Hello, everyone. Thank you so much for joining our webinar today. Going beyond the hype enterprise ready GenAI for industry leaders. My name is Bonnie, and I lead product marketing for Coveo's service line of business. And I'll be your moderator today. I'm joined today by our director of R&D, Vince Bernard, who'll be walking you through a product demo. Now I have a couple of housekeeping items to cover quickly before we get started. First, everyone is in listen only mode. However, we do want to hear from you during today's presentation. So we'll be answering questions at the end of the session. Please feel free to send those questions along using the q and a section at the bottom of your screen. Today's webinar is being recorded and you will receive a copy of the presentation within twenty four hours of the conclusion of the event. Now as we get started, please note that presentation may contain some forward looking statements, industry, and market data. So please review this disclaimer before you make any purchasing decisions. Now the agenda for today, it's simple. In the next thirty minutes, I'll give you a quick overview of Coveo, and our generative AI continues to shift customer expectations. And then Vincent will introduce you to the new Coveo relevance generative answering solution with a demo. And then, of course, we'll take questions at the end. Now for those of you who may not be familiar with us, Coveo has been building enterprise-ready AI for over ten years. And today, we deliver that at scale for around seven hundred leading brands around the world with the most trusted and highest levels of compliance and security. Now we do this in one single platform that connects an entire enterprise's experiences across commerce, service, website, and workplace. Creating more positive customer and employee experiences, and in turn creating transformational business outcomes And we do this through semantic search, AI recommendations, generative answering, and unified personalization. Now, our strategy when it comes to GenAI is really last to hype first to results. So we've been working really hard to build a generative answering solution that meets your business needs and your customer expectations. And over the years, we've really seen how large technology disruptors like Netflix, Amazon and Spotify really shifted customer expectations and redefined the ways that businesses interact with people. And the launch of OpenAI's GPT is no exception to this. You know, we've seen everyone from large enterprise companies to everyday people you know, playing around with chat GPT. And generative AI is really the common experience game trade game changer, which is really driving the need for enterprises to move quickly to start leveraging this technology so they can deliver on these expectations and not get left behind. So when we talk about these expectations, you know, what are they and and what what is it changed to? And Yes. It has evolved as technology has evolved, and we think these expectations have changed for good. People want experiences that are aligned around a unique individualized journeys designed just for them. They want experiences that are prescriptive and intent driven. So they don't have to start their journey from scratch every time. And they want journeys that are coherent and don't differ depending on where you search for an answer. So at the end of the day, the experience is what gives today's companies a competitive advantage. So when thinking about leveraging generative AI and generative answering in particular, it's important to remember that this is something that should be seamlessly threaded into your digital experience, not something bolted on the side. Now as we've been looking at generative answering solutions and and really, you know, having conversations around the difference between generative answering and search, you know, we have to remember that this is not going to be a separate siloed interaction channel. The worlds of intelligent search, generative answering, conversation, chat, all of these interactions are converging into one. You want one interaction placed for customers that will offer them answers along with content. The experience needs to be personalized and guided and have a single source of truth, and of course it needs to be secure and trusted. And that's really what we're looking at as we're evolving our platform. Now as I mentioned, we've been building AI for over a decade, and generative answering is really the latest step that's part of the natural evolution of our journey. So again, not something to think about in a separate silo or as a bolt on. We're very excited about what the future holds with this new technology and and our ultimate goal is to help enterprises solve key challenges that are found with platforms such as GPT. So what are those challenges? Well, let's take a moment to walk through them. The first one is security. So this is really about ensuring that you're respecting the permissions of the content and the privacy of user data. Accuracy is key. You know, ensuring the content used in answers is current, the answers provided are factual in the system that's not hallucinating. You also wanna be able to trace the answers back to the source of truth. When it comes to content, you know, we know that enterprise have multiple sources of content and a variety of formats from structured to unstructured. And we also know that being able to connect to the right content can really improve the answers and the value from