Alright. Thank you, Elaine. Good morning, everyone. Thank you for being here. It's a pleasure to be in Palo Alto where It all started for us in two thousand five. We had this little office nearby the water tower, so it was good to be back. Yeah, I did not seen those hotels in those days, but, you know, we're getting older and older, I guess. Okay. So in the next thirty minutes or so, I'm going to give you a quick product update what we have delivered over the past three months But then, and then we will switch directly to what we do around GenAI. If the Internet gods are with us, it will be a live demo, if not, I've got backup slides. Okay? So let's aim for great bandwidth and for everything to run smoothly. So always start with this slide that is a high level view of our Coveo platform. The search box is in the middle. We believe the search box is the universal way to interact with information, between users and information, and we believe that search, recommendations, and answers now go through that search box. This search box is becoming conversational, search box has multiple purposes, and we're going to show you a few of that in the next few minutes. I want to give a quick update on what we do in commerce, what we've released recently in commerce, In B2C commerce, so think about those large storefronts that are out there. Air powered like Coveo. So the goal here is really to align shopper intensive, so that's relevance. With business outcomes and that's merchandising. So what we're building in commerce is quite advanced, and we are moving that into our other lines of businesses such as service and workplace. So let me show you a few things that we're doing here. We're rolling that out to our large customers as we week. We call that intent aware ranking. So it's the ability to personalize results without having the user being logged in. So it's an anonymous fashion. The way we do this with a few clicks, so look at the products that the user presumably has seen or has browsing the website, With a few clicks, we can adjust results according to a user session. So looking for men hoodies, this is what I get. Is this good? Is this personalized? Well let's look at the next session here. I guess I'm in Quebec city. It's colder, right? So I'm in Canada. So I'm looking for two canned gloves and so on, the same query will give me totally different results. So how we do that under the hood with behavioral old data, we create product vectors that we could generalize into document vectors down the road. And with that, we don't need cookies. We have this kind of personalization that is built in at the indexing level. So this is something that is being rolled out now with our big customers, and we're quite excited about it. We also have Louis mentioned the acquisition of Qubit, so now called the Coveo Merchandising Hub. So think about it as an administrative tool for business users starting in context of commerce. So from there, you can build campaigns, you can manage boost and bury rules, you can personalize content, deal with recommendations, and advanced analytics. So this is an example of Boost and Burry rules in the context of a business user. It's a little bit easier to drag and drop and so on. And then what's really cool here It came from our customers. Let's open the black box and let's understand why your result is on the second position or third position and so on. So you see here, that's what we just released. We know that clicks have an impact, but the add to bag and event of commerce also had an impact and so on. And you see what kind of rules are also influencing the position of that result. So those are examples of what we have released recently in commerce. If we switch to service, So many of you will recognize the Salesforce service console with Coveo to the right that surface results based on the case that you're looking at, there are the search box at the top and so on, So, we've had a lot of comments and feedbacks from customers that it's great But, ideally, it would be even better if we could apply changes inside the Coveo admin tool, untie the layout of this from Salesforce and make it more generic. So that's what we've released, that's what we've released this quarter. We call that the host inside panel, so this part on the right of the screen can now be managed directly into our administration tool. So no need to go into this change management process within the Salesforce team. You can configure, you can do your own testing directly into the engine tool. So what it also means is that this inside panel is not while it's designed for Salesforce, it's not tied to Salesforce. It can be used elsewhere, If for whatever reason some some users are not in Salesforce, and they want to have access to this, it becomes possible. Okay? And finally, and that's more under the hood, right? Those are hosting regions, at Coveo, around the world. So we have North America, Europe, APAC and Sydney. We have a HIPAA cloud. Some of you may be using our HIPAA cloud. So I want to show you what we have done, what we have announced which we've rolled out this quarter actually. So that's North America, obviously. So we have a Canadian region. So presumably we don't have a lot of Canadian in the roof. They were in Palo Alto after all, but Canada is important for us, so we have this new Canadian region, but this is what is going to be interesting for potentially all of you. We now have the option of enabling activeactive. So for an additional fee, we are going to move from forty nine to fifty nine SLA. So for those of you who are powering a workload that is critical, the active active will allow us to go through. Any of you remember the AWS issued two years ago on US East, Well, we rolled that up in forty five minutes. We were quite good, but we were reactive. And we said let's make sure that Even if this happens only five years from now, let's make sure that we're covered. For less than thirty seconds of downtime. Okay? So that's what we have built here. We We're using US Midwest as a region. It's what's most central, so it will have very little to no performance impact. If this happens. Okay? So that's an option. Not all customers need that, but it's there, and it's Yeah. It's actually ready for prime time. Okay. Now, let's go to the main event. Covio generated relevance answering. So basically, this is, for me, one of the three most exciting milestones, the lifetime of this company. This is something that changes the paradigm of the search box and that will have an impact on basically all of our customers. So let me show you what we're trying to fix in terms of problems here, and then the demo. So We're trying to provide answers to those more and more complex queries. This one is kind of easy. How to add a new bank account feed in