terms that we have within our search index, but rather this is information from previous queries that have been valuable to other people. So when I come out here and I say, how do I pair or how do I how do I pair? And even as I start to have typos in my query here, Coveo is gonna say, hey. Time out, Kevin. We know you're a bad speller. Is this actually what you mean? So we can also start doing some typing and correcting here. And I can make the selection here and say, alright. Yes. I did mean Bluetooth pairing. Or I can come out, and then we're looking at the search results here. Here, I can then come out and say, how do I pair with an iPhone eleven? And then get my, you know, the a query that's listed here. So a couple things, again, from a machine learning perspective that are in place on this on this page. I I had a question about how do I pair my watch with an iPhone eleven. The search results here realized that this is a question that was being asked, and our smart snippet machine learning model was able to determine, hey. We have the perfect article for this based upon the question that you asked, and we're able to frame it into the top of the search results page here. So meaning now I'm able to I don't even have to click into this document to go off elsewhere to be able to view it, but rather now I'm able to follow along with the different steps, click the buttons on my watch, and I have my watch paired with my iPhone or my iPad. So things that we're accustomed to from a Google experience when we ask, you know, who was the architect for the Eiffel Tower, you'll get a little nice little box talking about mister Eiffel and his accomplishments in addition to the Eiffel Tower. The same thing holds true from a Coveo perspective. You know, a lot of our customers have been spoiled by Google and Bing and other search search engines that they want that same experience for their customers or when we talk about it from a workplace experience for their employees to be able to benefit from as well. And this smart snippet is just one example. Going above and beyond the smart snippet, we see that we also have questions other questions that people have asked, kinda helping the helping your customers, you know, interest in kind of going down a specific path or kind of going in the right direction for the question that they may have. But you'll also notice that the organization or the way that we're presenting the results onto the page here is machine learning as well. We're just not taking the query and saying, good luck. Here are all the, the content that we have that's related to it. Now we're scoring all of these different articles. And then based upon the scores from previous people's searches, what we're then doing is saying this content here is the most relevant. In fact, we're actually recommending this because it's been so successful in the past. So the nice thing about this is that your customers don't have to scroll down to the very bottom of the page. They don't have to move on to page two, page three, page four. It's because that automatic relevance tuning, what we call ART, if you're looking through our documentation or our website, you may see the acronym there. The automatic relevance tuning is making a determination where that content needs to be displayed. Visiting your website. But we also realize that you your company is run by humans and consumed by humans and has, you know, different requests that may need to come up. If you're offering a new product or a new release and you want to make sure that the new release, when people, type in version ten of the your software in the query box, so we always put this piece of content up to the very top, and we're able to go above and beyond that automatic relevance tuning that we're seeing on the page here. You're gonna have the ability to do that. If you had certain content that you wanted specific to queries that are being asked, that if they say, if they're querying for product documentation or new release and that's their query, then maybe push that content up to the top. We'll see some examples of that as we go through of how you're able to work with the levers within the Coveo administrator to be able to get the the business rules and the the search results that you would that you would want. All of which we're seeing here, both from a Coveo end user perspective and the back end side of things, there's no need to know a programming language. There's no need to have to know a a specific query or report writing to be able to get to the functionality that we're looking at today. So the hopes are here that your customers are able to ask questions. How do I pair with an iPhone eleven? Find the answer. If it's not here or if there's not an appropriate smart snippet for it, well, within the results, they'll be able to see the title, the summary, the the the, the content itself. And you'll notice that content for us can be anything. In this case here, it's a knowledge base article. It's a guided flow. It's a PDF document. It's a video. So you think about all the different content that you have from your on your support site that may be within a knowledge article. It may be your cases. It may be a community where people are asking questions. If you wanted to make that part of your Coveo index to make it searchable, that's very easy for us to be able to do. And we'll talk a little bit about that when we get into the administrative side of how do we how do we connect those sources of content into the index to make it searchable by your customers. So the nice thing about this is not only are we providing relevant search experience for the for the customers or for your customers, but also giving them the tools so they stay engaged on your website. So when I come out and I see this article may be of interest to me, let me go ahead and look at at it. Instead of clicking on the URL and being taken taken to this specific URL here, I can consume it within the confines of the search results here and then be able to see not only the query that I had asked, if there are any images that were still available, we'd be able to see the images there. We see the content itself, and we see the query that's highlighted and color coded here. So this then allows me to navigate. Even over on the right hand side, we see the color coding. So if I wanted to pop down to iPhone, I can certainly click the buttons, and it'll navigate me through the find feature functionality. But I can also look at the heat map over on the right hand side where I see the clustering of those colors. And then from here, I'm logically able to ascertain that this is probably where my answer is. So just