generative answering. Cost is a big thing. Right? GenAI experiences can be a hundred times more expensive if not engineered right, which is why we're taking an approach that will be inexpensive compared to a build it yourself model. Finally, the experience. Now this is really ensuring the relevancy for the user, unified experiences across the touch points and combining that search and answering for disambiguation within that experience. And at the end of the day, it's it's about trust, right, for both users and the business. You wanna be able to trust the technology and you want your customers to be able to trust you. So we've designed a solution. We've been working with over thirty customers on our design partner program to build a generative answering solution that can be embedded into your digital experiences. And we're excited to announce that this will become generally available on December fifteenth We already have customers using this live and seeing results. And we're very excited to to to share with you today what this looks like. So I'll pass it now over to Vincent who will be giving us an overview of our generative answering solution. Vincent? Thanks, Bonnie. So welcome, everybody. I'm Vincent Director in R and D. I'm very happy today to walk you through the technical side of things. I'll also run a few demos. And surprise today, we are going a bit further than the previous, iteration of this webinar, we're gonna go into commerce. You'll see why, and you'll see also, some very interesting ideas we have on the So why Coveo generated answering a few things that we realized is that throughout this year with the tsunami created GPT and all these LLMs, people started to interact in a different way with content. In service, for instance, we we get longer questions, questions that are more human made. Instead of going straight with keyword search, we see more and more people being more structured, and ask real questions such as how to add a new bank account, or my robot's not following my path, I have a roomba here, or another example, what's the difference between atomic and headless. If you don't know atomic and at least are the UI framework to be was building. So this is a query we've seen in our own analytics. So first off, we realize that the user behavior is changing. People are expecting to get more proper answers, regarding these rich questions It's not just keyword search anymore. We've also seen that in commerce. So you see here, and gentlemen, I'm a student entering law school. I'm expecting to travel a lot. I'm on tight budget. What's the right laptop for me? So we're expecting these relevance platforms to be able to parse these long queries. Identify over here what's, tight budget means low price and then traveling a lot, for instance, it means long battery life or or lightweight. So translate these, human made languages to something that is more attribute in commerce. Or, again, here, I'm planning for a fishing trip in Florida. What's the optimal equipment for beginners. So all these new behavior we see on the market. Lastly, same thing on the workplace where we we see people asking more and more things like how to connect my laptop to the TV office in London. This is my vehicle colleagues. I'm not really trying to connect in London, so, it's fun to see them in action. So long story short here, we see behavior shift. To, illustrate the next steps of this demonstration, we'll talk quickly about what we have today. AI power surge is what Coveo has been delivering on the market for years. We're talking here about predictive query suggestion. So when you go on Google or on a video powered experience and you start typing something on a search box, you'll see these different words that will we call it type ahead. So they're trying to understand where you're going with your query and and try to suggest your relevant path, for success. And then when you reach a search page, you have a lot of different things meaning multiple data sources, some some some filter navigation, a list of results that is ordered by relevance. So you gotta have, for instance, top results will be better. So that's the classic I'd say experience that we have today. Then the new generative answering generation is a little bit different. You see that instead of being a search box, it's more it's more wider. It's more like an input prompt, and people will will ask question in natural language. And then it's gonna stream or generate that answer. It is a little bit more complex than just search results. So those are two different experiences for sure. I think everyone agrees. How can you merge them? How should you merge them? What's the truth about it? So what we see right now is this. So could be a search on one end is a stack build like this. You're gonna have at the bottom some some secure connectors, the connectors we use to reach your data as an enterprise. So we have native connectors from a variety of systems. You can see here, Salesforce, like, or it'll be etcetera. So we take that content, and then we put it on the video index, which is unified And then on top of it, we bolt machine learning relevance. And this is how we get you a very good step. Then when you talk to clients or cross fix or anyone out there, they will try to build this, and those are the words you use a lot these days. So extracting the content, adding embeddings semantic search, having a vector database being able to throw that to an LLM. So those are all the buzzwords. And we think this is not a solution because on one