Howdy everybody, I'm Peter Curran. I'm Coveo's General Manager for e commerce. Welcome to Relevance three sixty. I'm happy to talk to you today about keeping AI afloat. We're gonna cover three basic topics. First, the pace of AI, then I'm gonna talk a little bit about history and then commerce in a liminal space. There's no doubt that AI is moving the world really quickly that has velocity impact and potential. Four thousand seven hundred percent year over year growth in Gen AI referral traffic to retail websites in twenty twenty five. Two hundred and sixty two billion dollars, were, attributed to AI traffic in peak trading of twenty twenty five. And according to Mackenzie, the potential of consumer revenue through AI, specifically Agentic AI is estimated to be five trillion dollars in just a few short years from now. So this thing is moving fast, making a big impact and has a lot of potential. And it's undeniable as a consumer. We can see in things like Gemini on the left, ChatGPT on the right, The development of these new AI surface based shopping concepts, UCP from Gemini and ACP here from OpenAI and ChatGPT. And this is really, really cool. It's exciting. It was just announced a little bit less than a month ago in the case of UCP and we know that people are already thinking about how they can integrate this if they haven't done so already. And we are keeping pace. Coveo has announced that we have a hosted MCP server that you can use if you are among the leaders who are integrating these kinds of capabilities in their platform. It's a thing you can do with us today. As you heard from Laurent, the personal assistance and Agentic AI platforms are both something that can talk to Coveo's MCP server as well as our agents and get back to data that you have deep within your enterprise. In this sense, search isn't just search anymore, it's really an abstraction layer for enterprise transformation. It's an abstraction layer that allows you to transform transform the back end without having to worry about slowing down front end evolution. It's also just a transformation in general of the front end experience. People can shop with agents, they can shop on traditional sites. Coveo is there and ready to integrate with both of those shopping modalities. But I wanna talk about history for a moment. I wanna talk about this dude right here, Gustavus Adolphus. He was the King of Sweden from sixteen eleven to sixteen thirty two. He was considered quite handsome. I think he probably did have a neck, you just can't see it here in this picture. He became the king of Sweden at seventeen years old even though the law said you couldn't become king until you are twenty four I believe or something like that. But he had been with his father on many military campaigns, been around all of the leaders, everybody had faith in him, so they made him king. And he was a very aggressive, acquisitive king. He was warlike, was trying to take more territory for Sweden. And he and some of the rulers before and after him were very successful in transforming Sweden from some, you know, kind of regional power to one of the most important powers in all of Europe. And one of the things that he realized he needed when he was going after the folks down in Poland, was he needed a better ship. And so what he decided to build was this ship called the Vasa. It would be the most impressive single war fighting machine in the world at the time. Notice it was started in sixteen twenty six, that's exactly four hundred years ago. In fact, the construction of this ship was happening somewhere between February and March of sixteen twenty six, exactly four hundred years ago today. And it was all commissioned and endorsed by the king himself. He wanted to see lots of guns on this boat. He wanted to be able to pull it up and sink rather than having to board and fight man to man. He wanted to be able to sink ships from afar using all of these guns and cannons. And you can see some interesting metadata about this ship here. He wanted that so much that he told the designers to go ahead and put a whole another deck on the ship with more guns on it to make it an even more devastating weapon of war. And if you don't know the story of the Vasa, it does not end very well. Just in, not exactly its maiden voyage, but just moving from the place where it was built to a slightly different place there around Stockholm. In a very light breeze without a drunk captain, it sank. And the sad thing about it is they knew it was going to sink. The engineers had protested. They even had taken to the admiral of the Swedish fleet, got all of the sailors on the top of the boat to run back and forth as one from one side to the other across the boat and it was going back and forth so much that the admiral stopped them because he was afraid it was going to flip over with just people running back and forth. And I think he remarked something like, if only the king were here to see this. But it was the king's baby, it was the king's dream. So the king pushed the teams to get this product into production if you will and it ended in disaster. Only thirty people luckily were killed but the Vassa sunk to the bottom of the harbor in Stockholm where it sat until the nineteen fifties or sixties where they raised the whole thing up, preserved it and built this awesome museum around it. It's my favorite museum in the world in Stockholm. There's