these small little steps that keep your customers engaged, but then also to allow them to get to the information as quickly as possible, hop out of it. And then if they wanted to consume this video, and I fall victim of this all the time. I click on a video on a cuss on a a support website. And if it takes me to YouTube, guess what? I'm watching puppy videos. I'm watching baseball videos. I'm doing something other than what my initial intent was. But here I click on the video. Let me just lower my volume here for a second. As we able to consume this, I can watch it. I can pause it. I can continue on. But then as I pop out of that, I'm not lost to YouTube. I'm still within the search results, and now I'm able to look at the PDF file or whatever it may be. So it's these small little things that will allow your customers to be able to serve themselves and get the answers to their questions. Thinking about this a little bit further now, how can we also allow them to filter this information, to be able to sort this information by relevance and date, segment this. And, again, in your environment upon implementation, your miles miles may vary here, but you may say, you know what? We don't want these tabs at the top, or we wanna add different fields to the sort functionality or these these different facets. Can we add different values to that? And we're gonna see some examples of some of our customers in a little bit of how they're able to do that. So the answer is yes. Essentially, the facet values that we're looking at here is all metadata. It's metadata that you have within your index from the content sources that you have. It's, you know, those that are out of the box, like the author's name, the create date, the last date that it was updated, the keywords, that type of the title, that type of information. But, also, if you have custom metadata that you're using for your different content sources or different articles, you can also map that into Coveo to allow it to be searchable within the search box and then also allow it to be used in the filters and the facets as well. So as you start to think about the metadata that you have, it can be leveraged above and beyond the search functionality of it to allow your customers to say, alright. I'm looking for a specific document, that may be a video or it may be on a specific product version or different product family that you may have. So the whole thing within Coveo is that level zero of support. Allow them to help themselves in hopes that they're able to avoid a ticket. Now let me go back to our home page here, and I'm going to authenticate in. And I'm gonna refer back to our visit browser at the very top again in a second. So as we go ahead and log in, and now we're logging in as authenticated user, we're gonna be leveraging Salesforce in this situation. This information that we have about Susan now, you're gonna see the page looks a little bit different. Why? Well, because Susan has done some searches in the past, and we have some more additional information about her. But we're also able to leverage information from Salesforce or ServiceNow or Zendesk or whatever system they may be authenticating to or where you have that information residing. So Susan's able to pick off where she had left, last left left off. We know that she had purchased a Speedbit Blaze, and we know that she likes running. Well, how do we know that? Well, if I refer back to our visit browser here and we'll talk about the actions history in a second. But the CRM profile is that information that we're pulling from the CRM system. So this information of her interest is then determining what appears here. And then because she's purchased a Speedbit Blaze, we know that she owns that product or products, and we're able to manipulate that as well. So this is where you start getting from that anonymous user coming in, starting brand new. And as they start to perform searches, we start building upon that in conjunction with the successes of other people for similar type of queries that they may have asked in the past. And we start to build that story for that is going to be unique even to that anonymous user. As that anonymous user then authenticates in, we're then able to start to stitch that search experience together. So then if they end up moving into a cell phone or onto a tablet and go mobile from their desktop or vice versa, we're able to continue that story. And suit from Susan's perspective, when she logs in on a mobile device, has that same experience, same search history, and we're getting that same machine learning models that are being recommended to her. So I mentioned before about being able to track the information and to make the much more relevant experience in conjunction with the machine learning. You know, we talked about the the query suggestion model. We talked about the event recommendation models and then the automatic relevance tuning, from a machine learning perspective. Well, what feeds that machine learning monster, that we come in and take a look at all these different clicks? And these are the different clicks that I had taken while we're talking on the phone here, talking in this this webinar. We started off on home. I started performing a a query. I clicked on some documents. I then performed another search. I then ended up coming back home to where we are now. This information is relevant because, as I mentioned before, it helps tell Susan's story when we start talking about stitching together and a much better search experience for her next and subsequent visits. This information, when we get into the administrative side of Coveo, also feeds the analytics data. And so all this information here helps from a reporting perspective. So now you'll be able to see in reports, well, what are the most helpful most queried topics? What where do we have con where do we have content gaps of where people are asking questions that we don't have content to be able to help them out with? But we'll get to that in in a little bit. So the hopes are that Susan's able to answer any questions that she has online. Unfortunately, there are people like me in this world who, if I have a question, I wanna be able to speak to somebody about the the the question that I may have. So I'm looking for contact us. I'm looking specifically for a phone number. And the way that we have the case deflection model here set up is we're going to ask the people, the your customers, to tell us a little bit about what the problem is. So I'll come out here, and I'll I'll type Speedbit Blaze. I am having trouble with pairing with my heart rate monitor. Actually, that's not too bad. One typo I think I've