Welcome everyone. I'm Olivier, lead product manager for the experience side of the product here at Coveo, and I'm with Elias. Joining from the product team as well. So we're going to talk to you today about how to design commerce Agentic capabilities essentially and what capabilities truly matter within the Agentic stack and the modern Agentic solutions. So Elias, take it away. Yeah, thanks. So today's agenda, we're going to be going through changing shopper and buyer expectations. We see a lot of that happening today and we're going to cover some of that. The chatbot trap, this is a big one. Yeah, for sure. It's a big one, right? Like how people are sort of confusing essentially Agentic solutions with chatbots, and a lot of vendors are positioning chatbots as Agentic solutions. But one is not necessarily warrant of the other, right? Like I think it's important to differentiate between the two types of technologies. Absolutely. As well as the five Agentic Commerce capabilities, right, that are really critical to making a solution truly agentic. Of course, that's not an exhaustive list, right, but we feel that those are some of the five most important capabilities that you need, as well as how to prepare your enterprise for the Agentic world essentially in these Agentic solutions. And we're going to look at a live demo as well as part of that enterprise preparation step, right? Exactly. And then closing it away with some key thoughts and high level overview of what we've talked about. So it's really important to note that AI has really changed shopper expectations, right? Customers are now shopping very differently than they were years ago or even last year, now where a lot of the commerce journeys now start on AI agents essentially. And it's an ever growing trend, right, that's going to keep growing in the future. And we strongly believe that to become now part of the normal buyer journey. It doesn't necessarily mean the rest of the tools are completely going away, right? It just means that now shoppers want the full optionality, essentially, of being able to choose where they begin their journey, where it ends and to be very fluid between these different channels. So one's traditional search experience doesn't fit at all, right? And we see giants like Google even today completely revamping their whole experience around AI modes and etcetera to allow customers to still retain market share essentially of some of those shoppers and to be able to adapt the experience based on the shopper's intent, right? Whereas before, we were stuck in a bit of a wireframe, right, where you had traditional search results and people were taught to search a certain way, AI is really shifting customers' expectations of now being able to express their intent in a natural way to them and then essentially to have the engine adapt to their needs, right? And that's really what we're seeing with these different types of interfaces now that you've probably noticed throughout your different digital journeys, that you have today. So essentially, if want go to the next slide, Elias, what we need to do is really to evolve discovery without disrupting the conversation, right? And I would say also to evolve conversations without disrupting discovery, right? The last thing customers want, and Google has shown that, right, by backtracking on some of its AI forward looking capabilities that they've put forward, is really that customers want the freedom of choice, right? And brands, what they want to do is really to preserve the revenue that they have today and preserve what drives that revenue to make sure that they don't just drop search, recommendations, discovery, navigation, etcetera, from their experience while essentially meeting their increasing expectations around agentic, conversational and being able to declare their intents essentially throughout their experiences. So a lot of our customers in our industries fall into what we call the chatbot trap, right, where they figure, okay, well, in that case, I'll just launch a chatbot on my digital properties going to be able to answer customers' questions in a conversational way. But that really creates a disjointed experience with the regular channels they already have in place, and it really creates two parallel tracks that don't really reconcile anywhere throughout the journey, right? So customers can land, do a search on their website, and when they go on the chatbot to do a similar inquiry or they want to dive deeper into some results, they have to start over. And once they're done on the chatbot journey, they then have to start over again in the traditional channels that they're used to, Right. And so this creates really a disjointed experience between a user starting on a chatbot or starting in search and trying to switch to either one. Exactly. And not only that, right, often these chatbots are single purpose today, and they are, you know, of course now adding functionalities and etc, But often, you'll have a different interface for commerce than you do for support than you do for discovery than you do for post sales support as well, right? So all of these different channels, we strongly believe, need to reconcile into a single surface. And the solution here is not to just slap more tools under a same agent, right? The solution here is to really have to rethink the system and how these solutions work in terms of giving that functionality to end users. So really what we're talking about is changing the dynamic layer and how to make commerce truly Agentic here, right? Right, and what are the key building blocks you would say in terms of achieving that, right Elias? Yeah, I think it's key that we focus on things like maintaining that context between both the search experience and the chatbot, but unifying it into one place where context is carried throughout. And what about other experiences, right? What about support