one like it in Southampton in the south of England, where you can see the Mary Rose, which is which is older but kind of a similar story. Absolutely fantastic, thing for you to see if you're ever, in Stockholm. So you're probably asking yourself, why is this dude telling me about a seventeenth century ship when we're supposed to be talking about AI search? I think this story is really resonant for me, not only in the fact that it's four hundred, years ago but, and and at the same time of year and it's an anniversary, but it's resonant because first of all, budgets don't create success. You can't just throw money at engineering problems. Secondly, you always have to separate business goals from your technical design. You have to follow best practices and listen to the testers when they tell you things aren't working. And like the ship, AI isn't just sculpture and guns. It's not just whiz bang features that you can see in a video from Google or from OpenAI, it is about the ballast, the stuff at the bottom of the ship that keeps it from tipping over. And in this analogy, that ballast is content. It's your enterprise content which is spread all over your enterprise in different line of business systems. So a lot of people think about, okay, they understand I need to get my content in order to do a search project. I would have you believe something just a little bit different and that is that your data project, your content project is never done. It's an evergreen program that lives constantly within the organization. So I wanna talk now about with that analogy put to bed, how commerce is in a liminal space. Do you know this word liminal? My kids use this word a lot, I didn't really know what it meant. I don't think they really know what the word means either, but what they mean by a liminal space is sort of a space that's caught between, that's kind of eerie and caught between two places. And I think that's where commerce is right now. On the one hand, we have all of this really killer AI stuff that Google and OpenAI and Anthropic and other people are talking about, but then we have the reality of these businesses that we're running on e commerce websites where billions and billions, trillions of dollars are being transacted all the time. So I wanna talk about that for just a minute. If you look at the UCP announcement from Google that came out in January of twenty twenty six, one of the things that you will see on the landing page, which the URL is there, and why you should implement UCP is that you want to maintain full control of your brand. But I would ask you, how do you maintain control of your brand when the AI surface, in this case Gemini, is a hundred percent in control of the customer journey. You're also not there at key moments where you can introduce cross sells and upsells the way you can in a website. So I would posit that you want to adopt these kinds of technologies because consumers will use these kinds of technologies to buy products from you, but you also want to invest in your existing on-site e commerce experiences. And as our CEO said, there is a trust gap. Hopefully you saw Laurent's presentation about this as a slide straight from his deck. LLMs don't know your enterprise. Enterprise data is spread out all over the place and that's okay because we have a retrieval platform that can make those LLMs not hallucinate and produce good quality answers. And let me give you an idea of what that might look like in the future. It might look like somebody on an AI surface or on your website asking a question and getting an answer. But let's look at a really meaningful kind of question. This one says, for our enforcer, which hydraulic pump replaces PPHP four hundred seventy two thousand one nineteen fits as maintained and can arrive within a week? So there is a pretty complex question. That question probably doesn't mean very much to most people in the audience but there are a bunch of interesting parts of this question. First of all, our enforcer. Enforcer is capitalized so we can infer that this is some model of maybe vehicle or big complicated machine. They're saying our so this person clearly thinks that they own this. So now when this question gets asked, we need to know who this customer is and which machine they're talking about. If you look in the garage of the enforcer, all the different equipment that maybe this customer has with you, Maybe they have several different enforcers and you need to ask a question back. Can you please send me a serial number? You may need to be able to accept a photograph of the serial number rather than somebody pecking out sixteen or twenty four letters and numbers. Once you understand which one they're talking about, maybe the question about the product that they're actually looking for, in this case a hydraulic pump, maybe there are multiple hydraulic pumps, maybe one of the enforcers in the garage has three and the other has four hydraulic pumps. So we need to know which one they're talking about. That may require a question back. Maybe when we figure out which one they're talking about and which hydraulic pump they're interested in, maybe we know that that hydraulic pump was just replaced a few months ago. So it shouldn't be bad right now, perhaps that tells us that something else is in fact wrong rather than the hydraulic pump. And then lastly, the question is, can this thing arrive