got in here that's pretty good for me. So what's happening is you notice as I was typing along, this area down here started to populate and show onto the screen. And a lot of times, customers are asked to be able to classify into your system what the category, what the reason, what the product that they own, or whatever the questions may be to fit into your information. And we can see, you know, the case reason and the product here. And if I click on the drop downs here, we see a listing of all the different, products that I can make a selection to. This is machine learning at work. What's happened here is we're making in these little ovals here, we're making recommendations based upon what's being typed here. So we see we're pulling out that this they're making the assertion that there's an issue with tracking, and so that's why those are in the little bubbles here. So I'll make the selection here, which is one of the values that fits into your case resolution issue or tools that you may have. And it pulled out the the speed bit Blaze. And so now I'm one step away from providing you the information that you your agents need to solve the questions. I click the next button. They know what the problem is. Now we're gonna put together a a search in hopes that we take that last gasp effort to be able to solve the solve the issue at hand. And I can then take a look at the or Susan, because that's who we're authenticated and asked, to say, you know what? This is great. Or I look at this specific article and, boom, found the solution. We we deflected that case. We didn't have to involve an agent. Great work. Now if they still have the problem, what we're doing is just validating the information that we have. We know that there's, you know, from what she had typed in, the information that we're seeing here, and then she can go ahead and and submit the case. So this then would not be measured as a deflected case. This would then as they submit the case, it'll end up in the hands of one of your agents. So what does that experience look like from an agent perspective? How can we pull that information in to allow now that we've involved the agent and it's gonna involve some of their time, how can we make that time as short as possible to be, give them the ability to effectively answer it and keep the customer happy with it before they move on to the next case that that agent may have. So let me go ahead and authenticate into the agent tool. So now I'm changing hats. I'm no longer Susan. I'm no longer a customer, but I'm a support agent. It doesn't matter. I can be a level one, two, three, or however you may classify them. But this is within Salesforce, and this is just the example that we have here. It could be ServiceNow. It could be Zendesk, or any ITSM tool that or tool that you may be using to manage your your cases within. The beauty of this is that your your agents are still gonna work within that single pane of glass. But from within that pane of glass, Coveo is gonna the Coveo search experience is going to be, become part of it. So we look over on the left hand side. We see that we have the subject, the description, you know, the con person the subject or the the contact's email address and some additional information. And you may have more fields that you're gathering information on that last form that we just filled out or maybe akin to what subject and description. Regardless of the fields that you have and whatever information resides within your case management tool, all of the search results that we're looking at here from that's being pulled into to whatever fields that you wanted to specify within the the case itself. In this case here, it's leveraging off of the subject in the description. So the beauty of this as an agent being an agent here, I know when I open up this ticket and I start looking at this within, the Salesforce case management tool, I come over here. I know that all the content, the articles that I'm seeing here are all for this specific question that was being asked by Susan. Now how do I get the this content into the hands of Susan or whoever whomever may be asking that question? Well, we have the integration for us to be able to click send us an email. Or if you're using any sort of, you know, chatter or, you know, conversational tool to be able to push that information or this specific link out to Susan or to the customer, or just to be able to copy the link if you needed to if you're corresponding via email or Outlook to to her, however may may occur. But a lot of tools that we have from a search experience that we saw for Susan also apply to your agents. So if they wanna take a look at this specific article in quick view mode, they can. They click on quick view. They look at the article, make sure it's the right content for the question that's being asked, pop out of it, and then they move maybe move on to the next one and they say this is the best article here. Let me go ahead and send this and attach it as an into an email. A couple of you probably have noticed already, even on this page here, that we see content that has been viewed by the customer already. That's the beauty of the tracking that we're able to do. As I mentioned before, it feeds the machine learning and helps to make better recommendations for your future customers and for Susan moving forward. But also when we're capturing from the analytics perspective, we're able to run reports, but we're also able to leverage that information into the agent insight panel for the case management tool that you may be using. And this here then helps and matters to your agents because now if I go ahead and send this article back to Susan, she's gonna say, well, you've I've already read this, and it didn't do the issue. And then they it resolves. It gets into a phone call or another email. Or that agent may say, alright. She's already read this. Let me point out the paragraph or the sentences in the document where she needs to refer to. So it allows your agents to be a little bit more pointed in their responses potentially in helping somebody that may be having trouble. Obviously, they're having a question because they opened up a case. Let's take that a step further now and talk about where that customer may be within their journey. And the hopes are that Susan has just come out. This is a single isolated incident question that she has, and your team is gonna be able to answer it in a way they go. In some other instances, the session history is going to give you a full history of what Susan has done in the past. What are some of the documents that she's clicked on? What are some of the queries that she's looked into? What