or case creation, or etcetera? Do you believe that needs to be maintained as well, right? Right, that's a good point, because as users transition from either a shopping experience to a support experience, that context needs to be carried throughout and the experience doesn't need to change dynamically in the sense that they have to have a totally different experience but that chatbot needs to be able, or rather our conversational product discovery needs to be able to carry through across both those unique experiences. Yeah, of course. So we're also going to look at grounding in trusted data. Of course, Coveo is paramount in grounding against that trusted data, and adapting the experience of the task. So like you just mentioned, whether it's a support case or whether it's a shopping experience, we're able to carry that trusted data across both those tasks. And that's really critical, right, because enterprises have really complex knowledge bases, complex data that resides in the enterprise knowledge, And that data isn't always known, even by modern models, right? If you think about a parts vendor in a complex B2B environment in a regulated space, often that knowledge over these parts is not made public, right, because it requires certifications and etcetera. So truly needs to be grounded in that user's accessibility of information and knowledge, right? Right, so we can't just point the LLM at a large data set, right? It has to be curated and indexed in a proper way, following those security permissions as well that are enterprise critical. For sure. We're also going to, of course, focus on keeping those business teams in control. We just spoke about the enterprise criticality of security, so keeping that control is going to be a key element. And then, of course, allowing those agents to act on the user's behalf is going to be very important as an end state. You have to make sure that the experience doesn't stop dead and that you can close the loop essentially with the different workflows and agents you may have elsewhere in your business. I think there's also an important point around making the experience adaptive to the different intents, right, to the different tasks on hand, right? It's a good thing to be able to delegate, to be able to ground in whatever, but still users' expectations are there, right? And users will expect to be served in a certain way and that if they express their intent in a certain way, to be responded essentially, considering how they express their intent, right, which is key in my So keeping those business logic and guardrails in will be very important. Let's look at an example of this. So if I might be a shopper looking for a new fridge here, I might start with an intent. I'm looking for a small fridge, but of course there's some intelligence layer under this that is looking not just at the types of fridges, but reading that secured metadata to find me the exact fridges I need. Yeah, because in the end, in that catalogue, there's probably no explicit mention mention of what a fridge for a small kitchen may be, right? Maybe you'd find that information elsewhere, but maybe not, right, also. You you have to make some assumptions in this case, right? That's right. So we have that grounded context in the types of fridges being presented in front of me. And so as the experience evolves here, and I may be looking for something else that's important to me, like an energy efficient and bottom freezer, now we're adding a dynamic layer here to the actual experience, and what kind of fridge I'm looking for will change. And this is still relating to the fridge you were looking for before, right? So to our point earlier around, you know, adapting the experience and managing context, yes, the answer here is changing, right, now to highlighting specific attributes, spotlighting, you know, that fridge, right, that answers this prompt. However, you know, it's also contextualizing it based on the fact that you're looking for a fridge for a small kitchen that you were before. That's right. And now that experience is that context is being carried across this experience here. And if I go into even deeper, here I'm looking for a comparison to a French door model from LG. Yeah. Okay. And probably referring also to an LG fridge which was returned right in that first query. If I remember correctly, it was the most expensive fridge there, right? Being able to also manage that context, not only to keep it going from the previous queries, understand here that there's additional context that needs to fetch. Exactly. And so now we have a comparison table that the user can go through and experience with, look at in plain HTML here. And here, right, even on the comparison table, I see here there's a few features that were selected and etc. Are these things from an Agentic perspective that should be hard coded, or is this something that the agent needs to reason over? That's right, that's the agent reasoning in a dynamic way over the context and over the security permissions that we're respecting of course and the user's intent most importantly. So essentially here, right, when we talk about Agentic, we often think about taking action on behalf of the user. And that's what we're starting to show here, right? These are simple actions, albeit, but being able to essentially reason about and create those correlations between what the user is asking for, right, like a small kitchen, energy efficient, extracting these attributes in a relevant way so that when you ask for a comparison, then here, essentially, it brings back all the relevant attributes to that context, right? That's right. If I had come in and asked for something else, albeit these tables would have adapted, right? Yep, exactly. Guiding that user to that experience, ending with an action, like you just mentioned. So we're able to come to a conclusive action here where a final product