within a week? That's a question that the order management system or some part of the fulfillment apparatus would have to be able to answer. So even a fairly short but dense question like this has a lot of detail. The answers to all of these questions in this short, the details in this short question are something that you could generalize as context. And that's what search engines do really well. They wrap non product and product content in context. We believe that the future of enterprise is context that's connected in real time. That can be accounts, inventory, SKUs, compatibility, fitment, products, all kinds of different concepts connected in the answer to one question in real time. And what that means by extension is that all of that stuff is stored in different backend systems, CRMs, ERPs, mainframes, databases, under Bob's desk, all over the place. And what search engines are really good at doing is not caring about where things are managed. So while some companies will argue that you need to put all of your data into a data lake in order to make sense of it and use it in AI, Coveo believes that you need to put it into an index where you can understand it contextually and manage it and provide it to AI surfaces. So that looks like these systems connected into that same diagram I showed you a few minutes ago and we have more than a hundred connectors that can connect all kinds of different systems, substantially everything you'd find in the average enterprise, index it and make it available to AI. That's the future of enterprise. It's not about your catalog and your price and how fast price changes and inventory. All of that stuff is important. It goes much, much, much deeper than that. And that's what we hope to bring to e commerce is an understanding of all that enterprise data. But the fact is the present of enterprise is still that search is messed up in a lot of companies. And let me give you an idea of what that might look like. And I don't mean to pick on these guys. I've done this in a few different presentations where I talk about companies that have search issues. Lots of companies have search issues and we're constantly working on solving those kinds of issues, so we totally get it. But here I've searched for electric wood splitter. And you can see that these are really relevant results, right? They're all wood splitters, no doubt about that. They're all electric. You can kinda see that the cord is wrapped up in each one of those different machines. So we know that these are relevant results. And look, they're three hundred, four hundred pound in price. So these are substantial considered type purchases. If I change just one word, if I go from electric wood splitter to electric firewood splitter, all of a sudden I'm getting this stuff and what is this stuff? You think what is like is that the axe or the hatchet that they're selling? No, it's the log that they're selling. Not a firewood splitter at all but the thing that you would put your firewood on were you to chop it with an axe. So just one little word can mean potentially a disastrous outcome. Like there are people in the world, I promise you, who search for electric firewood splitter and when they see that B and Q doesn't have it, diy dot com doesn't have it, they go somewhere else. And that still is a real problem that exists today all over the internet. We solve this by focusing on your business. So you won't hear us come in and talk about a ton of technology. We're focused on how can we increase the number of sessions to the website and make sure that it's good quality traffic? How can we convert that traffic with a set of tools that are AI based? How can we do that where we're increasing your average order value and thereby generate more revenue that you can keep and not return back to your customers because we know for example that it fits. Do all of that at lower costs, both cost to buy as well as lower cost to operate. We do that with a variety of different features, everything from precise semantic search, one to one personalization that's demonstrable in session on public websites that we'd love to show you. We can drive more relevant traffic to your site through a sitemap expansion play that's Agentic. We have conversational capabilities, so if you wanna have a multi turn, multimodal conversation with Coveo, you can do that. We have recommendations, so if you wanna put recommendations on content pages, on product details pages, anywhere throughout the site, we have more than a dozen different recommendation strategies that you can deploy. And then finally, we do all of that with total control. We believe humans are the ones who should be driving artificial intelligence, not the other way around. So we try to make it absolutely clear to the humans who are operating the e commerce site why products are coming back and why they're in the position that they're in. And through that total control, we feel that they can be more productive. Furthermore, and I think most excitingly, Coveo offers a copilot for its merchandising interface where a business user can state some kind of general business goal or concern and the copilot will come back and tell you things that you can do in terms of business rules in order to achieve that business outcome. I think it's really killer. I hope you'll come and see a demo of it and thank you very much. Enjoy the rest of the conference.