were some of the events? You know, the the click paths that she has gone through to be able to get to the point to where she is. Giving your agents a ton of insight into where that customer may be within their journey and what have they tried to do to solve this themselves at the level zero of support before they reached out to you, but also give you an idea as far as where they may be from, where their head may be. You know, are they angry? Or this is something that's been festering for a while? Is it something that has been, you know, every four or five months, they come back and visit it and allow your agency ability to make those determinations and how they respond and the words that they may use in conjunction with taking the content that they may be making recommendations to. So the same thing that they have from the search experience, as I mentioned with the quick view, they also have the ability to look at the different facets. And the facet experience or the search experience may be different for your agents compared to what it is for your customers because the access to the information that they may have internally may be a little bit more than what your customers may be seeing. If they're not authenticator, they may be authenticated. So the search experience can also change as well. So let's take a look before we jump in and put on our Coveo administrator hats. Let's take a look at some of the customers that I'd mentioned before that are using Coveo from a service perspective. We were just pulled out of Salesforce from a, demo perspective, logged in for the case management side of things. But Salesforce is one of our biggest customers. So if you've gone out to their website and perform searches in the past, chances are you've already used the Coveo search experience. So the likes of Trailhead, the events, of being able to get their from their documents and getting information from there. The same thing holds true that they have from a query suggestion, the automatic relevance tuning. Remember, that's the the machine learning model that'll push the most relevant content up to the up to the top. That's what they're doing. So if I just click on one of their suggestions here, give you a little bit of a look and feel of what their search experience looks like, You'll see that it has the facets over here on the left. They have some tabs on the on the top here, and then they're a little bit smaller as far as what we've seen, in the search example that we were looking at within the fitness watch, the Speedbit website that we were looking at. If we take a look at a company that we leverage for our data, on our analytics side of things, from a Snowflake perspective, they too are using the machine learning models that we've been talking about to be able to get information and questions from customers and push that information out. And we see that whatever I'd selected there really didn't have that much content behind it, but they're using the automatic relevance tuning. The thing that they that we see here that we did not see within Salesforce is that quick view functionality. So you'll notice that each customer that we have has a little bit of a different take on what that experience looks like for the customer. And to take it away from the, high-tech side of things and into the consumables, I guess, side of things of Formica. They're leveraging our search. And so I come out here and type they're not using the query suggestion, but if I just type in countertop, the look and feel of this page of the search results page is gonna be a little bit little bit different. They pull a little bit more imagery into theirs, as well. So they're taking some images that reside within the different products that they have here. They have the facets that are specific to the different sections from within their websites. They do not have or not using the the quick view functionality, but it has a nicest clean aesthetic look, that, you know, goes in line with the Formica brand. One of my favorite sites to show you is the KB, the knowledge base for VMware. And the reason why I say that is, one, they have millions of articles and content that they're indexing across multiple countries, across multiple languages. They're also using it for three different types of audiences. They have VMware is using it for their agents. So when you call in from the support side, they're using the the all that content, the millions of content pieces of content that they have is being used by the agents to answer questions. It's being used by their partners. And if you think about the hierarchy as far as who has the should have the most information at their fingertips, well, it would be the agents first. They have the most. And then the partners probably have a smaller subset of that information at their fingertips to be able to get information on. And then the customers have probably the most restrictive as far as the amount of content. We're still talking millions of articles anyway. But here from the consumer consumer side of things, as I come out, I'm not authenticated into the system, but I have done searches in the past. So we see some of my recent searches here. It's also then powering the event recommendation machine learning model for articles that are recommended to me. So all of these articles we're looking at here have been successful queries by others in the past based upon also taking into consideration queries that I've made in the past to end up on this page here. Now if I take this page experience that we're looking at here and I pull over an anonymous browser, meaning that it's treating me like I'm coming out here for the first time. I have no search history. Over on left hand side, we see the history that I have. Over on the right hand side, we see no history about me because I'm I'm incognito. I haven't actively done a search yet. But what you'll notice is the difference in the recommendations. These articles here are all related to adding memory and CPU throttling, queries I've asked from the past. Over here, it doesn't even understand that I only speak English, but it's also urging me to, hey, Kevin. Download this information or download this installer. Get started with it, then we can start to provide you additional articles recommended based upon me clicking on this. It's gonna say, alright. Well, the next article you need to look when you come to our home pages, how do you, not how do you download it, but how do you install it or how do you use the product? So this is one of the, my favorite sites. One, because I I use it a lot in my home here, but also I like the aesthetics as far as the recommendations. They're