Hello, and welcome to the service edition of new in Coveo, where we'll be covering our latest product updates for Coveo service solutions. My name is Bonnie Chase. I'm the senior director of product marketing for our service solutions. And today, I'm joined by our product manager, David Atala. Welcome, David. Hello. We get started, a few housekeeping items. This session is being recorded, and you will get a copy of the recording within twenty four hours after the conclusion of the event. And we will be answering questions today. We'll try to answer as many as we can during this session, but, of course, we'll be answering some at the end. So please be sure to drop in your questions in the q and a portion at the bottom of your screen. Now we will be sharing some product updates today. So I do want to, remind you of our disclaimer. So please note that this may contain some forward looking statements, industry and market data. And as a publicly traded company, we do urge you to refer to the disclaimer page before making any purchasing decisions. Series that recently aired. So if you missed it, you can go to our website to watch it on demand as well as register for an upcoming session about AI strategy with, the former head of GTM at OpenAI. And then finally, some additional sessions for our other lines of business will be taking place over the next couple of weeks, so please be sure to register and attend. Alright. So let's get started talking about service. Now when we look at our service vision and what we're trying to achieve with service, you know, it really is ultimately about delivering the best service experiences across the entire customer journey through data and AI. And there are really three things we're trying to achieve here, which you'll see our product updates align with. The first is making sure that we're creating those effortless customer experiences. Now this could be through, you know, the connectivity. It can be through the UI itself, but really about reducing that customer effort. The second piece is really about ensuring that we have autonomous admins. We wanna reduce any dependency on IT and developers so that our admins, can take things into their own hands and make those changes that they need to make to improve the experience faster. And then finally, something we're very excited about is our no code Gen AI. So what we'll be sharing with you today is, you know, what are those updates we're making with generative answering across the entire customer journey, and how can that really happen with no code, which David will share later on. So with that, let's start talking about some of our product updates. I'll give you a high level overview of some of the things that we'll be talking about today, and then I'll pass it over to David to to get started. Now we've kind of bucketed some of these capabilities into these these five different areas. The first one is around connectivity and integration. So you'll be able to see some of the improvements that we've made to the connectors that we have. Now we have hundreds of connectors available to for you to index content from across the enterprise. We're continuously maintaining and updating those. Search and UI, of course, some improvements that were made to the building experience as well as the display experience. Machine learning, we're going to be sharing a lot about relevance generative answering, including some of the latest updates and what's to come in the future road map. Data and reporting, and platform experience. So these are some high level things we'll be covering today. Now, David, if you could just take us through connectivity and integration. Yeah. Thank you, Bonnie. So, obviously, Coveo we at Coveo, we invested a lot in ML and generative AI, but we also invested a lot in our foundational capabilities, including our connectors. And I'll kick off with our pipeline online connector where we've enabled, our users now to be able to set up and configure, the SharePoint online connector, directly within the admin. In the past, it used to be done through, adjacent config, but now you can do it directly in the admin. And with that, you'll have the ability to filter out content by date and time period, ability to exclude content based on conditions, and as you can see, a lot more DD user profiles and many others. But we've also added some other updates to some other key connectors that we have, including our sitemap connector and our web connector. For these connectors, we've essentially improved, the we've added a better UI, and you can now simply look at common CSS exclusion filter, where we define and modify some of them whenever you want. We've provided some advanced to advance users the ability to easily apply and multiply web scraping. And then finally, we've additionally added a web scraper helper Chrome extension, that will help you, test web scraping configurations. And that's kind of the essential part of our of our updates on our connectors. We are looking to continue and update these as we move forward. And, yeah, moving on to our, integrations. Some of the key updates we've provided to our our integration is to our Salesforce integration. We've, added the ability to now leverage our hosted search page builder, which I'll go deeper into later on, to be able to set up experience silent service cloud search search pages completely out of the box, no code. We've also added the ability to effortlessly add, Gen AI to your, Salesforce insight panel. Again, completely out of the box and no code, and also improve our quantum components for, search capabilities as well. Moving on to platform experience, what we have done and what the team has been has worked on across the last few quarters is what we call the projects feature. The projects feature allows you to combine