is presented to the user after going through a series of questions, going deeper and deeper on their needs, to where now we can actually add the product to cart. And here, structuring it in a way which makes it achievable for users, and makes it more efficient to get them on their journey. That's right. Easy to understand here, easy to present, we see the final state product that's energy efficient, meets their needs, and easy experience. Very cool. So what I'd like to talk about now is really that dual layer of grounding we just saw in action. We just saw the catalogue and the content meshed together with the agent's dynamic experience. You can speak to that a bit, right? Yeah, of course, right? Like in this example, for instance, right, you're looking for a refrigerator for a small kitchen, right? But this is an enterprise, right? So they want to have control over the tone and what's being said essentially through their experience to make sure that we're saying the right thing and it's on brand and it's based on the products that they have, right? But albeit, right, on these fridges, there's probably no attributes called kitchen size, right, for a small kitchen or a big kitchen. So we have to make these correlations essentially between what does a fridge for a small kitchen mean, right? So what attributes does that mean? And that information can often be found but more in the unstructured content that these enterprises have. They can be buying guides, they can be customer reviews, they can be FAQs, support cases even, and etcetera. But here, what we're really doing is using all that ungrounded information to then be able to reason about, right, using the agent essentially and be able to draw lines between that ungrounded that grounded, you know, unstructured, sorry, grounded information that we have in customers' knowledge bases and use that to essentially discern what we need to look up in the catalog and to really blend it all together in Answers, right? So it's less about one or the other as we had before, right, and as we've seen before in the market where you had a solution to be able to answer your questions and then you would go and find products, now we can use agentic reasoning to actually bridge that gap and go directly from finding the right information and looking up the right products simultaneously, right, giving a much richer answer, of course. And of course, there's different types of information that can be found in the enterprise, right? So on one hand, you have a product catalogue, right, which contains all of your structured attributes for your products. This is really what gets merchandise, what powers your website, your traditional product search, right? So if you want a filter, etcetera, you need these attributes to be tagged and explicit and deterministic, right? Then you have the unstructured grounding in enterprise content, right? So this is really general topics of discussion, right? General cases, buying guides, etcetera, things that speak about products and their use but aren't directly necessarily related to them, right? We have large enterprise customers which have millions of support cases, for instance, but those support cases are never explicitly tagged with the products they relate to, even though have a lot of rich data, right? Yeah, of course. They only speak about products, right? And then you have a third category of data, which has often been underutilized up to this point in commerce experiences, which is really the product related content, right? So things that you can usually download on a detailed page, but that nobody uses or has access to when it comes to these more dynamic experiences. So you can think about specs, installation guides, ratings and reviews, compatibility documentations and etcetera, right? So everything that pertains to the information regarding the product that the agent can now use to blend with both the unstructured content and the product catalogue to give these rich answers. And so all of this really at the Agentic layer ends up being just a really dynamic experience for the user who's really going through a shopping experience, a case support experience, or just knowledge retrieval experience. Exactly, and a bit recreating that human experience that customers are accustomed to when walking into a store or calling somebody and speaking to an agent. So really trying to bridge that gap between computer driven experiences and human driven experiences by having an agent as the intermediary. Oli, we spoke about during that little fridge experience that we just showed, that adaptive design as we went through the experience. So we started with just one fridge, we went into a comparison, and then we started really defining what we were looking for in energy efficiency, for example. Talk to me a bit about that adaptive design, how we're able to generate and use comparison, different products, rich knowledge like you just spoke about. So yeah, and this is going to be a good segue for a demo of one of our customers' properties to see this in action. But essentially here, the industry has actually adopted a protocol called AGUI. So what AGUI is, it's a declarative protocol that allows agents to essentially treat front ends like a series of building blocks. So you need to couple it with a library that can accept it. In our case, we're using something called A2UI, which is also a bit of an industry standard and a custom library that essentially can be brought to the front end. And instead of presetting wireframes as we would before, right, so you'd have a website, you'd have a search grid, you'd have facets, etcetera, here these libraries essentially wait for the agent's stream to tell them what to render. So a GUI is that protocol that essentially alongside the Answer will have the agent say to the front end literally, oh, for this answer, I want a product grid with a table with