using the query suggestions that we're talk that I was talking about. But even as we get into the search experience here, the facets I find that they have are so rich. They have all these different commune they have communities or documents. They're indexing content from their blogs, even their downloads. But then when you get into it a little bit deeper, well, what product are you interested in? And, of course, of of those products, what versions are you looking for? And you speak Spanish? Oh, okay. Let's find the Spanish or let's find the Japanese, whatever language, and we can continue to whittle down the search till we get to a point where we have relevant content to them. Now if I were to authenticate in, maybe it does automatically know that I'm speaking English, and it knows what products that I've already purchased the same way we knew that Susan had purchased already a Speedbit Blaze of the watches that she owns. So finally, before we jump into the administrative side of things, the last thing I like about this, and I see this on a lot of custom a lot of customers where they are incorporating a community, where they have people asking questions and people answering questions, is the fact that they will not only index that content for the communities here let me make this a little easier, and I'll filter or facet this value out so you or into it so you can see it. But it will also let me know which one do we have the correct answer for. So I'm not wasting my time performing a query where I keep on diving into these discussions where there's no answer there. And I just continually get frustrated over and over and over again. Here, I perform a query. I dive into the community to see if there's a correct answer. I know that there's a correct correct answer here because it's it's documented on the search results page. And in fact, I might able to get a little bit of a sneak preview, as far as what, you know, the the the, the subject of that may be before I dive in and take a look look, at that. So I'm gonna hold questions. Usually, I pause for a second to to take questions at this point. I'm going to go ahead and pause, pause at the very end for and open it up for for questions. So let's take off our agent hat. Let's take off our customer hat now and talk about how do we set up all this information? How do we get this information searchable by you by my customers? How do we take a look at the reports and the analytics side of things? And that's done through our administrative console, which I'm logging into right now. Within the administration console, there's, I guess, three areas that we're going to talk about today. There's the sources, getting the content that you have into the index, meaning I have a Jira environment. I have a Google Drive. I have an on premise database that I would like to have crawled, or I have a file server that's on premise or a box account or SharePoint. All of those areas that are that those technologies that I just mentioned are sources of content for you that you may wanna index and have searchable by your customers, by your agents, by whomever. And how do we get that information into the Coveo index? Well, it's just as simple as coming out here and clicking ad source. We have about forty seven out of the box, connectors already at your disposal to say, okay. Well, yeah. You know what? I want to incorporate our confluence, instance into the, into the into the index. Or I wanna be able to take our sitemap that we have for our blog and pull that in and make that searchable. Based upon what is being selected here, Coveo is gonna ask the appropriate questions of how do we get access into it. And in this case here, we'll just give it a name. And in this case here, we'll also just give it a relevant URL. And the way it goes and Coveo would look at that, site map because it spells out each individual page that should be indexed. It'll go out to that page, get that information, the metadata about that page, the URL, the title of it, and put that into the index. Anytime we talk about security, and it doesn't really happen too often with, site maps, but probably more so something on, you know, Confluence or in the Salesforce where you or, SharePoint where you have certain content that is only for a specific group of people's eyes, meaning that it's only for the legal teams. It's only for your marketing team, or it's only for people that may be at a specific tier of your support of people that have purchased, support from you. So we always support the security of the document itself. So if I come in let me go ahead and close out of here, and I'll go into an existing I'll go into an existing, source that we currently have right now. So I'll open up this Salesforce and show you what it looks like when it's edited. And if you remember, when we're looking at the site map, it just said give me a name, give me the URL for the site map, and away we go. Well, here, once we provide the authentication for that Salesforce instance to get into it, now we start looking at over on the left hand side, the last alright. Well, what objects, including third party objects, what objects do you want to be able to incorporate into your index? And if I wanted to say, you know what? I'd be interested in taking cases. I type in case up here. I would then make the selection to case here, and then all the appropriate fields within that case would then come over. And I can then say, alright. Well, I also want to pull in our knowledge. And I pulled you know, come over here and type in knowledge. And I just kind of rinse and repeat for on the knowledge side of things. The beauty of this is not only we're able to get to any of the objects that you have within Salesforce or any of the content that you have from within SharePoint or ServiceNow or whatever that content source may be. From a security perspective, here we're able to make modifications to allow everybody to view it or to be able to follow the system permissions, meaning whatever system permissions are for these documents as it's being indexed, we're Coveo is going to support that. So if I'm sitting in the sales on the sales team or if I'm a customer and I try to search for something that is specific to legal that they have access to, I'm not gonna be able to see the search results. And if I can't see it in the search results, I won't be able to click on the link. Whereas if I don't perform that search and I'm authenticated in as a member of the legal team, I perform the search. Yes. I will be able to see that content, and, yes, I will be able to click into that. So it's not a matter of Coveo managing a second set of permissions to determine, do