The AI Agents Are Coming

Too many business leaders think they have to solve, like, the big picture of AI when in reality, all you have to do is give your people three to five use cases where they see value in it, and that starts compounding across the organization. This is Revenue Makers, the podcast by Six Cents investigating successful revenue strategies that pushed companies ahead. Adam, are you ready for this? I was born ready. I don’t know what you’re about to say, but I was born ready for it. The agents are coming. The agents are coming. They’re here. They’re coming and they’re here. We were in Boston walking around very close to where Paul Revere was. And so they the British came, and now the agents are here. Now the agents are here. And the agent conversation and the AI conversation in general is a bit of the wild, wild west. Everything is being slapped on with an agent term, slapped on with a copilot term, slapped on with an AI term. And so when you think about getting clarity, who better to ask than our guest today? We have Paul Raitzer who’s the founder and CEO of the AI Marketing Institute. And he has been talking about AI and writing about it long long before anyone was before it went chat g p t, before it went mainstream. He’s not dissimilar to sixth sense who’s been in the AI world for a long time before AI was cool. So amazing conversation, certainly about agents, about what people aren’t using AI for or really thinking about the power we talked about, some of the reasoning and strategy of some of these newer models and kind of where that’s going. And we could have gone for hours. I mean, we weren’t trying to do Lord of the Rings trilogy here, but just so many amazing insights. And possibly the best answer to the most ridiculous thing he’s been asked to do in his career. And so I’ll leave it at that. Let’s jump into the episode and have folks listen in. Let’s do it. Paul, thanks for joining us. Thanks for jumping on. So excited to have you here. And this is gonna be a beefy episode for us today because we could have a topic that no one knows anything about. That everyone’s seeing everything about. We’re gonna talk about AI agents here in a second. But before we dive in, could you talk a little bit just introduce you know, you’re part of the, an interesting organization. Talk about your background. Yeah. We don’t sometimes, we don’t go deep on that, but is it a cool story here in a great organization? Yeah. It could probably be really just a little perspective. So I started a marketing agency in two thousand five, became HubSpot’s first partner in two thousand seven, wrote my first book in two thousand eleven, the marketing agency blueprint, and that was the IBM Watson one on Jeopardy. And so I, out of curiosity, started exploring what is Watson and how can it help with marketing strategy in particular. Could we eventually go for we wanna generate five hundred leads in this vertical. What’s the best way to do it? And I was trying to understand, can this Watson thing do that? Can it actually predict what we should do in, like, a go to market way to actually drive leads and sales? And so I started exploring AI two thousand eleven, two thousand fourteen, wrote my second book, mentioned this idea of a marketing intelligence engine, which was something I had now been working on for a couple years as a, like, a concept to build this thing that would predict what to do with strategies and campaigns. And it was a very small part of a fifty thousand word manuscript. It was, like, one thousand words, and that was all anyone wanted me to talk about. The US government called, asked me to come speak about it. I was traveling the world, like, talking about these ideas. And so we eventually, in two thousand sixteen, created the Marketing AI Institute to tell the story of AI and try and figure out what is it and where is it gonna go. And then eventually, I sold my agency in two thousand twenty one and dedicated all my time to AI when I realized, like, it’s we’re at an inflection point, and it’s gonna change everything. And timing wise, it just worked out a year and a half before chat GBT arrived and changed all of our worlds. Changed all of our worlds. Like, an explosion. Like, some of the dates you gave. Right? Twenty eleven, twenty thirteen. I’m like, I feel like there’s been this explosion obviously in the last one or two, but you’ve been in this for a while. It still seems like the wild, wild west, especially in terms of how folks in go to market are talking about the key concepts that they’re dealing with. Right? Agents, co pilots, models. Like, can you help level set? I think we all really need a level set. Honestly, like, it’s still so new. So if you think about it, like, the moment in time of two thousand eleven, you had Watson, but you also that year had, something called AlexNet where Ilya Sutzkever, Jeff Hinton, two very famous names down in the ice space, and one other guy created a system that could recognize objects, like computer vision, like a breakthrough in computer vision that actually shepherded in the deep learning movement within artificial intelligence. And so two thousand twelve was, like, the watershed moment that most people weren’t paying attention to. Two thousand seventeen, the Google Brain team publishes a paper called Attention is All You Need, which invented the transformer. The transformer is generative pretrained transformer, GPT. So two thousand seventeen actually started this whole origin of the generative AI phase. We’ve had artificial intelligence for seventy years. We’ve been working on giving machines the ability to understand human language, make predictions, and it was being used in businesses in the form of machine learning. Data in, prediction out. So you like, in in marketing, we used to see it in, like, email subject line writing would be like an early example of how we were all using AI. Lead scoring was an early form of, like, machine learning, but it couldn’t think, it couldn’t create, it couldn’t understand, it couldn’t reason. All of that came into our lives in November thirtieth two thousand twenty two with Chatt GPT, was the origin for the vast majority of marketers, salespeople, customer service people, where we all could actually go and use a generative AI tool. And then we had two thousand twenty three. Most businesses are just figuring out what is this? Like, how do we do this? Like, how do we integrate it into our programs? And then two thousand twenty four, you started having more adoption. But, like, we really are only, like, two years into this whole generative AI world where we can now create these things. And now we’re, like, agents everywhere, and, like, nobody knows what they mean and what they actually do, and it is the wild west still. You’re right. I mean, that’s a great transition too because, you know, we’ve we’ve had we talk about it on here, of course. We’ve had guests. We’ve talked about AI. I think one of our AI guests was a year ago, probably completely irrelevant at this point because everything has changed. But, again, obviously, now and we can attest to this in our own world and the marketers we talked to, like, it was content creation. You know, that was the quick one. That was everyone started. And now, at so many different places. And