blah, blah, blah, right? And that gets rendered on the front end. This has a lot of benefits, right? Because the first iterations type of experience were really around generative UI, right. So this is about having the LLM at inference time generate components dynamically on your front end, which brings a lot of drawbacks, right. First, it's super expensive, it's slow. Also, it lacks control, right, because agents are nondeterministic beasts. So even if you had a lot of guardrails around it, a lot of systems around it, there's never truly full ways to fully control these experiences. With a GUI, you're sort of flipping the tables around, right? So the agent is aware of what's available in the front end. It shares a state with the front end, right? So it's also aware of what the front end is currently displaying and rendering to the end user. And then it can tell the front end and instruct it on what to render next and what to do next from a preset series of outcomes, right? So we feel like it's a really good compromise between having a fully dynamic experience but keeping the enterprises in control experience and having a deterministic outcome that they can show their end customers. So really giving the agent a tool belt with all the tools needed to really render that enterprise level experience. Exactly. And let me show you that in action on one of our customers' site, which is Freedom Furniture. So Freedom are a retailer based in Australia that sell furniture and home goods essentially to be able to style your home, right? So Freedom have recently deployed our Coveo conversational product discovery and Intent Box solution on their property. And they've deployed it in two separate ways, right? So on one hand, they had an Ask AI button here that allows you to reach the agent directly, and we'll get there in a second. But on the other, they've also implemented our Intent Box, which is really what we believe to be the right evolution of search, right, within these new paradigms. So as we said before, right, we don't believe these channels are going away, right? We believe that search just needs to evolve by using these new technologies to be able to answer all these intents, right? The example Elias showed earlier, right, if you said, you know, I'm looking for a fridge for my small kitchen, could be achieved with traditional search, but you would end up with very fuzzy results or partial match being the main driver of that search, right? You need dense embeddings and etcetera to be able to run a semantic similarity search, and that would really lead to maybe hitting on the right products but throughout a bunch of noise, right? And there's a bunch of papers and techniques out there to try to optimize for that, but it's a bit of a never ending game because essentially it's a bit like using a hammer to knock in a nail a screw, sorry, right, instead of a drill, right? Like, you know, you need the right tool for the right job. In the case of Freedom here, what they've done, and I just want to draw your attention here to this little toggle, which we'll talk about in a second, is they've implemented our intent box solution within their main search box, which essentially blends search and conversational together. So search actually becomes one of the potential outcomes of this intent driven experience, really retaining the best of both worlds, right? So if I go in here and use our query suggestions to just click on leather sofas, for instance, right? Of course, this is doing a full trip around the world to Australia, in this case, but this will, of course, give you results for leather sofas, right? So traditional search, which customers are used to with facets, filters, availability, yada yada yada, right? So we won't spend too much time here. But essentially here in this case, I typed a simple query clearly showing an intent towards viewing products directly. And instead of putting an agent in my way, right, we're just showing you relevant results as we have been for a long time. So keeping this sub second, more in the hundreds of millisecond range, right, and making sure that customers can get to their results really quickly. However, if I were to go into this box and say, oh, I'm looking for a new table for a small apartment. Can you help? You can see here already that we've detected that the user's intent to switch to asking a question, right, where we're showing some popular shopping questions that customers may have had. And then here, we're going to take you to this Agentic interface instead. Now one important thing to note is that both of these interfaces are actually the same, right? The search we showed before is actually rendered as an AG UI policy within this A2 UI surface, meaning that while we're rendering these results here, we can actually completely morph the interface based on users' needs depending on the intent that they show essentially early on. And here I'm falling into this Agentic task, which, as Elias showed earlier, right, is able to essentially disambiguate my query and find the right products based on what I'm looking for. So here it's suggesting a bunch of options, right, of these small tables and a bunch of different types of styles, as well as suggested next steps, all under a single API request, right. So this is just one API request to Coveo, no orchestration needed on the customer's side, and we render this full interface. But since it's a GUI, right, it can be styled to any customer's needs. So if you go on other Coveo properties, you'll see different interfaces and different experiences, all driven by that same unified experience. Also, agent looks like it's asking clarifying questions. So that experience really feels like an in person shopping experience all through Freedom's website. That's really the goal. Right? And it's fully context aware. So if I go in here and I say, I want something in light wood, it's