The Future of AI in Marketing

Tom Nixon (00:01.602)
Curtis, as we welcome people back here to Bullhorns and Bullseyes, we are one day removed from getting the chance to see each other out in the wild in the real world outside of a podcast outside of a screen.

Curtis Hays (00:10.906)
We do.

I have picture proof that we got to see each other out in the wild here recently. Great event, by the way.

Tom Nixon (00:19.786)
Yep. It was. So listeners might remember we had Doug and Meghan Mans. Has that already aired? Yeah, that’s right. That’s right. Yes. So that’s right. So it’s something about that client that gets us out behind our desks and we were in the real world meeting with real people. It was a great event. Congratulations to them again, celebrating their 125th anniversary.

Curtis Hays (00:29.479)
yeah, yeah, last couple weeks ago. Are only in person.

Tom Nixon (00:48.536)
But speaking of the real world, we are going to talk about something that is seemingly more and more real feeling like the real world, but still has the word artificial in its name.

Curtis Hays (01:00.772)
Yes. A lot of people have asked us to talk about this topic, talk about AI and bring on some experts. And just so happened, I had Carl Wines who we had on the show earlier this season, who’s got a lot of background and experience in e-commerce and building brands and startups. And

Carl introduced me here to Jesse Flores and Jesse is the founder of SuperWebPros. And we got to talking, had a great conversation with Jesse after that introduction. And he got to learn a lot about me and I was like, I didn’t really get to learn a lot about you. We ran out of time, had to kick up another conversation a few days later and I got the opportunity to ask him some questions about what he’s doing.

Jesse Flores (01:44.569)
You

Curtis Hays (01:54.113)
over there at super web pros and showed me some demos of some conversational AI tools that he’s got. We talked about, you know, the AI marketplace and what’s happening. And then a few ideas that we’ve been playing around with Tom. And, so now we’re working together. So we’re going to bring on Jesse who’s helping us implement some AI inside of collideascope. And, it’s great to have him on the, on the show. So welcome, Jesse.

Jesse Flores (02:21.015)
Hey guys, how are you? I’m super excited to be here. Thanks for having me.

Tom Nixon (02:24.59)
Yes, everyone’s favorite topic from a marketing standpoint. AI, right? And it seems like there’s something new to talk about every single day. I’m old enough to remember the Internet Revolution. Curtis, do remember when they called it that? And we were surfing the World Wide Web. my career is seat. Right. I remember why one case. that’s it. It feels like it.

Curtis Hays (02:25.19)
Of course.

Curtis Hays (02:39.206)
I remember Y2K, I mean that’s how old I am.

Jesse Flores (02:42.324)
Hahaha

Jesse Flores (02:47.097)
You

Tom Nixon (02:49.486)
But you know, so our careers as marketers has gone through evolutions like this in the past, but I don’t think Jesse anything has ever happened this fast this dramatically with as much impact as we’ve seen in the last what 18 to 24 months?

Jesse Flores (03:04.149)
Yeah, I think you’re probably right. I I also remember Y2K as well. You can tell by the gray in my beard for those that are watching this on a video. Yeah, I can’t think of anything else that’s moved this quickly. And I think it’s because there’s this kind of nature in which data compounds, And particularly artificial intelligence leads to compounding outcomes. So I can’t think of anything that’s moved that quickly either.

Tom Nixon (03:29.55)
Do you remember, can you pinpoint a time in which, know, maybe the started a few years ago, actually, where we started to see some of this stuff roll out. Do remember this first like aha moment? Like, okay, wait, this is different. This is something I got to be paying attention to.

Jesse Flores (03:43.541)
Yeah, you know, it’s funny because it was actually a buddy of mine that was doing research in machine learning at Georgia Tech that turned me on to chat GPT when it first came out. And he was like, Hey, you need to take a look at this. It’s now kind of available. And, know, before I had my company, we were doing a lot more software engineering. So I’ve been paying attention to large language algorithms for a long time. It was only when

open AI, so people think about the chat GPT moment, but really what happened at that moment was it just kind of became accessible to everyone in a way that was easy to distribute and the economies of scale made it more or less affordable. But I’ve been thinking about this stuff for a very long time. And so my buddy was messaging me and is like, Hey, you’ve got to see this thing. It’s incredible. And that was, know, chat GPT with 3.5 or something, which had a

ton of mistakes, so people that are still complaining about AI or chat GPT today, like just imagine 3.5 was even worse. And I remember looking at that and it was one of those things where I was like, okay, this is really here. But for me, when it became really clear this was valuable was when you were able to start.

connecting those models to your systems via APIs and starting to augment these kind of fundamental large language models in really kind of easy ways to be able to create totally new experiences. That for me was when it went from like, this is really cool to this is going to be transformative for everybody.