Canada. My robot is not following the path define in the map. How can I solve this issue, right? Or explain the difference between atomic and headless. When should I use atomic? This one is from Coveo. Documentation. Some of you may be familiar with those kind of things. So how can we do a better job with this? You can also apply in commerce where guided shopping may be may be more and more popular. Right? I'm a student entering law school and expect to travel a lot. I'm on a tight budget. What is the right laptop for me? I'm planning fishing trip to Florida in June. What is the optimal equipment for beginners? And then good old workplace issues, right? How to connect my lab to TV monitors in London office? Okay. Alright. So how to deal with that? Well, first, a lot of those queries will not be as precise to start with. Search, good old search is still critically important. Because if you think about my bank account feed question, a lot of people are lazy, or efficient, I would describe myself as efficient, right? Yeah, yeah yeah yeah. So why think about a long query when you can discover, you can refine your intent with a few keywords. That needs to remain and be supported. People do not stop to think about a long question when it's not clear. They want to help, they want to help have refined their question, they want to understand what's out there, Is this a valid question to ask to the system? Right? So that's search that most of you are using and very familiar with. SO, its discovery, its navigation, its content from multiple data sources, its refinement, its relevance using different techniques. This is not going away. We need to support this. Now what's now becoming expected also is when you know what you're looking for and you're going for a question like this, this is what you want. Right? This is the answer you want. Well, today everybody is excited and everybody is having a search system, but is building something to respond to these queries at the same time. And that's a problem for us, right? Because basically you've got Caveo here on the left side of this slide. So you've got connectors, you've got the index, you've got relevance, And then there are initiatives where going to build those system, those GPT systems with a lot of homegrown components. The issue is you're duplicating the search box you're trying to access to the same content using different pipes and different methods. So at the end of the day, for the same question, you may have a different set of facts, and that's an issue. Okay? And it's complex to build also by the way. So here's what we are doing. We are integrating the right side with the left side. And by doing so, on the back end, we've got depth and breadth of content because we have all of those connectors. Right? So we have access to fresh content. There's security involved. Which is critically important for Coveo. And I'm sure for most of you, and you've got admin administrative console and analytics. SO WHAT I SHOWN YOU A LITTLE BIT EARLIER ABOUT THE merchandizing HUB. THAT'S OUR BUSINESS. THAT'S WHAT we do for a living, right? So all of this system becomes available, but on the right side, By leveraging large language model in the system, we can have unified search for all queries. Keywords, questions, and conversation. Get answer to when there's an answer required, and we'll leverage all of the personalization and the relevance that is already configured in the system. So if you don't want certain elements of answers, if you don't want that to appear in certain certain documents you don't want this to feed your answers, it's fairly easy inside Coveo to remove that from the search queries, search results, and so on. So at the end, it means that there's implicit protection against hallucination. Under the hood, we're using a generic large language model as GPT-three and GPT four, three point five and four, running on our own infrastructure, so there's nothing that goes outside of Coveo Cloud, first, Second, we don't train the large language model based on your data. We use a generic model and we use it for the extreme ability and power to generate text based on advanced prompts that we provide. And the secret recipe here is all of the content that we have access to. All of the relevance, the secret and so on, we build a great prompt that will provide an answer. And then we understand the relevance, we understand the quality of the answer and so on. Okay? So that's what we're building. We're starting at beta for this next few weeks, but we're building this for scale. So we're building this not for not for a query once in a while. We're building this for industrial scale like most of you are using Coveo. So what I'd like to do at this point is switch to my laptop and show you live how we're looking at this. This is running on a demo environment inside our own cloud. So working on scalability, working on a few additional elements, but I thought it would be great to show you where customer is zero. It would be great to show you some queries on top of our online documentation. Okay? Some of you are familiar with those sites. So I will show you some queries that we happen to see more and more often. Okay? And what's interesting here is that there's specific context to Coveo. This is something that you asked to Google or you asked to bing GPT, you won't have a good answer. But if you ask that on the Coveo documentation, it kind of makes sense. So let me show you a few things here. So one reasonable query to ask is how does Coveo determined relevance, right? It's a question that we have often. So we have all sorts of results in the results list. But our large language model will provide a great summary of how determined relevance. And of course, you can interact with it and so on. Right? And you look at it, look, we're in demo environment. We're not in production yet. Performance is not yet it's not yet settled, and you can see that it's pretty fast. Let's get a little bit more complex. Explain the difference between atomic and headless. So atomic and headless are two different ways to build user interfaces with Coveo. And this is something that our partners, our developers, are always asking, we're always trying to understand When should I use headless, when should I use atomic? Well, okay, so let's try to see this. Explaining details of differences between atomic and headless, and when should I use it. That's pretty cool. And it's all coming from our online documentation. So if we add new documents or if we refine ways to deal with atomic or headless, it will surface in those results. Okay? We're using the large language model to use all of the content that we have and make that's about basically. Let's get a step further. Okay. I'm on my iPhone. I'm like, have sleeping. So I'm typing