people have access to it? But rather, we're looking at the security of the document itself to make determinations if people have access to it. What we see after we went through if we did the site map and then Salesforce and pulled in all the different sources that we have within the system. What we're looking at here is the index. It's all these different con source contents that make up the index, including if of those forty seven plus out of the box connectors, if we don't have one that fit fits technology that you're looking to pull into your index. We have a what we call our generic REST API. And that generic REST API allows us allows Coveo to be able to go out to those endpoints within that API, index that content. Obviously, it's not as simple as I'm making it out to be, but, usually, you need some sort of token and authentication to get into it. You have to spell out the endpoints of where you're looking to get. But once you get there, it'll then index that information, and then that becomes searchable, and facetable. And everything else that we've looked at, it just becomes part of the the Coveo index. Now once we have the index in place, we wanna put some rules around it. And what we do, we do that through what we call our pipeline. And pipelines allow us to pull the machine learning components into the, the the experience for the search experience. And we looked at a couple different search experiences today. Right? We looked at it from the end users, from the anonymous user, and from Susan. That search experience may be different sir may be a different search experience than what the agents were looking at when we're looking at that case for Susan and being able to respond back to it. So there may be needs in your organization to have multiple pipelines. The functionality is the same. It's just a matter of what are you pulling in to make those business rules unique for that audience, the customers themselves or the agents or the partners or however you may be leveraging, Coveo. So from a pipeline perspective, we're able to pull in the different machine learning models, and I can come here and associate a model associate a model, make the selection of what that model may be. And as I make that selection into it and here we see we have the automatic relevance tuning. That's the machine learning model where it provides the result in the best order. And makes those determinations for the end user. And then the query suggestions. When I come in and type in whatever that query may be, it it helps me with some suggestions. We also talked about the event recommendations as well and the smart snippet functionality. Those are also machine learning models that can be introduced into the query pipeline. Realizing, and I mentioned this before, that machine learning is great. It makes for the different experiences for each individual user. Well, we also want to ensure that your customers are getting content that is relevant based upon your business needs. And this is where I was mentioning before about new product releases of having new content that you may wanna push out, or updates to different applications that you may have. And so the concept of featured results and within the pipeline, you get into being able to boost or bury content. And this is a case here when I come out and I you know, I'll give it a a very creative name here. I'll call it name. And so if the query is speed bit or if the query contains or if it matches it, you see it then within the drop down list of different, options that you have. If the query is speed, but we want to go ahead and add a specific piece article or a piece of content. And I'm going to spell out exactly what that may be. So I'll come up here and I'll do a speed bit. Told you I wasn't a good typer. Go ahead and make that selection here. I'm going to select this article here, and I'm gonna do this a second time. And I'll type in, heart rate, And I'll make the selection here, and I'll pull in this, we'll pull in this case here. And And now when I go ahead and edit this or add this, anytime somebody queries for speed bit now if they do if I change this to contains, if they do speed bit blaze, if they do speed bit watch, it'll then populate these two articles at the very top regardless of what the machine learning is telling. So this is ways to be able to go above and beyond what is in the, the machine learning may be mentioning. And there's other ways to be able to boost and bury based upon, you know, the create date, the update date, keyword proximity, things along those lines to make the determination that this content is much newer. It needs to be pushed up to the top. This content that may be a little bit older, we're gonna push this down because it may not be as relevant, for your for your customers. And that's the beauty of the experience that you'll have with the professional services team of Coveo to make and help you fine tune the pipelines here knowing full well that you have the ability to do a b testing for, you know, the different pipelines that you may have to see which one's working better, which one's more efficient, and then push that one that is most efficient into production if necessary. Some of the other tuning that may occur, in addition to the boosting and the bearing and the, the machine learning models is the ability to add thesaurus and stop words. So we think of stop words from, yeah, to, from the b n a words that really don't add anything to the the query itself. I've worked with customers where they've actually included some, bad words in as stop words because they didn't want somebody on their website to go in and type in a bad word, and then, god forbid, they have content that had that word in it, and then they're exposed on social media for having some of that language in there. So they've included some bad words as their stop words. Sounds kinda silly, but very effective of saying, you know what? We don't really wanna get involved in it. So what'll happen is it'll just drop that. If someone were to type in a query of the or and or one of those bad words, it'll just drop it from the equation. So we don't have to worry about, you know, being exposed that way. The thesaurus is the ability to take acronyms and being able to do this the entire spelling of what I've mentioned, automatic relevance, tuning, a r t, at least early on to determine what does ART mean spelled out and allow the system to develop that connective tissue. Over time, machine learning is gonna pick up on it. But early out of the box, it takes about a thousand queries before the machine learning really starts to provide some valuable results, and the thesaurus is helpful. It's