multiple resources, within the Kubera admin console and combine them into a project. We've heard from our customers in the past that they could have up to fifty, sixty, seventy sources, multiple reports, multiple query pipelines. With the projects feature, you're able to combine these into a single project and name it at at your will. And this will allow you to control a lot better, the projects or the features that you have live with Coveo, and as you add more to your capabilities. Search and UI. Within Search and UI, you've heard you might have heard us talk a lot about our Coveo builder across the last, year. In twenty twenty three, in January twenty twenty three, we've released, the hosted inside panel, builder. We've then added the in product experience builder in April, of last year. And then last year, we've asked we added the hosted search page builder, as of, December of, twenty twenty three. What the builder allows you to do ultimately is, to be able to set up, as an example, hosted search pages, completely no code out of the box, and without the need of a developer. You get access to the latest and greatest capabilities and, obviously, to all of our machine learning capabilities as well, at a click of a button. With that, we've also provided update to our in product experience, which has now the ability to leverage version history. And you can now automatically create pipelines and machine learning as you go through your creation process of an IPX. We've lastly enabled, relevance center answering to the IPX as of late twenty twenty to late twenty twenty three as well. And now I'll pass it on to Bonnie who will dive a bit deeper into machine learning before we go into a little bit of a demo of what we're talking about. Morning. Much, David. So sounds like some really exciting things coming out with improvements to the builder and our end product experience. You know, ultimately, all of this is is still, affected by the machine learning capabilities that we're coming out with. So, of course, relevance generative answering is top of mind for us and one of the things that we're most excited about. So when we talk about relevance generated answering, you know, I think most of you are familiar with the typical kind of search and and results experience. You've started becoming more familiar with generated answers being added to the search experience, whether it's on your website or Internet, you know, within a CRM, within a full search, application as well. And now we wanna make sure that you're able to extend that within a bot. So whether that's a chatbot, an insight panel, in product experience, we want to make sure that you're having a consistent generative answering experience across all of your touch points. And one of the things that we talk about is, you know, sometimes people get excited, and they want to implement new technology, and they wanna add you know, let's say, I wanna add a generative answering capability quickly on my website. Well, you wanna make sure that for your customers, that experience is seamless across all of your digital touch points. So what better way to do that than to implement it with a checkbox? So you implement it once with Coveo. You you can easily extend it across to all of your other sites, applications, and product experiences, what have you. So that's what we're really excited about. Now today, what we're sharing is generative answering is now available in your agent experience. We went GA with generative answering for self-service at the end of last year, and now we want you to be able to easily add that into your agent experience. So this is now available within Coveo's hosted insight panel. You're able to empower your agents to get answers with a click, and really overall enhance your service management application. So you're able to integrate GenAI not only in the self-service experience, but also in the agent experience. And, of course, all of that context gets carried over. We talk about generative answering in product and in app. So, again, you know, the same great generative answering capabilities, accessing all of your enterprise wide content, applying all of that security to avoid hallucinations and make sure that the right people have access to the right information, and then generating that relevant answer right within, within those apps, for you. And then finally, you know, if you have a commerce experience, you can still bring that relevance generative answering experience into that guided shopping experience as well. This really does close the loop when you think about, you know, from a commerce experience where you're shopping and buying products. People have questions. They wanna make sure that they they're choosing the right thing, and and that's really what this helps you achieve is, you you know, improving those conversions by educating, your customers as they're, going through that shopping experience. So we've covered quite a bit on slides, but I think what we really wanna showcase to you today is a demo. So we'll give you a a quick show and tell, really showing what we're able to do with this generative answering experience. So I'll go ahead and stop my share, and, David, you may, take over and show us the product. Awesome. Thank you, Bonnie. And, hopefully, you can see my screen. Yes. Fantastic. So as Bonnie just mentioned, we went through those slides quite quickly because we really wanted to deep dive a bit into the product and show you those capabilities, directly within our product. Some of the updates provided around in product experience, were quite were quite innovative for us, and we're really proud of having adding those to our, to our offering. First of all, I will say that within the Kava admin console, we ourselves leverage, an IPX, which is what you're witnessing