I'm typing very poorly. Okay. Let's try to let's help a little bit the Internet connection here. Okay, the God's helped us a little bit. So, okay, in one second, we switched to the -- in ten seconds, we switch to the slides if it doesn't work. Okay. Let's switch to the slides. Sorry about that. So, How to install Coveo for SalesDFRC? Okay? I'm sleeping, I'm lazy, I'm on my iPhone, I mean, I've got the French keyboard, doesn't make any sense. Well, look at the answer, it makes a lot of sense. But let's get a step further. Let's click on steps see the bottom on the bottom right corner of the excerpt, steps, and steps and bullets. So I clicked on steps. So look at this. This is pretty good. It's actually pretty impressive. Okay? Let's get a step further. Let's go to really what's the core of this whole thing. What is the difference between a trial and a partner or? And how to create one? Right? So our CSMs will always get those kind of questions. Our support team will always get those kind of questions. And if you see on our online documentation, there are decent results here, but there is no answer. Why is that? Because we don't want the system to hallucinate. So we have thresholds in place where we're not confident about the answer that is being provided by the large language model. So here's what we'll do. We will log in to the system, we will log in as a partner, so we have access to additional content, right? So if I'm logged in, there you go. Now I have the answer, and look at the first result here, it's internal only. I have access to this document, and because I have access to this document through Coveo, it will change my prompt, it will change my ends. And that's critically important. So you're leveraging all of what Coveo provides in terms of security in terms of relevance, and so on. And then let's get into a, let's expand that into bullets And that lady as a gentleman is pure power. So in a few clicks, in a few seconds, I have a clear, a crystal clear answer that is coming from multiple documents, multiple sources and some of them are protected and available presumably only to me or to my group of people. So that's what we are building in production, That's what's going to be available to our better customers. WE ARE GOING TO MAKE SURE THAT THESE NEW SYSTEMS DO NOT HELICINATE we are going to find a few things here and there that we're going to fix, and we believe that this will be a game changer. Okay? So that was the demo. Right? So maybe, okay. Now, that's what's working today with the help of WiFi gods. So what's coming soon? We're going to add citations and sources to this. Okay? So in the answer, we're going to explain where are the results coming from. So that's coming soon. And then what's critically important is how can we start having a conversation with a search box. So is this a chatbot, or is this a search? We believe that there's a lot of conversation that will happen in this interface, because the search box will always be more used than a chat box. Typically, right? So there's more data, there's more content, there's more there's more relevance into this, so we think that a lot of conversation will happen up there. And down the road, we may also connect that into a chatbot, if for whatever reason, the chat real estate is preferred to death. But we think that this is critically important. So ask follow ups, these are the additional areas that we suggest you look into and so on. And of course, you've got the results at the bottom to help you also. If you want to discover, if you're not sure, if you want to slice and dice with the facets, that's what we do for a living. Some examples in commerce that we're also experimenting with So semantic search is about understanding queries like this one. What are the best petal boards for a beginner? So those are the kind of queries that we may see more often, and we are likely going to surface certain elements of summary of guided shopping. But at the beginning, there's also the concept of discovery and understanding those queries in a very advanced fashion is critical. So in conclusion, some final thoughts. Lewis mentioned the cost aspect of all of this. Running those models is extremely So when the GPT honeymoon is over, we'll all look at this and say, whoa, okay, there are certain use cases that may be more critical than others, right, in the sequence of investments. So, Louis mentioned, it's 1000x running a single query of Coveo versus getting an answer. Yeah, so Let's say that a Coveo query, each and every time you click on a facet and you refine a query, it counts for a query, while the answer is pretty complete, it's pretty large. So let's say that you've got ten queries for one answer in terms of value. Okay. Fair enough. So it's one hundred times more expensive, great, than what? Then we're going to optimize the hell out of it. Just like we have done for the past fifteen years with Coveo. Now we have done it for the past, let's say, thirty days. With those large language models. There's a lot to do. Let's cut that by another ten x. Because I'm highly confident and I'M BullIish on this, So let's cut that by another 10x. It's still 10x more expensive. Okay? So you are going to select the right use cases for this. There will be multiple choices of large language models. Right now, open AI, GPT three point five and for are amazing. They are the best, but others are coming. So the Google ones is coming. Neta is also coming with has also released an open source version of it. So we believe that there will be some sort of commoditization of the generic large language model, and it's great, because what will happen is that many of you and very large businesses will apply some of those generic large language models and retrain them or fine tune them for your own domain. Or your own businesses. This is expensive, but you do that once or twice a year, right? And kind of makes sense. So and the open source, I mentioned. So what are we going to do with this? The way it's architectured, we can connect to other large language models. So we can connect to a large language model that is on your premise that has been fine tuned for your domain or your own corporate corpus. And we're going to use that to provide better answers, but still leveraging all of the content from the search to support that. Okay? So we believe that's the future. We have architectured our system to be basically, l and m agnostic. So the final thought is for us, it's all about relevance, It's all about relevance on the entire enterprise content and for all of the interactions. That's what Coveo provides right now. So we're going to add on top of that the ability to provide long answers and rich answers, without any compromise on security, on relevance, and reducing to minimum hallucinations.