also helpful if you have people coming out, and analytics will help with this, typing in information about your competitors. Well, if they're typing in information about your competitors and they don't see anything coming up, they may go back to their competitor. So what you can do is add a thesaurus to send them here of your competitor's product or competitor's name to the the what your equivalent, product name may be within, within the pipeline that you may have. So there's a lot at play here. Yeah. Once we start talking about the index once we get the index, that's great. But then what do we want that end experience to look like for your customers, agents? And that's where the query pipelines come into play. Finishing up, let's talk about analytics. Right? We saw a little bit of the analytics when we're, clicking on that little vignette that we saw over on the right hand side of the page. We saw a little bit of the analytics when we're in the agent insight panel providing what has Susan done in the past? What has she clicked on? What has she done? Well, at the very base level, the micro level from a reporting perspective, we're capturing every step of what your users are doing, what your visitors are doing. And we see that we have, you know, the date stamp, the visit duration. If they've authenticated into the system, we see, their name. Otherwise, it may be anonymized or anonymous. Generally, their location, the technology they're using, and then this this number here, the event count. What does the event count mean? Well, those are the actions that they've taken within the within Coveo. So we take a look the same way we saw it before within that little, vignette up in the top right hand corner, the search, the click, the facet value changes, looking at a preview, a quick view of it. We capture all this information to see that, yes, Richards or Wallace, I guess, did a query for pair my phone. How do I pair with an iPhone eleven? He clicked on this specific document. In fact, let's go a little bit deeper. For the different queries, well, when he asked about pairing my phone, how long did it take to respond? How many results came back as a result to that? And then when we take a look at the document side, it's all this rich information to be able to see what the language of that document that they clicked on, who the author may be, and where that document may have appeared on the page, the search results page. All this information at a very micro level then bubbles up to the analytics side of things from a reporting perspective. And so people probably are not going to run a report on the the information that resides within the visit browser on an individual by individual basis. You could, but probably not. What most people are most interested in from a reporting perspective is the ability to look at a report, and our reports here are all templatized. They come out of the box. You can make the changes to the templates or you can, you know, create yours net new if you wanted to as well. But from the examples that we're gonna be looking at here today, you know, getting at a very high level, well, where are people coming from? What are the top devices that we're looking that they're using? And this is information that we saw within the visit browser of, you know, geographically where they are and are using, you know, Chrome versus Safari. Let's dive down a little bit more because now we're able to see not only the visit counts, you know, the accumulation over this period of time you know, about a month here, that we've had a hundred and eighty thousand visitors, but then also take a look at queries that have been asked. And, actually, not only queries that have been asked, but also that have involved a specific click. And then when they click on that, those specific content sources or the specific articles, what are the the sources? What are the most popular sources? Is it a knowledge base? Is it a your website? Is it information that may Coveo from a community? And measuring that information. But then taking that a step further, and this is kind of the pat yourself on the back moment. What are the most popular searches that you have? And what are the most popular documents that have been clicked? And allows you to look at where things have been going well, and how can we replicate this and, you know, provide additional information for the customers on this these specific topics to be able to build upon that if necessary. You'll notice at the very top that we have different tabs as well. And this just becomes another way to organize the reports for you. Each one of these different index cards can be moved around and changed, can be filtered if necessary or clicked on to be able to look at additional information. So if I wanted to look a little bit more, a little bit deeper into one of these lines here, I click on the line and then filters, you'll see that the filter up here had changed as well. And now we're looking at only that information that I had clicked on. If I wanted to further filter it based upon the day or the week, I can then go ahead and and do that. Now we talk about how can we provide a good experience, and we saw the information within the reports of the top documents, the top sources. And if I was putting together the re this report, I probably would have the the the most requested documents or the popular documents, top popular queries. And then on the other side of the the ledger, I would have the queries with no results and keywords with no results. Because this then becomes an opportunity for those folks that are looking at the analytics and the reports here to go to your content creators and say, please create this content for us. We do not have content to be able to serve our customers or whoever the audience may be. And, ultimately, what's happening potentially is that they're going elsewhere to find their answer. They're going out to Google to find their answer, or, god forbid, they go out to your competitors to find an answer to question that you should have. Or they just give up, and then they become a a frustrated frustrated customer where they reach out over the phone to try to find that information or an email to find that information. So we wanna give visibility to both the good side of things and the bad side of things. We didn't talk much about chatbot or live chat, but that functionality can also be incorporated. The index can also be incorporated into both of those as well. As you see, we have some chatbot metrics that we're able to pull in, but also giving visibility on the agent side of things as well. To look at the productivity of the agents and how they're