end, you get your search stack, which you already have, which you capitalize on for many use cases. And on the other end, you have your LLM stack And then you get duplication of content. For instance, you get different search boxes. You get two infrastructure to support. You get separate administration interfaces, different set of facts. So it's I mean, twice the cost, basically. What we came up with is actually this way to structure sure of it. So we kept our very relevant stack, but we decided to bolt on top of it the right part of the equation. So now with our content, were able to do embeddings and extractions of snippets straight at the index level. We're able to store these vectors and then after with the results that are powered with machine learning, we're grounding the LLM based on this. So this powerful architecture is actually what makes us leader on the market. We're the first to deploy these interfaces at large for self-service experiences, for instance. Because it's a bolt on extension to Coveo, and Coveo is a very scalable and easy to deploy platform. So we think we really have something good, a good recipe. I'll show you in a few minutes. Basically, what it gives you is a grounded experience where you have your fresh content that is extracted from our connectors in the index you have security. So each user let's imagine you deploy that in the workplace. You're gonna have, like, a generative answer based on the content you're entitled to see. Which is very good. And then we have also administration analytics. So it's a very complete stack right now. You're gonna be able to serve unified queries, unified results from all queries, which means that your LLM, your generative approach will have the same set of facts as your search engine or your chatbot or your IPX and product experience. So, it's gonna generate answers based on the most relevant part of the documents you have index it's personalized and then it's a very strong protection against a hallucination. So this is a stack. This is how we built it. And now I told you in the beginning, we're gonna talk a little bit more about commerce. In commerce, what we see, so this is a question generative answering feature. So when does people have question, usually at the beginning of their journey. So this is a quick map of the Coveo features and when they are in the lifecycle of a client, and where we see generative AI is at the very, very early phase of the cycle, at the beginning of the funnel, when the client has questions regarding, I don't know, this is this guitar better than this one? Should I get a fender or a last bolt? So these questions where you want to talk usually to a human, nowadays in the e commerce world, there's no, not much humans. So it's it's fun to be able to talk to something that has good facts that is grounded to the content of the enterprise where you can really improve discovery. So the way we see it is something like you have your search box or your input box. So you need something that that looks good. That is right on the middle of your website, and then a user can go there and ask questions such as you have any tips to build an outside kitchen with a barbecue. Very legit query. Then, based on the content we have from that client, we can build a beautiful answer like this one. You can see at the bottom, the citations, so this is the part of the documents where we found these answers. And then you have a very neat detailed step by step. I'd say bullet points where for each step you see what's the the tips you should you should use. So that's cool. But what about it when how can I make some purchases and what's the next step from a commerce journey? Which we see at this point is that you're gonna be able to bold on top of that, the top categories. So in the answer, we're gonna find some some hints regarding, like, here outdoor kitchen or even here a barbecue grill. So those are all ints of what category should be highlighted at the bottom. And then when you continue, you're gonna have, like, the categories and the full interface of the bottom with the facets so you can really start digging and understanding what you need. For the question you asked above. If we take a step back here, you realize that first off, we're extracting the attributes from the generate events, sir, And then secondly, we're gonna connect these attributes to the catalog fields and then recommend some categories from the catalog. So this is how it's gonna work. I'll show you a demo of it. It's not, it's not in the final form. It's still an early pro an early prototype, but we get something pretty good going on. So at the top, you see the generative approach. At the bottom, you're gonna use query of their standing and semantic just to make sure that we can connect both universe together. And then enough slides, I think you're all here for demonstration. So let's just get started. The first thing I'll do is use the, Coveo partner community. So if you are not yet a Coveo client, you may never use that community or it yet. It is a Salesforce community. That Salesforce, it's public community, This one is dedicated for our partners. It's a beloved partners that are helping us implement surveillance across, the world. So on the home page, you're gonna find the full Coveo classic experience with recommendations. It's all personalized depending on what you see. It's all gonna gonna morph into something that the that that is actually adapted to your visit. And then in that box here, you're gonna find these, query suggestions