Tom Nixon (05:21.964)
Yeah. Curtis, same question to you. Do you remember like was there a moment when you’re like, holy moly, this really does change everything.

Curtis Hays (05:29.102)
Yeah, I remember playing around with it at first, just having it like write me poems. I had to write a poem about my dog, you know, and then and then a poem about Stephanie. And, you know, just trying to give it some backstory and background and see what you know, kind of creative ideas it would come up with. And I think like a lot of people, it was was more of our like creative writer in a way. But yeah, I mean, same thing with Jesse. Now I’m starting to see this this

I would say within, it’s been about a year now where to isolate what the agent is looking at, what the AI is looking at and have it only stay with inside those guardrails and have it interpret data or interpret transcripts or, you know, know some sort of information. It’s really powerful at

the retrieval it can do. And honestly, for us, I would say one of the simplest things has just been starting to transcribe all of our customer meetings, have all those transcripts documented, and then having the AI agent having the ability to recall information that’s in those conversations, which has helped me countless times and the amount of time I’ve saved and having to manually go and potentially retrieve that, or it’s just lost at the end of the meeting. It’s like, I don’t remember what we said.

or what we decided, but well.

Jesse Flores (06:57.653)
Or if you’re like me, like I’m probably the worst person in the company at keeping our CRM and our systems updated. And so I’ll go on with a meeting and I’ll promise a bunch of stuff where I’ll talk about a lot of things and then forget to tell my project manager, by the way, I committed us to this and this and this and this and this. And now she can actually go and talk with our transcripts and be like, you didn’t tell me we told them this would be a seven day delivery for, 60,000 hours of work. yeah. So that’s kind of where they found it really valuable is it’s that ability for.

Tom Nixon (06:57.751)
Yeah.

Jesse Flores (07:24.373)
everyone to kind of triangulate around a source of truth using this information or using this technology.

Tom Nixon (07:32.398)
Yeah, absolutely. Curtis, you mentioned the word agent, which is, think, like the next level for me. It’s like you called me one day. I’m paraphrasing. And you said, you know, websites aren’t going to be websites anymore. And I said, what do you mean? Well, think about what happens now. You go to a website, you’re looking around, you’re trying to get information, you’re trying to understand what this company does. You’re trying to determine is this even the right fit for you? And you’ve got to do some reading and some clicking. And you said that’s.

Not tomorrow, but someday soon, that’s going to be a totally different experience. You’re going to come onto a website and you’re going to chat to an agent and an agent might actually look like a real person, talk like a real person, and you might just have a conversation. So this is like now blowing my mind because I can see this coming really quickly. And I’m curious, where are we on that time horizon?

Curtis Hays (08:19.166)
So for me, it started with Google marketing live and Google marketing live. I thought long and hard about where we were really at and where Google was going and, in the AI space. And I think that was VO three that they released and, was it VO four? don’t, I don’t know what version they released that day, but you know, just the, the power, in the AI tools at that moment and said, we’ve got to do something about this. And then as I toyed with it, and then a few months later, Matt.

Jesse, we started having this conversation. said, hey, Curtis, actually, let me show you something. And I think it was Comet that you first brought up. so I’ll let Jesse explain the specifics. But he brings up this browser. And he goes to my website. And he starts having a conversation with a browser about my brand. And it’s repeating extremely accurately. And all he’s doing is sitting on the home page, asking it very specific questions.

And I was like, holy cow, how are you able to do that? so conceptually shows me that and then shows me plugins and tools and those types of things that you’ve built, Jesse. So let’s, let’s kind of share a little bit of that, that story and journey with me and what, what you’ve seen and what you’re doing now.

Jesse Flores (09:37.047)
Yeah, so let me take it high level and then I’m happy to share some of that. So I think Tom, to that point that Curtis, because he and I spent a long time talking about this idea of what we call the agentic web. It’s kind of only a historical accident that the internet is document heavy and text.

And what I mean by historical accident, is that, you the early days of the internet, I mean, you, think all three of us might remember what the internet looked like in 1995, right? Those really ugly text documents back in those days, the, in fact, what they call the document object model, the internet was just about sharing documents in this kind of networked way. Now, over time that evolved to what became CSS.

and HTML that gave you a presentation layer. And then JavaScript gave you an interaction layer, but all of this still remained very text-based. Now, as compute got cheaper, you started to see more media. think all of us remember when images used to download pixel by pixel on America Online, right? But eventually images would show quickly and then video could stream. So as the piping has gotten faster, multimedia began to be able to keep up.

Tom Nixon (10:43.598)
Mm-hmm.