here, and we also leverage, Caveo generative answering within that specific experience. And it works completely out of the box and no code. And the way that works, and I'll show you that very quickly, is I can add an end product experience, as I go. And you have now the ability to compete there and automatically set up those query pipelines and machine learning models that are specific to Coveo, and you you might not have the experience with or usually would have needed the help of a developer to set up. And all I then need to do is click ads. And through that experience, I'll be redirected directly to my specific inside panel, APX builder. Within there, I can start configuring it as I need to. I can start adding specific result actions to it. I can start, tailoring specific components and change the theme if I want to. Completely out of the box, I can make it as red as I want or as blue as I want. You can start making it it will almost start looking like a Christmas tree if I continue like that. And, ultimately, all I will then need to do is click save. But what I can also do is in my settings, I can straight away, set up a RGA model if I wanted to. As long as the right model is set up and the grid pipeline is set up, within, your Azure admin console, which is something our professional services team can help you do. You can simply click that button, and, ultimately, you're ready to go, and I can click save. When I apply those changes, these are automatically and directly set up. And then I can then go into my version history, and I can go and set it up as I want to, and I can call it, I don't know, demo or webinar. And I'm done. This versionist review will help me, audit moving forward. So if I ever make changes to my Apex, I'll be able to see that. And it would also help you restore any versions that you would wanna restore to as well. But that's not it. So that's the first touch point for IPX. But as Bonnie just mentioned, we also have, CRJ available within our search pages and and also our hosted search pages. And the first example of that is our own docs page. Our public docs pages have access to generative, generative answering. And you can directly see if I ask it how to set up a query pipeline, it will give me an answer. Now that view might not best be the best view for me right now. I might wanna see it in the list view so I can just have a step by step process of how to set up a group pipeline. And that answer will be generated, as expected. I can then use that specific information to go ahead and set up my grade pipeline. I can also see the citations that have been, that have been used to generate that specific answer as well. With a and, however, you can see kind of a little summary of that specific, citation. With that, I can then thumbs up or thumbs down the center, or I can even copy it to clipboard and share it with whomever I want. Now as Bonnie just mentioned, we've just, added our, our RJ capability to our agent console as well, our agent inside panel. And you can see now the inside panel takes into context that we have a code error twenty two, in our barcode skipper demo property, which is a boat, parts, kind of company. And I can see that generates an answer, specific to, the code error twenty two. So this will enable the agent to ultimately help and answer, your problems as fast as they can and ultimately reduce the time to, time to time to close those specific those specific tickets. I can they can also move ahead and ask it how to upgrade an antenna, an antenna, and the model will go ahead and provide me that answer on how to upgrade a specific antenna. So the agent at the tip of their finger can also copy to clipboard that answer and submit it, to their specific ticket. We do mention that we do wanna caveat that agents will always have to review those answers before showing them, because, obviously, some, answers might need a bit of Coveo oversight as as you move forward. And to show you how easy it is to set this up, I have a demo, a demo agent inside panel that I've created on on specific dev environment. And in there, what I can do is, leveraging our specific our preview capabilities, within the within the Coveo builder, I can try and see if, hey. Why not while whilst I'm setting up RJ, is it actually working? So this specific demo leverages, some content about wood, and specifically about, band solvo. And I can see that, when I type band saw ball, I don't get any answer. I do get the result of how to turn a band saw ball from, from from a board, but there's no really, no generators capabilities activated. Well, if I go to my settings and, again, bearing that I have the clear pipeline set up with the right relative generative answering models, by simply clicking this button and clicking save, and I apply the changes. Straightaway by just typing this up and typing pencil bulb, you can see that now the the answer is being generated. And this is directly mimicked in your Salesforce, environment as well. So if I go back and remove it once again and I go back to, my Salesforce environment where, this specific insight panel is set up, you can see again if I type Bansalvo, no answer is being generated, and I go back and I activate my RGA, model, again, as as quickly as that, I just click refresh. And whilst the inside panel loads, I can just type in bandsaw ball, and the model will start, generating an answer. So it's as quickly as easy and that to set up, and it's really at the tip of your finger to be able to go and set this up, for, for for yourselves and for your agents and, again, to your into your search pages or your, end products experiences. I'll take I'll take the lead now to share my screen with the remaining slides of our specific presentation. If you have any questions, please put them in the chat. I know