that are personalized So again, if you click on one of them, how to add Coveo on the, smart snippet on the page, you'll see the top widget, which is the Coveo generative and serving approach. If you followed earlier in the slides, you realize that this is grounded on the content that you see below here. You see that, heck, I'm I'm one of the reference if you click on this, you'll realize that on one of the forums probably got responded a few years back. There you go. That day. Hey. This is the snippet you need to use if you wanna use it in your query pipeline. So that's the kind of, that's the kind of content we can we can use for generative approach. It's all it's all possible. So let's go here and try something else how to, index content. So if you go again on the search page, we're gonna have that generative approach. It's gonna be based on the content. Like I said at the bottom, it's pretty generic. But then the hard the the cool part of that feature is that it's it it works with the authentication. So let me show you what happens when you have access login content. How to, create a partner orgs. Yep. How to create a partner organization. So this is something that we do. If you're a partner at Coveo, you can get a partner organization. It's a different license type with a few perks, for for for our partners that are doing testing. So you see here that it's not able to respond. Got a few hints, but if we're not confident enough, we'll just say sorry. The information is not relevant enough. We can't really process with that. So it's kind of a safeguard. But then if I log in at this point, I'll log in using, a partner name call, barca. So same question how to create a partner organization. You'll see at this point that since I'm logged in, I I got access to a different set of content, some support articles that are dedicated to partners. And then since with my permission, these are being thrown in the result set. I can use them in a generation. Well, that's how we do it. That's how we can scale permissions, security, in very secured, enterprise systems. That pandemic has never seen a day on internet, so it has never been learned by a by BT or any large language model. So this is why it's very interesting to see that we are using kind of a rag pattern providing grounding the model with the information we have. And we have these native connectors to reach whatever content you may have in SharePoint, in Salesforce, NSC, whatsoever. Another demo I want to show you today is the same concept but then apply to commerce. Again, so this is example, we, crawled some guides, buying guides, or how to articles, on the internet. And then for a specific client, we can see they have three different pipes, of, of articles. And then if you ask a question like how to install a bathroom, exhaust fan system, we're able to get the information across a variety of sources and then just merge them in a single article. So this is the first path toward product discovery, the next phase after, like we illustrated in slide is really to have the products at the bottom and making sure that we can have that link. It works very well, so far. We we gotta have, like, this is interesting. So how what kind of it's, yeah, I did a typo here for Christmas Lightning. So Christmas lights outside, you'll see that, it's very interesting to see getting the content from, like, cord and wire management, extension cords, and then Christmas lights. So it it's the big advantage of that feature is to generate a custom built answer based on multiple documents that usually a user would need to click and read the first, read the second, read the third. With that system, you can just ask a question. You gotta have it straight off the bat. Another good example here Should I buy a gas or electric or electric range? What are the differences? So we're able to provide a very good answer. Last thing before I leave the demo, I want to show you. I'll just go back here on our latest and greatest. So this is docs dot Coveo dot com. It's documentation website. We think this one is very cool. So how to create a query pipeline? If you try it, you'll see we're gonna have multiple, answers depending on what you're looking for all grounded on our documentation, but the latest and greatest here, you're gonna see a little bit more. So you have a copy paste button. Let's say you're a support agent and you're answering the phone or you're, on a email. You can ask these question, copy based and send the response right away. You can interact positively or negatively if you don't like the answer. You can collapse the component. Let's say you don't want it. You're just using a search page for quick references. You can collapse it if you And, we're also offering some rephrase option, which is the this one is the executive mode. So it's just gonna make it very short, if you really wanna something that is short and sweet. Otherwise, you can have bullet point mode. So there is a variety of outputs that are possible, and and We think it's really useful so far, for what we're using. So that's it for this quick demo. I'll go back over here and show you a bit the next phases what we're working on. So here, you see in the future at least from a mock up or perspective. So if we zoom in a little bit more, we wanna have inline citations. So this is where we're gonna see, like, you've seen the citation at the bottom. They're not associated to specific part of it. So this is where we wanna go. Really have, like, pinpoints citations on the right spots. We also wanna