serving their their, their customers through the cases that they have. And looking at the activity that they have, the click through rates that they have, the different searches, the different tabs, and how they've interacted with the product itself when they're going back to the customers. What documents are are some of the most top documents as we saw before that are unique just to the agent experience? But then let's go up another level with that. And if we're taking a look at the in the from an agent's perspective, let's look at a report that'll actually give you an idea from a case deflection of how much money you're actually saving by deflecting cases here. Looking at you know, from our perspective, we consider a deflected case when somebody performs a search, somebody clicks on a document, and then that's good enough. Now you may have different rules that we'll be able to incorporate into what a deflected case may be. But for this case in this example that we're looking at here, we can see the total number of cases that are created, the numbers that we deem to be deflected. And then of that, if we put a dollar amount to the cost of one of your agents, either what your l one, or t one may be or all of them together, get an idea of how much has been, deflected or that you avoided and that you actually saved here. In this case, you're at twenty five dollars a case. So here we're looking at some information from a case deflection as far as by country, by product, by deflection outcomes. And so, really, you're at your your imagination's end as far as what type of information, how you wanna measure the success of your case deflection and the work of your of your agents. So I'm gonna Maggie, I'm gonna go ahead and, let's open it up for any questions you that we may have. I did see that there was a question or two that may have come over, in the chat here, so I'll handle this one first. The cost the question was, well, where does the customer say that this is not helpful? So you know quickly you're on the wrong path. So, typically, when we talk about the the use of a thumbs up, thumbs down, a star system like that, one, customers typically will not engage in that. I know firsthand that I I usually don't when I see those. It's just I read what I need to consume, and that's it. I'm not gonna go the extra mile and and thumbs up, thumbs down on something. So we have we found that that really isn't an effective way to do it. The most effective way for us to be able to determine relevancy on a specific article is did they move on to the next article, or do they move on to the third article, or they ask another query that's somewhat related. And then that allows us from a machine learning perspective to make the determination that this specific article was good because they didn't ask another question that was similar, nor did they go on to the next option or look for click on additional, additional article. So we would determine that. And then what happens is from a scoring perspective, every article that we have within the index, within your content, is scored. And so in that case, you're the scoring for that would then be adjusted upward. If they did go on and click on another article or and and, you know, two or three other articles, we will continue to move that scoring down for that article. But then the last one that they did, we would move that one up because we deemed it to be a success. Hopefully, that was a a hopefully, I understood the question that you're asking now. I hope so. I think so. I mean, it's so impressive what this machine learning can do. Maggie. I don't think we can can hear you. Nope. That was my volume. Sorry about that. That's okay. I was just saying how interesting and awesome it is what Coveo can do with the machine learning to do these minute small details, but it impacts the customer's experience so much. It yeah. It really is. It's somebody that's been in software for for quite some time, and I'm a bit of a techno geek. I'm always spending time online at night, and if I can find the answer quickly, it's like, wow. Great. This is so much nicer because it allows me to get on to my next project a little little bit quicker. And so seconds that add up to minutes, that add up to hours over the week. Yep. That's so true. I have one question, and we'll wait to see if any more come in. And people if you have any questions, put them in now. But how long is the implementation, and how quickly will the machine learning start learning? So let let's start with the the machine learning first. The machine learning, it it all every all few questions were based upon the amount of content that you have, that you're looking to index. Typically, a machine learning model will take about a thousand at bats or a thousand queries, let's say, to start to be relevant. Now that's not to say after the twenty fifth query, it may start providing if we think about the, query suggestions, you know, the at the very top, will you pull on this here for within so within the query suggestion is this drop down list that we're looking at here. You may see results within twenty five queries, but it may not be the desired results that you're looking for. After about a thousand, you'll start saying, alright. This is this is Intune and on point. And there's ways to to help that along. But if we just started net new, it's probably about a thousand queries to to get there. And then the the first question was about implementation. A lot of it depends upon what you're looking to do, what how much content that you have, where does that content reside, how fast like I said, I've been doing this for a while, not just at Coveo, but it's how fast can the customer move compared to the, you know, in this case, your Coveo, to be able to do. But we're looking at typically about three months. Some of the more simple implementations, maybe two months, and some of the longer ones where you're doing things across multiple, multiple continents, multiple, different websites. And if you're pulling in your cases and if you're pulling in your intranet and if you're pulling in service, it it could add to the complexity of or length of the project, not the complexity, but the length of the project. Perfect. Good to know. Doesn't look like we had any more questions come in, but, Kevin, this has been so insightful. Thank you so much for doing this. Thank you everyone for joining us today. As I said at the beginning of the call that I'll email this to you a recording. And if you have any questions that you think of in the meantime, please just reply back to that email. And with that, I hope everyone has a great day. Great. Thank you, everybody.