going to be able to understand that I'm still looking for my new table for my small apartment, right, and really narrow down these results essentially for me, dynamically. But the context management essentially is key, and adapting that experience is also key because if we weren't doing that, here, you would still be in your wireframe in a traditional chat model, which realistically doesn't use up enough room and really leaves a lot of money on the table from customers' perspective, right, and sort of distracts from the main experience. If I then go in here and ask, you know, what are the differences between the Gaurizia, Amarillo, and Manjuri, because I want a round table, right, and those are the three round tables that are here, it's going to be able to essentially morph the experience once again, show me a comparison table, you know, with those three products there and help me really make my choice. And because I really ask here a broad question like what are the differences, the agent is going to be able to use its reasoning skills to actually be able to figure out what are the relevant attributes that it needs to show me to help me make that decision, right? So you see here in this case, it went for standout, trade off, best for and price, meaning it's giving me a price indication, right, because, of course, customers are influenced by price, but also trying to already summarize a lot of these attributes and results and how they compare against one another to remove a step that previously was owned by humans, right? A lot of our customers have previously generated comparison tables using search results, but they're always hard coded, right? So you use facets and you return facet information, but you're still letting that user make that final judgment call, right, on what that means for them. Whereas in this case, we can go one step further, right? If I were to go in here and say, you know, could you recommend some chairs? I like the Amarillo. So now what we're also talking about is really changing the direction of the conversation. We're not only just focused on one linear objective, now we're switching completely. Well, not completely, right? Because if you think about it here, I said I like the Amarillo, right? So I'm saying, you know, essentially, I like that second table that we previously returned, right? But it's giving the agent a good directional indication of the type of chairs I may actually want to look for, right? So you see here, I like the Amarillo, and if I scroll back up, you can see the Amarillo here is actually a square table, right? Like and it's actually peak and timber and etcetera. It's warm with bright tones, as the agent has described. So here, it's actually showing me chairs that pair well with that Amarillo based on Freedom's style guides, right? So to our point earlier around grounding and enterprise content and everything, Freedom have stylists, right, that actually help customers make these decisions in real life. And what we've done here is we've trained the agent to actually use all of that styling information that they use to train their internal employees, but this time to train the agent essentially to deliver some similar experience on their interface. I really want to double tap into what you just mentioned that now the agent is really recommending products that earlier in the conversation you alluded to being your favorite. So it's carrying that context throughout the entire conversation here and as it develops, really remembering what you spoke about. Yeah, and the more this goes on, the more I will know about who I am and what I care about, right? And last but not least, it even has access to a lot of other things from Freedom's site, right? Freedom, in their case, have a lot of customer reviews, for instance. So if I ask here, you know, and I say, oh, you know, I like the Camellia, know, what are people saying about its, you know, durability over time. So now really tapping into that rich enterprise content that sometimes gets left out of these experiences traditionally. Of course, and there's two things that we can demonstrate here, right? And I must say, know, I don't know if there's reviews specifically for the Kamalia chair, but that's a critical piece as well, right? So here, we can demonstrate that if there are reviews, right, it will tell me essentially what customers are thinking about. But if it doesn't, as you're seeing here, it's going to prevent itself from answering, right, because in the end, you know, it doesn't know, right? And that's where the grounding piece comes in. Retail enterprises don't just want to deploy LLMs on their websites, right? They want to deploy Agentic Advisors on their website. And that's really what sets the bar between these two different things, right? All right, let's now have a look at another example, right, where maybe the customer doesn't know exactly what they're looking for, but they have certain constraints that they want to follow, right? So let's say, you know, the customer says, know, oh, I like contemporary, furniture. I want a sofa but have a two thousand dollar budget, right? Want to sit three people. What do you recommend? So here you see, like, we put two constraints in place for the agent, right, with a directional, essentially, guideline. So I said I like contemporary furniture, right, so it understands, you know, based on the Freedom Style guides that it needs to recommend some contemporary sofas. I also gave it two constraints, right? First, I told it about a budget I have, right, so it knows not to go over two thousand dollars and also knows that I want to set three people, right? So it's going to look for that attribute in these sofas and actually recommend a sofa that can seat three people, right? So here it's giving me an assortment of sofas that follow my constraints and actually work well with this example, right? If I now ask