Jesse Flores (10:55.629)
But under the hood, all of the data moving back and forth, if you look at any given website, 85, 90 % of the content on a website or the code on a website rather, not the content, the code on a website is built for you and Curtis and me to think it looks pretty. Almost none of the information or a very small amount of the information is the actual content that machines need in order to make sense.

of a website or an experience. And for anybody that develops, anyone that’s familiar with APIs, you see this whenever you look at an API. If you look at an API request, you’re going to see what’s called a JSON object with a bunch of small little brackets, and it’s very thin. And if you were to look at the number of characters or kilobytes a JSON object takes versus a website, it’s dramatically different. And so when Curtis and I first started talking about this, it became very clear that we as humans

love to look at stuff. like things to be pretty. We’re human. That’s how we’ve evolved. But machines don’t need any of that. What machines just need is that pipeline, that very clear data pipeline. And for a long time, we’ve kind of built this network only for humans because only humans could kind of make sense of the documents that you saw online. But with the rise of AI,

agents and AI that can interact intelligently back and forth with content, particularly with structured content, it became very clear that there will be two highways on the internet before too long. There’s going to be the highway for humans. So I grew up in Houston, Texas, mega highways, right? You got these eight line highways. love the cowboy hat. You got these eight lane highways, right? And then got the little feeder rope, right? And the little feeder of both of them will run side by side and get you to the same destination. But

Tom Nixon (12:40.216)
You

Jesse Flores (12:49.933)
For a long time, we’ve only had kind of the one highway. Now we as humans, actually think we’re on the super highway. We are definitely not on the super high. We are on the feeder road because you can run multiple AIs on that one mega highway running very, very fast processes in parallel. And that’s where the web is evolving. One to where the pipelines are going to be.

Twofold, you’re have humans kind of navigating some stuff and AI is navigating a lot of things. And the way I describe this to people is if you look at your browser and look at your tabs, all the tabs you have open, all the things you need to come back to, those are all things you as a human who work in linear time need to get back to at some point. But what if you didn’t?

What if you actually opened up the tab and you could say, do this job for me, open up another tab, do this job for me, open up another tab, do this job for me. Five jobs can be happening at the same time that you’re focused on one thing, five things can be happening. And when I was talking with Curtis about was that these browsers like Comet are starting to move in a direction where you can interact.

with, with content in multiple tabs, each doing something different while you as a human are doing whatever you choose to do, reading your email or, or, you know, reading about the writer cup. while, you know, the other things are, are, are running all in the background. So that’s where we see kind of the agentic web moving one where it’s not just humans. it’s going to be humans and AI and these data pipelines will be two parallel highways. One that is visual for humans, but one that.

Tom Nixon (14:06.595)
You

Jesse Flores (14:24.73)
exists on the same plane that is only used by AI.

Tom Nixon (14:30.626)
Talk about then back to the human experience, right? So how your the website of the future, let’s say, will be an agent experience for the in very personalized, right? So if I come to your website and I want one thing and Curtis comes to your website, wants something completely different. You don’t have to change websites. You just have to have the agent figure out what the person is there for. Right? So how is that going to work in practice?

Curtis Hays (14:52.838)
Let me add to that time that we’ve mentioned this a few times on the podcast that that that I have the belief that the AI which has memory on the users will use that memory that personalizes that experience. So like what Tom said to his experience on that website is going to be different if you are using personalized AI like my own chat GPT which knows I’m married with two kids.

I live on eight acres, I own a tractor, I own chickens, I do gardening. Like it knows all of these things. So my experience in working with that AI agent is completely different than Tom’s. These browsers are personalized as well and have that same opportunity to do the same thing.

Jesse Flores (15:36.749)
Yes. Yeah, I think that’s exactly what’s gonna happen. So I think there’s kind of three layers here to answer your question, Tom. So the first I’ll take Curtis’s is that, our browsers already know a lot of things about us. If you look at your browser history, whether you want to or not, I don’t know. But if you look at your browser history, like there’s a lot of stuff in your browser that Chrome or Comet, whoever just kind of knows about you.

Which by the way, I tell this to clients all the time when they’re asking about search engine optimization. It’s part of why SEO is so difficult these days is because even what you get back in Google is somewhat determined by what your browsing history is and where you are and how you’re coming to it. There’s a remarkable amount of personalization. So on the one hand, there’s gonna be a lot of personalization that’s just kind of built into the browser.

The second is, think when you get to a site, what’s going to start happening is your website, which today we call web pages. We like to call them, they’ll be web pageants. You’ll have pages that are agents that are able to adapt based on the interaction pattern of the user. So what people are used to today as a chat bot, which is kind of an appendage bolted on to a website.