have a follow-up slash conversational kind of aspect. So you see here the question was like how to do machine learning and inquiry pipelines. And then we get the answer. You can still reformulate, but then you also have that ask follow-up where you could ask, yeah, but if I'm in the service industry, does it change something? Or even on the bottom, we have follow-up question example, which is where we call them question suggestions. So, if you're not a typer, if you if you're on a mobile device or if you just don't like, you have to type on your keyboard, you can just click and start digging in a specific topic. You could ask what is Coveo for Salesforce, and then it's gonna show you at the bottom how to access my content in Salesforce. And then after how to expose it on a search UI. So you can just go there and and waste a good amount of your afternoon just digging in content and making yourself actually more knowledgeable about a topic. Final thoughts on the topic before we jump in, for the round of question, We think that generative is here to stay. We think it's a very good technology. There's a few things we won't do, like generating content is not what we wanna do. We don't wanna build new articles. We don't want to help you write something. That's not our job. Our job is to be a good service provider and the relevance provider. And we think that cost can be very prohibitive. If you try to do it, we do and just call GPT yourself, you'll see that it's it's hard to build that infrastructure to ground it correctly, have the accurate results and call it So we think cost is a very important factor, and we think we have a good, feature, for the right price. The LLM choice is also interesting. We built the platform to be agnostic. Right now, no surprise, we're using open AI GPT. But the way we build the whole stack, we can plop in new LMS without any problem. We're investing right now in another provider as well. We're gonna give you the choice. Hell, we can even let you connect to your own subscription of GPT if that's what you want. The way we build it is also that we wanna be domain specific. Right now, GPT is a generic model. You guys are working in a specific verticals. So we wanna make sure that we're able to tap in your vertical. Are you in the, high-tech hardware industry or are you in the medical industry? These should have different icon different models in our opinion. We are contributing a lot to the science. So we're taking these models that already exist, but we're also using open sources model for embeddings for vectorization, and we're also giving back to the community. So it's not just about thinking, continue to be a lot of agnostic. I touched that, and then it's all about relevance across all in direction. Don't forget that if you build a new generative approach one end, and then it's not on par or it's way too powerful compared to what you already have, user's gonna be, not satisfied actually. The goal here is to have something unified across all channels. So that's it. I see there's quite a bit of, question, the chat. So Bonnie. Can you help me with that? Absolutely. We have a couple of minutes to answer a few questions. So we'll start with with this one. So, I wanna test this out. Can you say the URL again of the site that has this? Of course, of course, I'll paste it right here on the chat. So everybody gonna have it. Docs dot com is our public documentation website. So go there, have some fun. The other ones, those are communities. So I'd rather not send you over there, especially if you're not a client because we're getting people that are asking for login and create some accounts, and it's a little bit more complex. So, if you are a friend of Coveo, partner or a client, please send us a email and we'll send the others. But the docs dot Coveo dot com is a hundred percent public. So go ahead and try it. Awesome. So you mentioned different content types. Can you give some examples of the different types of content you could include? Yeah. So what we see usually right now with the first, beta clients, so the the first design partner wave is, mostly HTML files are very popular. So it could be knowledge articles, how to, etcetera, as a web pages. We also see a lot of Salesforce content in there. So it can be KBs or internal knowledge. Theoretically, the features support PowerPoint word PDF, no matter It just need to be free text searchable content. We are performing better with smaller documents. So if you have put all your knowledge in a thousand page PDF, it's a little bit harder to chunk and to extract. So obviously we prefer, like, better structured content, but it's very flexible. All the content that can be indexed with Coveo is compatible, basically. Obviously, some of them are better than the other, but that's the rule of thumb. Awesome. Yeah. And I would I would add to that, you know, along with being able to pull in these con the the content from the external sources. It's not just throwing it all in a bucket. It's really combining it into that unified index that has the machine learning on top of it. That's really what promotes the relevance and the better answers. Now we do have a few more questions, but we're just out of time. So for those of you who have asked questions and didn't get an answer, we will We'll get back to you via email with some answers for you. I wanna thank you Vincent for for joining me on the webinar today and sharing the demo. And thank you everyone for attending. Have a great day. Thanks, Bonnie. Bye bye. Thanks folks.