it, you know, ah, okay, I like the Clio and, actually, no. Let's say I say, you know, what are the differences between the Clio, Tomi, and Arrow, It's gonna be able to essentially, you know, structure these comparison tables for me, you know, highlight all these key differences directly in line as I've showed you before, right, and really, you know, once again pick these attributes. I I once again ask for a very broad and generic question. So it's going to be able to select from the list of pre merchandise fields what the different attributes and comparative attributes and functions that they want to go for, right? Choosing again the standout, trade off, best for and price here. If I then go here, right, and ask, you know, okay, but what are people saying about the Aero, the agent's going to be able to go and look up the reviews that we have in place regarding the Aero sofa and actually structure and answer it directly in line, giving a bit of an overview of what the Aero sofa is creating in terms of follow-up and support and everything, right? So you see here once again, we're blending that dynamic surface, really spotlighting on that Aero sofa because that's what we care about. And here, it's able to summarize these ratings and reviews, saying it's four stars with customers praising its comfort, structural design and Buchla fabric, as well as highlighting some of these key features. So this is really what I wanted to show you in terms of what Freedom are doing to really elevate their new search experience, right, and being able to address all of these different new intents that are being created by customers' expectations shifting because of AI, right? So how about user control, right? Because some of these experiences are great, but at the same time, we deal with enterprises, right? Freedom has a ton of stores, has a lot of people working there, right? They want to have a deterministic way to essentially drive these experiences. Otherwise, they wouldn't have confidence in deploying these, and these would just remain prototypes forever, right? So what are we doing to address that, Elias? Yeah, that's a great question. Our enterprise customers today really are focused on that enterprise security we provide at the agent layer that both carries our merchandising rules as well as regular query pipeline rules. But what you just displayed, of course, shows a lot of important things. In the background we have guardrails and business logic that's happening, so really the agent you're interacting with becomes really a Freedom, a storefront agent that Freedom can be confident about deploying. It knows what context is being handled. It knows what kind of layouts to render. And we saw all of that in your really, your example here. And I really wanna call out the fact that you were able to go through this dynamic experience with, of course, conversations as a historical backdrop that you can go back to and reference as part of your journey. But really, all of that really grounded the enterprise security layer and catalog and enterprise content. Yeah, and there's a lot of layers when you talk about control here, right? Like there's the content guardrail layer that you just mentioned, but then there's also the agent's personality and tone, right? And how would you say that what would be the best conceptual way to think about controlling these types of experiences in your mind? That's a great question. So today, if you were to hire a new employee, someone like Freedom would hand that employee a handbook, and that handbook would include things like, here's our pricing catalog, here's the things we can and can't talk about, here's the specifications for the product. A huge multitude of items that a new employee would need to follow, and it's no different with an agent. An agent needs to be handed a guardrail, an employee handbook, and of course branding and information guidelines that it needs to follow. Of course. So once that UI is styled, right here you can see some components being reused, product tiles, tables, spotlights, AI generated summaries, and etcetera. The agent's able to orchestrate these. And through the guardrails, behaviors, and the personality that we create for it, it's able to essentially navigate, right, all that rich content that's already indexed in Coveo and the same content that powers that search experience, right? So if I were to go back and search on Freedom's website, for instance, or in your earlier example, for that other retailer, we'd see consistent results and recommendations, right, depending on the user queries. Right, and sometimes more importantly than what the business logic and what the agent actually does is what the agent doesn't do. So of course at an enterprise level we can't take any chances with hallucinations, with non respecting of actual business logic, so things that freedom actually excludes from the conversation is sometimes even more important than what it includes. And we know industries like that that have huge repercussions if those exclusions aren't met, right? Yeah, for sure, for sure. Okay, that's interesting. So I want to transition now really into the cut off point here, the evolution of search. Let's bring it home, we discussed a lot of how search has evolved over the last couple of decades, and really what ends up happening is that traditional search box that now becomes a dynamic experience. We go from a place of really static search, static layouts, users have to really dig through information to find what they're looking for, to this, what we just saw with Freedom, becomes a totally different experience. Can you talk to me about that overarching development that happened over the past couple decades? Yeah, for sure. We have Coveo and a lot of other vendors as well. We've been delivering search experiences, discovery experiences, listing pages, recommendations, etc. The whole surface of presenting