We think that’ll actually move into the website itself and the buttons where you click learn more, for instance, on a website or, yeah, learn more is not just going to take you to a page. It’s going to take you to an experience that says, Hey, I would love to tell you more. Tell me a little bit more about what you’re trying to do, or maybe it takes local storage.

from your browser and is like, this guy just came from this third running store. All of these asking about marathons, he probably wants to know about marathon training. So when he clicks that learn more, it’s going to be like, have you been interested in marathon training? Let me help you walk through that experience and start to bridge that gap. The third thing I’ll say about this is there’s an assumption in your question that you even go to a website.

Jesse Flores (17:30.743)
there will be an element in which the web will begin, will continue to become multimodal, right? So we all will joke about how terrible Alexa is, and I she’s about to go off my office now, but we all joke about how terrible she is. And she was like an early, early, early, early prototype, Siri also as well. But this idea of us having to go and navigate,

Tom Nixon (17:50.7)
Also terrible, sorry.

Jesse Flores (17:55.761)
only to a website or to interact with the business only through their website, only through a browser is also going away because as these agentic layers get built, so Google’s working on this thing called the A to A protocol, which effectively allows for agents to communicate with each other with low latency. What’ll happen is all of these machines that today are not really networked will start to become networked. So what I imagine a future experience will look something like this.

actually one of the times I was talking to Curtis, I think there was a leak and he was trying to, to, to take care of some stuff that was happening. Right? So here’s what will happen. Your, your, your water heater will leak. You will come downstairs, your basement will be flooded. You’d be like, my gosh, I’ve got to take care of this. You will then go to your personal assistant. Maybe it’s Siri, maybe it’s Alexa, maybe it’s 11 labs. It could be any number of things. And you’re going to be like, Hey, we’ve got this thing broken. I need somebody to come out here to, you know, as soon as possible to come out here, fix the thing and fix all of the flooding and blah, blah, blah.

That one request will go out to probably half a dozen local, ServPro, whatever companies can do that. That agent will broadcast to all of those agents in parallel and be like, hey, here’s what happened. Jesse just had this problem. Here’s what’s going on. Can somebody be there within, what’s the soonest somebody could be here? First guy says, I can be there tomorrow. Next guy says, I can be there in 30 minutes. Third guy says, I can be there in three days. And then it says, okay, great. What are your rates, right?

That kind of experience, which right now today would be, making phone calls, I’m on Yelp, I’m going to browsers, I’m Googling stuff, I’m asking ChatGPT. That won’t be the case in five years. Each of those experiences, each of those websites, what we call a website today, will be an agent that’s receiving that from your phone or whatever your agent, your client is, and ultimately responding back.

and you’ll just get an alert that says, hey, these guys can be here in 30 minutes. They’re going to cost X. Those guys can be there tomorrow. It’ll cost Y. These guys can be here in two weeks. It’ll cost Z. I think you should go with number one. Do you want me to go ahead and book that? And I’ll say, sounds great. I’m over here scooping water. Please just take care of it. And then I’m going to get a knock on my door in 30 minutes from, you know, Serve Pro or whoever. And they’re to be like, hey, we’re here. We’ve got the truck. Everything’s ready to go.

Jesse Flores (20:19.479)
let’s go ahead and start getting this taken care of. So even the underlying assumption that it will be a website that I go to, think is also, I don’t buy that premise in five years.

Tom Nixon (20:31.79)
Hmm. Which raises a lot of questions. I don’t know if we could get to them all today, but like I’m thinking, then will we have websites or will we just have intelligent agents? You know, so I mean, it’s going to be interesting to see where it goes. You just described a scenario five years from now. What should companies be looking at now in the next 12 months, 18 months to say, I can’t wait for five years to figure out what the ending of the story is. I need to start doing this now. What should we be doing now?

Jesse Flores (20:48.515)
Yeah.

Jesse Flores (20:55.959)
Hahaha

Tom Nixon (21:00.406)
as marketers, as business owners.

