Everyone is talking about "agents," "assistants," and "autonomous AI;" however, few marketers have a good understanding of what these systems truly do, or what they mean for how consumers discover and interact with brands.
In the third episode of The Visibility Brief, Rebecca Colwell, SVP of Marketing at Yext, and Christian Ward, Chief Data Officer, discuss one of the most trending concepts in AI today: agentic AI. They discuss the real definition of agentic AI, how these systems behave, and how marketers can prepare for a world where AI doesn't just answer questions, it takes action.
From executing tasks to personalizing decisions, agentic AI represents a major shift in how brands show up, compete, and stay visible today.
Christian and Rebecca break down:
What agentic AI actually is. Christian explains what agentic systems are and how to distinguish simple automations, guided assistants, and fully autonomous agents, and why that difference matters.
How AI agents will reshape consumer behavior. We explore the emerging reality of AI assistants researching, recommending, and even completing tasks for consumers, and what that means for visibility throughout the buying journey.
Why structured data determines your visibility. Recent Yext research across more than 6.8 million AI citations reveals a huge finding: your website, and how you structure your data, is now the strongest signal large-language models (LLMs) use to answer questions about – and eventually take actions on behalf of – your brand.
How memory and personalization shape results. Christian unpacks how AI agents leverage chat history, use location, loyalty, and context to build preferences on behalf of users to determine which brands to surface.
How to evaluate "agentic" tools. Christian shares the key questions every marketer must ask their technology vendors to separate true agentic AI from marketing fluff.
In this episode, Rebecca and Christian provide a practical, non-technical guide to understanding agentic AI, and how it will fundamentally reshape search, loyalty, and the customer journey.
If you're a marketing or digital leader preparing for an AI-driven future where AI agents actually take action on behalf of consumers, this episode is for you.
Episode Links
Transcript
Rebecca Colwell (00:02) Here to help me unpack all of this is Christian Ward, our chief data officer. Hi, Christian.
Christian J Ward (00:07) Hello, how are you Rebecca?
Rebecca Colwell (00:09) Great, great, thanks so much for joining. I'm really excited for our conversation today. ⁓ So, agentic AI, total buzzword, right? I see it everywhere, but I'm actually not convinced that when people use that phrase, they mean the same thing. So, when people talk about agentic right now, what exactly are they referring to? And ⁓ how would you explain the difference between a chatbot, an assistant, an automation, an agent, ⁓ to someone who's not really in the tech?
Christian J Ward (00:36) Well, first off, think no one has a perfect definition right now because everyone's running in a thousand different directions. At the same time, I think it is really helpful to come up with some sort of taxonomy, at least in your mind of how these things relate to each other. I tend to think of it in sort of very basic terms, which is if you start with automations, those are the things that we've been doing for a very long time. Almost any programmatic control of a process is an automation.
And I think we've lived in that world for a couple of decades. So people know that that is a thing. ⁓ Sort of next to that is this idea of the AI assistant. That's what you see a lot of times if you're talking to your chat GPT or talking in a dialogue where it knows things and you can even ask it to do things. And then an agent, when you really think about an agent, probably the biggest distinction is autonomy. So when we talk about what's an agent, an agent is something that should be able to go off. do work like legitimate work for you. Now that could be it's running several automations. It's making decisions of is this ready to be done? Is it not ready to be done? So it's sort of like if you had this autonomy layer and the reason why I separate them is I think you really what you're going to see is a lot of platforms like Claude code and many others. They're really built on this idea that you're talking to the assistant and then it calls agents to do things.
And so it gives you this sort of stratification, which I think for many reasons is so important because it's sort of like how we have an org chart. I'm allowed to do certain things at work. I'm not allowed to do certain things. And same for you. And we all have roles and responsibilities. That's a great way to think of the assistant as a manager, watching the agents and making sure they're doing the things they should be doing. But that's one way to frame it. I think people will continue to evolve this, but at least right now, that's very much how we see it.
Rebecca Colwell (02:26) If we were to put this in practical terms for a consumer application, ⁓ Google recently announced that users can book event tickets or wellness appointments. I know they've had a restaurant researching and booking option for quite some time. What other use cases from a consumer perspective might be coming for agentic AI? which ones do you think people are going to adopt first?
Christian J Ward (02:50) Yeah, I mean, honestly, at this point, almost anything that you're used to doing via website. So all the things you just discussed. So booking, finding information, finding information is probably the most basic agent. We don't think of Google search that way. But today, if people haven't tried this, you should ask some one of the main consumer AIs to go research something for you, like a car purchase or, you know, furniture for a dorm room, whatever you want.