I went to that demo quite quickly. And, again, I will reiterate. If you do wanna test some of those capabilities, you can go to docs dot co dot com and have a look through that. So if I go full screen here, I'll go through what's coming soon as well for our machine learning capabilities. So it's great that we have RGA within all of our touch points. But RGA is just the first component. It's just the tip of the iceberg for us. We started to look and the team has started to look into, what we call follow-up questions, which is ultimately, looking into generating questions to what's, to your to your to your problem and kind of following this this process of creating conversation with this with the generative answering component to go deeper into a specific problem. So let's say in this scenario, you're asking what are the benefits of using smart snippets. Well, the model will auto generate upcoming questions that can help you continue your setup for smart snippets. So let's say you're having issues with your query pipelines or whatnot. Well, you can start at the top of the funnel and then at a really large scale question and then start narrowing this down. This will ultimately help also, some some users that could have difficulties in prompting those, those LLMs. So, obviously, prompting an LLM, you cannot just ask it one single word. You need to kind of set a specific kind of questions and a specific way to prompt it. And with that, you can help narrow that conversation moving forward and as you go deeper. But that's not that's and that's not just it. There's also the ability to also have, a conversational experience where you could, again, ask it a question. The model will generate an answer. And then with that, you'll have the follow-up question being generated, but you can also go and go deeper. So it will mimic almost, you know, when you're on ChargeGPT and you have that experience where it's all contextualized into your specific conversation, well, this is what we would offer in that specific component where that conversation is contextualized to the information you're looking for and you're trying to get out of. And you could imagine this experience being leveraged within search pages, within an app, and you can also be seem to be leveraged by your agents where they're looking to go deeper into this conversation. Ultimately, we believe that that's kind of this kind of experience is the future for for, for our RJ component. And with that, it will help ultimately reduce the the time to resolution for agent and reduce and improve the self-service experience as a whole for your customers. Now I will caveat one thing for the follow-up questions and the conversation experience. These are still, at the moment, in development on our side and in testing. So we are, they're they're not ready they're not ready to go GA. As you can see, there's an early access call tag at the top. We are collaborating with some customers to try and help us define the solution. You might have seen the demo that, Laurent has made at our at at at r three sixty as well. So it's still early doors, but we're hoping that in the next few quarters, we'll make some significant progress in this and provide and add this to our to our to our GA capabilities. Also upcoming within, the RJ component itself is rich text formatting. So today, as you saw in some of the answers that are generated in the demo, the the text today is a bit is a bit simple. It's just a block of text. It could manage bullet points and stuff like that in list, but it's not really, that advanced. So what we're looking to add to that is is rich text formatting, where it will be able to manage Coveo snippets. It could manage tables. It will have titles. It will and this is where the model will automatically identify how to structure that answer to make it more visually appealing and easier to read as well as you kind of try and go and leverage the component itself. This is, this is a capability we're looking to compete in q two, and should be ready, and we'll make a further announcements when GA. Lastly, for, general answering, we are we have been we have heard the feedback of our customers, and we've heard a lot about the ability to support multilingual. So as you know, Cabello is a product Quebecing company, so and we have a lot of French speaking customers. And so the first multilingual capability we will try and offer is French. So in q two, we are aiming to, have, the our first few customers trial this, and trial the ability for for French support. But we'll also heard your feedback for other languages like Spanish, Italian, German, and a lot more. So these other languages might come later in the year. Our first focus is French for the upcoming quarter. But we are and this is but it's something we are working on consistently, and we will continue on add, more lang more languages as we move forward. With that, I'll know that you have access to additional resources for the for for new in Coveo where you'll be able to end level up, look back in some of the sessions, and have to go through those learning modules, and go through the new incovell if you want wanna rewatch any of those webinars in the new incovell page where you'll have access to all the resources that we've just mentioned. Bonnie, I'll pass it back to you. Awesome. Thank you so much, David, with a great overview of what we have and what's to come. Just as a reminder, we do have more sessions coming for website, ecommerce, and workplace within the next couple of weeks, so be sure to go and register there. Now we did get quite a few questions, so I wanna go ahead and start digging into some of these. So, the first question is around quantifying the average accuracy of your generative answers. So that's something I can actually speak to. So, you