Jesse Flores (21:00.535)
Yeah, so I think right now it’s, yeah, great question. So I think right now it starts with bringing some of those agentic experiences into your current website. By the way, I don’t think websites go away. I do think that there is an element of,

where we as humans, all of us have different temperaments. Like I’m very much the kind of human that’d be like, yeah, I trust the ad, tell me whatever. Some people don’t have that temperament, right? They’re gonna wanna trust but verify, right? And that’s probably the responsible thing to do. And so I don’t think websites go away. I think that the interaction just becomes, going back to my highway comment, it becomes…

two tracks. It becomes the agentic track that can be used in some instances, and it becomes the human track that is used in other instances. So in the short run, part of what we’re doing and working on with clients is building out what I called these pageants, these agents that are able to do specific things with and for humans bolted directly into the website. So for example,

we’ve started releasing conversational lead magnets. So we might send someone to a squeeze page like we used to do in the old days, but instead of when you click the button, it popping up an email, it just takes you to a conversation. And that conversation will start to naturally try and figure out your pain points and at the appropriate time, decide whether or not to get your email address and whether or not to book a call and or whether or not to refer you somewhere else. Right? So all of those things that would currently take up some time. Now the agent is just taking that

taking that lead from what is a traditional kind of a squeeze page. Like that’s something we’re doing today. For other customers that are in e-commerce, for instance, we’re building what are called AI product guides. And so here, particularly when you’ve got a large catalog or you’ve got complex sales cycle.

Jesse Flores (22:47.481)
Um, there can be a large gap between I’ve got a problem to solve, or I know I want to, I know I want to be able to run a marathon. I’ve never run a marathon before. How do I put together a training path? A training path? Well, in the old days, you might’ve had a blog on, how to run a marathon or train for a marathon, or there’d be a whole content thing on that. But I could today, and we’re working on with this on a running company, go in this click, you know, uh, I want my, you know, my, my couch to 5k guide. I clicked that thing and it says, you know, super excited for you to, take this journey. Um, have

ever done anything like this before? No, never. Okay, well let me get you started here. Where are you in the journey? I’ve never run my day in my life. Okay, let’s put together a plan and by the way, in order to start this plan you might need these walking shoes or these running shoes or you might need this gear. Here, do you have any of this stuff? Nope. Okay, well guess what? You can just click add to cart from inside of that conversation. Well we’ve got another client.

that is targeting caregivers of people with dementia. And they’ve got tons and tons and tons of resources. And so I can go to their website, this is the Greater Lansing Care Foundation. I can go to their website and I can start asking questions about like, hey, I just got told this, this and this from the doctor.

What should that mean? Or, know, mom’s acting kind of strange, you know, or here’s a list of common behaviors that might be going on. Does this, any of this apply? You tap the button, that one looks like it applies. And now it’ll start walking through, man, I’m really sorry to hear about this going on with your mom. You know, let me ask you a couple of questions. Boom, boom, boom, boom, boom, boom, boom. Right, so all of these experiences are starting to bridge that gap between all of this content and data that exists over here and humans who…

who want or need to close that gap.

Jesse Flores (24:27.501)
But do it in a familiar way where you aren’t looking at the blank screen of chat GPT. You’re on somebody’s website with their expertise and you’re interacting in the way you normally would. You’re clicking learn more. You’re clicking help me. You’re clicking whatever buttons you’re used to clicking. But now instead of going to a page you’d have to read, which nobody wants to read anymore, you’re now engaged in a conversation that’s making product recommendations, recommending resources, qualifying the lead. And that’s where we see it happening. Not in five years, but like.

today, we’re launching these today.

Tom Nixon (24:58.06)
Yeah, in a branded way, right? Do it. So you make the point. It’s not just chat GPT. I think we’re going to laugh someday. What chat GPT look like, you know, when we think about how AOL used to look like it was so simple, but it’s effective. Speaking of one, two K or why one K Curtis talk about something that’s come full circle. Remember when like chat bots were like all the rage until companies realized that I don’t have anyone that can run this thing.

And when I try to program some logic into it, it’s a complete horrible experience for the user. Let’s get rid of it. Now look where we are. like chat bots are all the rage, but it’s just a different animal now.

Curtis Hays (25:35.527)
Yeah, we have come full circle that the chat bot of the past was based off of a bunch of if then statements. So if the person had this answer, then go down this path and you had to build these trees to anticipate what the conversation might be and create all of that logic. And that took a lot of time. I think, no, and people knew that they were talking to a bot.

Tom Nixon (25:58.095)
It never fully got it right 100 % of the time.

Curtis Hays (26:04.802)
and wanted the personalization. And so it seemed novel at first and had some traction at first. But then people realized that I actually want to talk to somebody and get an answer much quicker, sort of like the attendant, where if you call Xfinity or, you know, some provider and you’re taken through 10 prompts before you can actually talk to somebody. And, you know, you hear somebody’s like, give me a human.

I want to talk to a human, you know, and it’s like, yeah, representative representative.

Tom Nixon (26:35.448)
Speak with a representative.

Jesse Flores (26:37.175)
No.

Tom Nixon (26:39.062)
You’d like to know your balance. No representative directions.