But it's absolutely amazing how long you have to put it in thinking mode. It really does that research. So that's an agent in its own, in its own right. ⁓ because you're giving it permission and the autonomy to peruse the web and come back. But so that one, think everyone's already kind of experimenting with, but in terms of the actions, when we think of then taking an action on your behalf, not just in knowledge, ⁓ but also in activity and actually engagement, I think all these elements of things that booking hotels, booking rooms, booking a menu, ordering food, anything also that if the agent knows about you, meaning it has the context of what you like and you don't like, it's pretty fascinating when you tie memory of these agents to what they can do. And so I would expect almost anything that people are very comfortable booking or engaging with online, they're comfortable doing here. I think it gets much more murky outside of the customer journey where you're talking about compliance.
You're talking about ⁓ responsibility within organizations. B2B is a little different. But with the customer journey generally, I think everything you've been doing for over a decade is very likely to be pseudo-agentic very soon.
Rebecca Colwell (04:28) So there's a future where my agent could book my travel for me. So I have to come out to the New York office. It knows that I like Alaska Airlines and a window seat, and it's just going to take care of it based on the schedule that it knows I like.
Christian J Ward (04:40) I think that's exactly what's happening. And on top of that, I think what's very likely is similar to if you were speaking to an assistant, the assistant is the one you're talking to. They are then going to talk to an agent that does hotels, and then there's a different agent that does airlines. And yes, you might merge them all. But one of the benefits of separating agents into very specific areas is then it gives the assistant and you the ability to critique ⁓ exactly what was done in a more chunk or chunkable format. Meaning, if you look at the whole trip, you may not know which part of an agent that does everything made a mistake. But if you know the airline was wrong, and there's one agent for that, now you know who to deal with. Again, this is why I really think even though people don't love org charts, it's very much a design around, hey, this is an organization. It's working for you. And you need to know and to be able to pinpoint where something's working well and where it's not.
Rebecca Colwell (05:35) Interesting. Do you think that there are other ⁓ applications that might be coming? Because I think booking airline tickets is annoying. ⁓ Are there other things that it might be able to help me with?
Christian J Ward (05:46) Well, in many cases, when you're talking about buying goods and services, the thing that we see is that the input of memory along with these agentic capabilities or these autonomous states ⁓ really gets very interesting because for one that came up when we were traveling an awful lot was this idea of loyalty and engaging with loyalty programs. I think that is a huge opportunity for most brands. Loyalty was really big maybe a decade ago and I feel like it fell off because nobody wanted the keychain with a million little things, trinkets on it, you know, because you have to scan to be loyal. I think all that goes away. I think the reality is, is loyalty and engaging in loyalty are things that you do on a regular basis is a wonderful place to allow agentic processes because there's enough data and enough context because you've engaged with that brand enough times.
Rebecca Colwell (06:17) Yeah.
Christian J Ward (06:36) that there's some really good stuff there. Autonomy is a little nerve wracking when you've never done it before. So for example, you hand your keys to your teenager to drive for the first time, right? You're giving them autonomy, but the risk is very high because they've never driven in this area before. But if they've been driving in this area for 10 years already, then you're like, you're probably fine driving in this area. So that's a little way to think about it again, which is the memory, the context, the understanding, those will get better. But for brands, loyalty is a big opportunity.
Rebecca Colwell (07:06) Right, right. So what does the brand need to do to show up in these searches? I mean, now it's trying to communicate with an agent instead of the end consumer. So I imagine the strategy is a little different.
Christian J Ward (07:17) Yeah. Well, for most brands right now, one thing that we're seeing in all of our research, we had done the study on roughly 6 million citations and that continues. think we're up to like 24 million citations in the data set now. And so we'll keep publishing on this, but what we see tends to be holding steady, which is you should have highly structured pages of data because you can control your own website.
And the website is the number one cited source for anything around brand loyalty, engagement. really quite powerful. I see this, it's sort of this fascinating dichotomy, which is in one way, less humans are going to your website, but in another way, the more you put knowledge structured on your website, the better you engage with the humans through the AI or through search. And so this will be a little uncomfortable, but I think the reality is, is if you structure that data, let's say you build a page on being loyal in Jacksonville, Florida for a particular brand, why does that matter?
Well, because for the person asking the AI, remember, they're in a place. And it always helps that if you're structuring data in these chunkable formats where you're sort of powering that answer. So for example, I can't fly American out of my airport. It's not here. But knowing my location makes that really important if I'm an American Airlines loyal customer. So it's got to get me to another airport for me to take benefit of that. These are those nuances that if the structured data from the brand is very clear, The AI can leverage that to provide a better experience to the consumer.