know, as we're developing answers for, generating these answers, one of the things that we wanna make sure is that we're not generating, inaccurate hallucinations. We do have thresholds to make sure we're not spitting out an answer if we're not confident in that answer. Now when it comes to being able to quantify the accuracy, you know, that's going to vary depending on a bunch of different things. Right? So that depends on the context. It depends on the question that's asked. Depends on how it's asked, the type of question that it is. But we are looking at establishing a baseline of, you know, getting to an eighty percent accuracy rate. We've seen that with our customers who've implemented it. Again, it's not a, a one number that it sticks to at all times. There are a lot of things that, go into that number, but that's what we're looking at. For us, accuracy is is more important than getting just putting an answer out there. So you will see that we are are, setting those high standards for accuracy for those answers. Let's see. If we are interested in French language, who do we contact? I would say for that, that's a great question. We are, you know, right now, it is English, but we are looking Yeah. For now, just reach out to your customer success manager. We are, again, at early early doors for for the live French language in q two, but, obviously, as we move forward, we we will look to add a lot more customers. So please reach out to your to your CSM, and then, we'll be able to engage in conversations on that front. Awesome. Alright. Another question. Can you speak about how to report on the success of the generated answers, especially when there is no content for the user to click on? Obviously, this is an industry wide challenge, but I'm curious as to Coveo's approach. This is a great question and something that our business values team has been working really diligently on establishing. There are a couple of things that we look at when we're looking at measuring the, the success of the generated answers, and a lot of this takes place in that that testing portion. So when you're evaluating vendors and you're, you're looking at what, how do you how do you rate success. Right? Well, some of what we've done is comparing, implementations with GenAI and implementations without, gen Part of it is looking at the success of Gen AI versus no Gen AI. So are they are they going and and creating a ticket after doing a search? Are they spending less time searching? These are a few of the things that we're looking at. But, again, you know, our business values team is is ready to jump on board, help you set those benchmarks for what that would be for your organization and those key metrics that are are are, focused for you and your objectives. Let's see here. We do have a question about the, the hallucinations. So how are you preventing hallucinations within your generated answer? I don't know. David? Yeah. Absolutely. So so what we do today, and this is something that that that's critical is we focus on and all of our are grounded. So, essentially, what we what we ensure specifically indexed, within Coveo, and then we, make sure you create a clear pipeline with the specific content you would like to, you would like the LLM to learn from. And so, ultimately, by grounding, the LLM to a specific set of content that you've selected with specific, the specific type of information, we ensure that the model only learns from your specific content that's specific to you and learns to the specific information you've provided. What it ultimately means is that it will not learn from anything from the web. It will not learn from, from any other, quote, unquote, junk that could be out there. And so it will reduce the hallucination rate quite significantly as well as as you try and go live with the specific solution. Alright. And we have time for one more question. Do you have any suggestions for a company, that we should work with to help structure documentation creation in more of a template form, so that data, so that Jenny and I can have better data fed into it. So we can definitely share some best practices for you for creating that cleaner dataset. You know, shorter documents work better. So keeping in mind, you know, you wanna make sure that each document is maybe answering one question, and the shorter, the better. Because as that, as that model is reading through that content, you know, the the the first, you know, know, twenty five pages, I would say, of a document are considered. But so we really wanna keep those documents as short as possible. Less is more. You really wanna make sure that as you're taking into consideration what you're including in your generative answering model, the the content that you're included. Make sure that you're keeping in mind, you know, the audience that you're trying to reach, what type of content will be relevant to them, what type of content would, make sense to have those answers generated. You don't wanna just throw everything in there. And then another key tip that I would add is, you know, as knowledge managers, we're used to adding metadata and and, doing all of that. For generative answering, the body is really what's important. So making sure that those keywords are in the body of the text, not just in the metadata, You know, making sure those headers, butters body is really clean, is really important. Alright. Well, that is all the time that we have for today. Thank you so much for joining our new and Coveo service session. Now as as I've said a couple of times, we do have more sessions, coming up, so please register and join those. And for those of you who didn't receive an answer to your question, we will follow-up with you after this webinar. Thank you all so much, and have a great rest of your day. Thank you.