Curtis Hays (26:42.864)
So if the, I think there’s two parts to this. One is you want it to be conversational. You want it to behave like it’s human, even though it’s not, but you also want it to be on brand. And this is coming off of a conversation we just had with Nolan about Paradox and Olivia. So Paradox is this AI tool for recruiting and talent acquisition where this

AI agent stays with you through the hiring process and you have this opportunity to customize this HR agent. Give them a name, give them an avatar, give them a personality and that personality I think, well, it needs to align to your brand. Otherwise that Olivia is no different than somebody else’s Olivia, right? And so number one, that’s why you and I are talking a lot about brand with companies today and making sure they have alignment.

I think the other critical part for Jessie that I want you to just talk about briefly and what I’m starting to understand as I play around with AI is this need to document and really clean out or figure out your data and your processes and all those things, right? That you cannot just jump into AI and like…

Jesse Flores (28:00.665)
Yes.

Curtis Hays (28:05.186)
throw a bunch of documents and spreadsheets and think that it’s going to, you talked about structured data and those types of things that you have to take the time to structure everything. Once you do that and you get it structured, then you can get it into these models, these APIs, or you haven’t said MCPs yet. So we’ll throw out the acronym MCP.

Jesse Flores (28:26.793)
Yep.

Curtis Hays (28:29.732)
But that allows the agent to actually communicate and talk specifically to, there’s a term now I learned is called a RAG agent, R-A-G, where it’s an agent that is isolated. It only has this information to work from that can’t go out to the internet and find some other information, right? And then come back with an incorrect answer. It could if you wanted it to, yeah.

Jesse Flores (28:50.947)
Well, could if you wanted it to, but it’s still trained on the corpus of whatever the fundamental model is. What it does, so with the retrieval augmented generation, what it does is it adds specialized information or specialized knowledge on top of the default. And then through the system prompt or the directions you give the agent, you might say, hey, don’t ever go out beyond this system. So by default, it will.

Curtis Hays (29:17.042)
And when you comment, like source where you got that information. So now it’s showing you where it’s retrieving the factual information and those types of things.

Jesse Flores (29:19.864)
Yes.

Jesse Flores (29:25.633)
Yeah, that’s exactly right. So when I was talking earlier about, for instance, these landing pages or these squeeze pages, and then the product guides and the one for the Greater Lansing Care Foundation.

Rag is how we do this well. So the way that these processes work is, first of all, we have multiple agents, not one agent. I love what you said about brand. And actually, when you and I first talked, it really got me thinking about this in a new way. Because when you are setting up an agent, and I think this is one of the failing of chatbots, is that you get like kind of one little chatbot that you can put there. But in reality,

your experiences, it’d like you have one page for your website. Like there’s nuance to each thing that you do. And so you might want to have different agents that do different things based on one, the kind of organization you are, and two, where somebody is in that buyer’s journey. And so the way that we engineer these is each agent

has its own kind of system prompt with its own kind of tone. So the Greater Lansing Care Foundation Alzheimer’s agent is very different than Jesse’s. Let me try and get a lead agent. Those two things do not sound the same because they are very different, very different humans coming to those sites for very different purposes. But what they do have in common is they both are using that rag system that you described.

under the hood to source their information. So as you can imagine, when someone’s coming to a site for care information, which we have disclaimers on obviously, but still you want to make sure it is directionally correct and thorough and based on everything that the site has or that all the content that they’ve already created before. And so when somebody is interacting with the Greater Lansing Care Foundation, they’re getting citations as to where that information is coming from. If you’re talking to one of our product guides, doesn’t do me any good to recommend products from somebody else.

Jesse Flores (31:17.997)
I want you to buy my products for my client on their store. And so what we’re doing is we’re augmenting that understanding not only with their content, but we’re also augmenting with their product catalog so that if I’m trying to do couch to 5K, I’m getting the actual, you know, hokas that they sell and not some like Nike that they don’t sell, right? Or in our case, I don’t want it to, you if I’m doing lead generation over here, I want it to have very specialized domain knowledge on how our product guides work. So it doesn’t just like, well, conception

a product guide could do this thing, right? It’s got to be a lot more specific than that. And so the combination of the system prompt, which sets the personality, we’ll call it, of the agent, and then the data source, the rag source that we give it, are what give that agent its uniqueness, and also allows it to have the appropriate voice it needs for the combination of site and buyer’s journey.

Tom Nixon (32:15.406)
And just for the listener who’s wondering if we’re ever getting back to it, what was the acronym? Dijour MPC?

Jesse Flores (32:21.543)
MCP. Yeah, so MCP is how agents interact with other systems. So it stands for Model Context Protocol. This was developed by Anthropic in November of 24, I believe. And what it does is it allows for AI agents to interact with other systems. So for example, let’s take our lead magnet agent that I was talking about before.