Rebecca Colwell (08:50) Totally makes sense. ⁓ Let's shift to the workplace really quickly ⁓ for our final segment. Every single application that I use today has now rebranded as a Gentic and just like, this, is this marketing hype? Is it actually real? I tend to be a bit of a skeptic. So I am curious on your perspective. Do you think these applications are truly a Gentic? ⁓ Do we still need a human in the loop? And ⁓ What do you think it will take, at least in the workplace, for us to trust these agents? What is it going to take to get us there?
Christian J Ward (09:28) Yeah, there are rumors that Gemini 3 is coming out this week. And I think you'll see a pretty big leap in some of the capabilities I'm about to discuss. But I think generally when we think about in the workplace, number one, think agentic is sort of, it's a horrible buzzword at this point and everyone's calling something agentic. That being said, I think most of the things we're saying are agents are really automations with a slight layer on top of them.
But I expect that to be kind of how it needs to be in the workplace for a while, probably even for the consumer journey. And you can actually see this happen. So for many people who haven't tried Clog Code, Clog Code actually, it's very much an assistant, but it has two modes. It has the action mode and a planning mode. And one of the reasons why I would bring that into this discussion is we find that if something's truly agentic, it runs off the rails really fast. Like anyone that's tried Clog Code, If it's in action mode, it'll write you a million lines of crap really fast. So you actually stay in planning mode, which is more of you overseeing what the plan is.
And I think in the workplace, particularly for businesses or anyone dealing with the type of work we do, where they're trying to make sure, how am I showing up? What am I doing? How do I structure my data? It really does help to have that planning mode or to have a team of humans doing that. Cause right now I think to let things be truly agentic, is a big mistake. I'm not saying it's not cool and it sounds cool, but it's really not the right way to do it. So if you're talking to any vendor that says, it's agentic, ask them some really important questions. Number one, how autonomous is this thing? Number two, what does it show me on how it's doing or what it's going to do before it does anything? Number three, based on its behavior, what changed and can I sort of audit any of those changes very quickly?
And number four, and probably most importantly, will it unveil or reveal to me every step in the process it plans on taking? If you can't do those four things, you get clear answers to them, I would not go full agentic. I would ask for a different version of this where you're in control of it. You can actually see this in almost all the platforms today. If you use Google or Gemini inside of Google Sheets, or if you're it in a document, you'll see that moment where it just went way, way further. than what you were thinking and it's not right. It's really wrong. And so I think you're going to see a lot of that in the workplace continue.
Rebecca Colwell (11:58) It's interesting. It's a bit like an early intern where, you know, they're very eager. They want to please you. But if they're sprinting off in the opposite direction or not getting those like little moments to touch base. Yeah, that's that's really interesting. ⁓ Well, Christian, I think this is a great place to wrap it up for today. Thank you so much for your feedback. And before I was just going to ask holiday shopping around the corner. Are you going to be tapping into some agents to do some shopping for you?
Christian J Ward (12:06) Yes. Absolutely, I already have, which is, it's really fascinating, but if people haven't tried this, number one, I think it's so powerful on helping get new ideas or new items or research things for you. But the other thing that I find that's really kind of wonderful about is I convert much faster going that route than I ever did opening 27 tabs on my Chrome browser, hoping I'd find something. So it's really, really powerful. I think we're going to see a massive shift. with Black Friday in terms of maybe less traffic, but a lot more people coming from these AI assistants and actually checking out. So we'll see, but I'm pretty sure that's what the data is going to say.
Rebecca Colwell (13:03) You know, that's such a good point. There were a couple of things I'm trying to put on my Christmas list right now, and I'm going on the rabbit hole of over researching every single possible thing. I think I'm just going to ask my agent to do it.
Christian J Ward (13:15) I think you should. one thing I will tell everybody that's a fun trick is try to tell your agent, put everything in a table with the pros and cons of each product. And when it's in a table, what I find is the AI tends to get much more concise in giving you the reasons for and against. But what would have taken me 20 hours in buying my teenage daughter's gifts now takes six minutes. So it's really, really beneficial to ask it to do that too.
Rebecca Colwell (13:44) Fantastic, and I love a good pros and cons chart. ⁓ Wonderful. Well, thank you so much, Christian. If you like what you heard today, subscribe, share it with the friends, or head to yaks.com to learn how we help friends stay visible in an AI-first world. ⁓ We are going to take a break for the Thanksgiving holiday, but we will return in December with more insights to take you ahead. Thanks for listening.