results to customers at the right time and the right place based on their intent, right? But now we've been able to really expand the set of functionalities that can be offered here, and this brought us to essentially this unified search experience that we've shown you today on Freedom Furniture, right? So here, what we've shown is really, as this slide is saying, right, a way for customers to go from all the way from finding, buying and discovering products all the way to completing their purchase and essentially even we plan to go towards post sale support and other types of use cases to really complete the journey, right, and fully unify that user experience, right? When a customer walks into a store, they know they can ask essentially almost anything to the human on the other end because, worst case, that human is going be trained to know where to go and look for that information. Customers have been trained for decades on how to use search, right? So they come in, they try to narrow down their intents into a few keywords, try to enter those keywords and be as precise as possible. But often, that fails on the first try, right, because unless the brand has very good search and a very good search provider, it's very hard for customers to know how to align their intent to what the brand's corpus of information and knowledge holds, right? So through this agent, we're able to have the best of both worlds, where we can have very fast search, right, that's very efficient, very scalable, and on the other hand, have the agent on the same path, right, and through these multi agent frameworks that we've shown you today as well as these orchestration layers, be able to dynamically adapt and morph the experience based on all of these different outcomes simultaneously. And maybe if you can talk just about the frictionless experience you just showed. It wasn't something that you really had to work to get into. It was almost frictionless in how you went about your shopping experience. Yeah, exactly. And to an earlier point, right, our goal isn't to remove functionality. It's only to make it easier to achieve, you know, that end result that they're looking for. Right? In the end, people, when shopping online, when getting support, they they're always looking to do a particular job or to find something specific. Right? Even if that particular job is just browsing an assortment of products. Right? So in this case, you know, we want to make it as seamless and easy as possible. That's why, you know, we allow for this interaction to happen from search, from other surfaces as well, and, you know, nothing could stop you because of what I talked about earlier around AG UI, A2 UI from deploying this in your own chatbot, in another agent, in a side panel, in a chat interface, right? But consistency is key, and this experience here allows you to get that consistency by providing the same level of recommendation and depth no matter the channel you're interacting in. And this also comes automatically with security and permission for each of the users because in the end what the agent's doing here is really impersonating the user throughout that experience. Right. So let's maybe cap it off by talking about that seamless flow and sort of orchestration that you just went through. You went through a flow that both included regular search as well as that dynamic buying experience that ended with a review summary of the sofa you were looking for. Talk to me about that. Yeah. And as this slide is showing, right, it's really, you know, just to complete the loop, right, it's really about these different building blocks that we've been showing today that amount to the experience that we've been showing today, sorry. So on one hand, right, you have the shopper's input and context, right? So you have that search box that's able to quickly and dynamically upfront detect whether that user intent points towards a traditional search intent, right? This is a search box in the end and customers' behaviors don't change overnight. Or a conversational intent, right? And once you're in that conversational intent, then you can trigger an agent, not wanting to trigger it at all times because in the end, well, we can, but customers don't necessarily always want to make the complete shift because it is more expensive, right, at an absolute level, right, to run these queries. Once you're in that agent, you can actually start selecting these detecting the proper user intent, selecting the right outcome essentially that you want to drive for that intent. And then fulfilling that outcome through the different tools and frameworks that are available, either from Coveo's rich knowledge base that we've already indexed, the catalog customers may have, as well as eventually in our future roadmap, data points, right, such as customers' ERP, OMS, or any other enterprise system that they may have which has data in it which resides and is relevant for the experience. All that closing the loop with that dynamic canvas, right, that's using AG UI from the agent side to have a shared state and to be able to convey what we want the A2 UI library to render on the front end. And it's really that full stack that enables us to deliver this seamless and dynamic experience that we've shown you today. All right, well, thanks, everyone, for being with us today, right? That's really what we had prepared to share with you in terms of what we believe are the five key pillars to building concrete and scalable agenting experiences for enterprises. As a key takeaway, we have a leave behind here around the Agentic Commerce Readiness Checklist, the checklist that the Coveo team have prepared, right, to see if your enterprise is really ready to start adopting these agentic technologies. And we'll see you in other webinars that we host on a consistent basis. So thanks for being with us today, and thank you for your time.