Tom Nixon (32:22.967)
CP.

Jesse Flores (32:44.119)
So what’ll happen is it’ll ask you a bunch of questions or ask the visitor a bunch of questions. And if it determines it’s a qualified lead, it will send that information directly to our CRM through, through MCP, which is the MCP connection goes directly to our CRM so that we don’t have to like go and fill out a form or the person doesn’t have to fill out a form more accurately. What’s happening is as the conversation is evolving, the AI is recognizing, I got a name.

Pop it in there. I got an email address. Pop it in there. And so it’s making these decisions and it’s using these tools on its own. It’s deciding when to use a tool to send information to other systems to either our CRM or email service provider or deciding which part of the catalog it wants to query. So you can think of MCP as the way that agents use tools. And this is the way that I describe agents to people that have never used or thought about them. Agents have four properties. They have

a name, typically like what you described it, they have the system prompt, which sets the personality. They have the tools, which is how they interact with computers around the world. And then they have the rag system or specialized knowledge, which is what gives them like the unique piece of information that they need. So that way they’re able to, to act in a certain way, use certain tools and make sure they’ve got the right information when they’re doing it. So those to me are kind of the four characteristics of an agent.

Tom Nixon (34:07.254)
Yep. And I guess I this is my time to be the brand narcissist. Just remind people that the future is probably agentic. You know, the experience on the web. Each one of these agents, though, is going to be a representative of your brand, just as if it were a person that you would hire. So if the customer experience agent is lacking, so too is your brand. So all the more reason, Curtis, why you’ve been emphasizing brand lately is because we have let’s get this right now before we start deploying armies of agents out there.

Curtis Hays (34:33.19)
.

Curtis Hays (34:37.042)
Right, because it’s not going to be necessarily who’s using the best technology, but as we’ve seen in social media with like, you know, that’s just a silly viral video that goes, you know, bonkers online, it’s who’s who’s being creative, right? So I see the same thing with the agent or I sometimes call it the avatar, that if that is unique, and personalized to your brand, here you have this opportunity to differentiate yourself from

competitors, but how do you create that unique avatar or that unique agent if you don’t already know your brand? And so if the future is this and you haven’t already done this over here, how are you ever going to get to the future? Which is why I think brand is now, if you don’t know your brand, you have six months to a year to get that straight or you are going to be in real trouble. That’s my warning message.

Tom Nixon (35:28.088)
Yeah. Yes. From your lips to everyone’s ears. All right, guys, let’s like. Yes.

Jesse Flores (35:33.101)
You know, can I underscore that point really quickly too, because I was giving a talk to a group of HR professionals not too long ago. And this came up, actually what came up was the question of…

You know, the way, a criticism, not an unjust one of AI is, we’ll call it the homogeneity of the data set, right? It’s kind of got one voice and it’s all been trained with, you know, a specific kind of perspective we’ll call it. Right. And so she was asking me, you know, how do you account for diverse perspectives in AI in order to, you know, be able to make sure that there’s representation and all these other kinds of things. And I think Curtis, what you said is kind of what I told her then I was like, you know, I think as the web becomes more agentic.

the agents are where you’ll be, is where we’ll be able to start to see a diversification.

of that’s more representative of humanity than maybe the way that chat GPT always talks. Because as you’re putting together your brand and your identity and you’re coaching your agent to behave in a certain way or speak in a certain way or act in a certain way, which is reflective of your brand, you’ll also be giving it a new reflective experience of and for humanity. So I think there’s kind of this element as well that people have a just question around.

And I think agents are one way in which we will also be able to showcase the diversity of humanity, not to mention the brand that they represent.

Curtis Hays (36:57.382)
It’s a great point and great example, Jesse. Yeah.

Tom Nixon (36:57.592)
Well said. Yeah. All right, gentlemen, let’s leave listeners with a final thought that is very much rooted in today. Like literally today, the day that they’re listening to this. Is there a favorite tool that you guys use? I know you’re in. I’ll call you an eager adopter, Curtis. I don’t know if you’re still an early adopter, but definitely an eager adopter of a tool, a practicality, a method, something that they could go out in either hack to punch above their weight or just implement.

Because it’s going to supercharge their business tomorrow if they do this today. I’ll start with Jesse.

Jesse Flores (37:32.225)
I was afraid you were going to start with me.

Tom Nixon (37:35.048)
They’re good. You’re our guest. I’ll start with Curtis.

Jesse Flores (37:37.691)
Ha ha ha!

Curtis Hays (37:37.895)
Okay, I will say which is hopefully, because I’m maybe more, hopefully more advanced than the average user now, thanks to Jesse, who’s taught me a lot over the last 30 to 60 days. I would say the biggest thing for me inside of ChatGPT was projects. And the example that Jesse just gave with the four things, I won’t be able to repeat the four things, but conceptually, I understand the four things. Those four things are, you can pretty much, I think, all accomplish in a project in ChatGPT.

Essentially in a project it allows you to give a set of an initial set of instructions so kind of like give it that that personality and that focus in the instructions and and then upload files I know OCR, but what’s the correct term Jesse for these these files that the rag files the context files that you up?

Jesse Flores (38:30.573)
Right, yeah?

Curtis Hays (38:33.024)
So you can upload to say this is the background I want you to use. I kind of want you to stay within this context and any results that you produce for me. But you can always upload additional files for it to research, for it to give you feedback on or whatever it is. But now you’re having a conversation with something that has more context and information than just opening up a new prompt and starting a new conversation with ChatGPD. If you start experimenting with that,

And then I would also say the second, hopefully pro tip is your first question to chat GPT should be help me refine this prompt. So I want to ask you this question, but refine this question for me first before I ask it that if you can train chat GPT to get you better at asking it questions, you will become better. I think the power is in the prompt. The power is in the instruction. And so,

You know, learn that, get better at that, but use the AI agent to help you get better at that.

Tom Nixon (39:38.7)
Yeah, that’s really good. I haven’t got a project yet. I have, though, started prompts by reminding the AI of a past experience. Remember when we did this whole thing over here and you ended up changing it because I asked you to do this? I’ll do that same thing with this and you can kind of cut to the chase because they will forget they have good memory, but it’s not perfect. It’s better than mine.

Curtis Hays (39:48.006)
Mm-hmm.

Curtis Hays (39:58.331)
You’re tapping into the memory, but it just hasn’t recalled it yet for that conversation versus getting that memory into a project. And now it has that memory that it’s likely going to use to recall. Yeah.

Tom Nixon (40:03.18)
Right, thanks man.

Tom Nixon (40:10.574)
All right, let’s ask the expert now favorite tool hack. What should we do to mark?

Jesse Flores (40:12.729)
So my self-serving answer is that I think we should go to our websites and check out our pageants because I think these are new and they’re interesting and I think particularly for marketers, this is a way to differentiate web experience in a way that’s really interesting and personal.

I think beyond that, I would say there’s two things that I’ve been doing a lot of lately. One is Syntesia. So this is an app that allows you to record video of yourself and then it turns you into an avatar that you can then make it say stuff and do things. And it took me three or four tries to get it done really, really well, but.

My content team kept complaining that I don’t record enough content for them. And so I made an avatar so they can just like start having content of me out there. And it actually looks pretty good after like the fourth time I did it. It works really, really well. I’d be interested to know if you could tell which ones are me, which ones are not me, because there’s a mix when you look at our profiles. But the other one that I would say is fall, AI, FAL.ai. This is another really interesting thing that we’ve been playing with a lot. does, it has a ton of different

AI models for voice cloning, image to video, video to video. From a multimedia standpoint, it’s got a lot of really, really good.

models. And so for instance, like Google’s Veo three or Veo for whichever one it is very expensive. If you buy with Google, you have to spend 200 bucks a month or whatever it is. But if you go to fall AI, it’s available via API. And so you can, you know, do it for like three bucks a video. And so you can kind of do on a pay per usage, as opposed to having to get subscriptions, all these things, you just kind of work off of a credit basis, and they’ve got, I know, 6000 different different models. And so we’ve been living in

Jesse Flores (41:56.471)
I’ve been living in Synthesia and fall.ai for the last couple of weeks.

Tom Nixon (42:00.896)
exciting and if I ask you in two weeks, it’ll be something different but well, we’ve run out of time so I don’t have to share my favorite AI. I will share an Easter egg though with the YouTube audience, the viewing audience. So, if you’re listening to this on the podcast, you gotta go check out the last two minutes of the video because I there’s been an Easter egg in this episode the entire time. Do you see it Curtis? What is it?

Jesse Flores (42:03.595)
it will be something different.

Curtis Hays (42:22.703)
I do.

Well, looks like an anchor.

Tom Nixon (42:26.958)
That’s all you need to know for now. Alright, we’ll see you guys next time. thank you. Jesse Flores, founder and chief web pro at Super Web Pros. We will link to that, your website, and your LinkedIn on the show notes and to learn more. Come see us at Bullhorns and Bullseyes and we’ll see you next week for another AI infused episode. See ya. Wait, was that was that really Jesse or was that his was that his avatar? We’ll never know.

Jesse Flores (42:47.385)
Awesome, well thanks guys for having me. Take care.

Tom Nixon (42:56.857)
See you guys.