agents now, like, that is becoming the thing. Right? And I think that’s sort of really at least the term got really, really popular maybe towards the end of last year or really when Salesforce started talking about it. But, like, let’s start with the basics. In your mind, in your definition, what actually is an AI agent? An AI system that can take actions to achieve a goal is the simple way. So if you think about ChatGPT and your traditional experience with it, you give it a prompt, it outputs something. It creates something. It does not go and do twenty one steps to send an email. That is action based. So that is I I want you to send an email to this list, and it goes and builds its own plan of what to do, where the list is, how to personalize it, how to do send time optimization. Like, all the things that we do in our jobs, it in theory does. And so the premise of an agent in how it’s currently being understood, and we can rewind a little bit of what they were originally supposed to be, which is where the confusion comes in. A human, like, gives it a goal. A human connects the data sources to it. A human decides the actions it’s going to take often, and then it’s autonomously then doing the execution part. So the problem is and where the confusion is coming in with agent force and all of the talk about agents is they’re being presented as though they are autonomous. They are not. The analogy I always give is if anyone’s been in a full self driving Tesla, it is not full self driving. It is supervised, meaning there’s still a steering wheel. You have to have your hands on the wheel. You have to be able to take over if it makes a left turn when it’s not supposed to. That’s basically where we are with AI agents. The human still has to be able to know what it’s doing, know when it’s going wrong, stop it from doing it, review it before it even does it. So a really good example of this is if anyone’s tried Google deep research yet. This is one of the the most advanced AI agents I’ve actually seen. You put in, I wanna do research on this topic. Like, let’s say, how many businesses are there in the United States, for example, and and give me a bit breakdown by industry. You can give it that prompt. It will build a research plan. It’ll say, here’s the seven steps I’m going to do, and the human can edit and approve that seven step process. Once you do it, it goes and looks at the hundred sites. It reads everything. It summarizes it. And three minutes later, it creates you a seven page brief of the business ecosystem in the United States. That is an agent at work. The human, though, was heavily involved, and the human then decides what to do with that output. That’s where agents are. So that’s a really good example of an AI agent at work where it goes and does the thing for you, but the human tells it what to do and then assesses the output of what it did. So with the human in the loop, I don’t know how many jobs I’ve heard are dead this year because of AI. Right? BDRs are dead. How much do you buy into those sort of statements? I’m a realist when it comes to job disruption, and I have been touting this on our podcast for a couple years. I do believe that in two thousand twenty five, we will start seeing disruption to jobs. No agent in its current form can do the job of any human. It is incapable of doing a complete job of a human. What it’s capable of doing is when you break our jobs into tasks. And so if you say, like, as a BDR, you have thirty five things you do every month, then you can break it into these are the thirty five tasks. There is a decent chance that the AI can assist you with the vast majority of those tasks. Some of them, like, I have a call today at noon with Joe Smith, prepare me for that call. It might be able to go and pull data from the CRM and create a brief on that person and go pull some public data, and here you go. Here’s your brief. That’s great. That just saved me thirty minutes of, like, my prep time. So if you break into all these tasks, it can assist you. In some cases, it can do eighty percent of the task. In other cases, it might be ten percent. So right now, we’re at the stage where we all have to assess our jobs based on the number of tasks we do and where AI can help us. Now the challenge here becomes let’s say you have fifty BDRs, and you optimize the use of Copilot or ChatGPT or Google Gem or whatever it is, and you drive forty percent efficiency for the BDRs this year. You reduce the time they need to do their current workload by forty percent. The debate that every company will face is, does that mean we need fewer BDRs? The BDR job is not gone, but the supply and demand so it’s like if the demand for their work stays flat that we only need them to do with this much because we only have so many leads to work, then we need fewer BDRs to work the same thousand leads a month. If, however, you can create ten thousand leads because you’re using this AI to drive growth and revenue and innovation, then those same BDRs can work three times as many leads maybe as they previously did. So this is where we’re gonna be faced is we need fewer humans doing the same jobs. But if there’s the demand for what they do, the output they create that allows us to just let them be more productive and more innovative, then you can keep those people on staff. The beautiful thing to be doing today is starting a new company because you can reimagine it from the ground up with just fewer people. And you don’t have to get into this challenge of do we maintain the staff, but a lot of companies, a lot of c suites and boards are are having this exact conversation. How many humans do we need in each of these roles if these AI agents really work? I mean, again, a lot of it’s clickbait and all that about fear mongering and all. But you’re saying is that AI could actually potentially increase human jobs in a way because of that sort of multiplier effect of more more AI doing more, giving more quality for, like you said, a BDR or something working. Yeah. So my big thing and the reason I’ve been very vocal about the need to be thinking about this is I don’t think it’s a cliff we fall off of. Like, I don’t think we all wake up next month and all the jobs are just disrupted. It’s gonna vary by industry. So some industry are gonna be way slower to move this. But if you think about public companies, private equity backed companies, and VC backed companies, they think quarter to quarter. They look at numbers. They think efficiencies. And if there’s opportunities in the short term to reduce costs, they may not have the long term vision. And so my feeling is we have time now. In most industries, you may still have a year or two before we start seeing, like, the true impact of this stuff. Let’s plan for it now. Let’s think about how do we drive innovation. How do we drive and accelerate growth so that we don’t have to reduce staff? We may grow more efficient. Like, Jensen Wong at NVIDIA says this all the time. He’s like, he’s envisioning millions of AI agents within NVIDIA. They will keep growing, but instead of hiring a hundred thousand people, they may hire twenty thousand over the next five years. That to me is the ideal scenario where you have visionary leaders starting to think about what does our org chart look like when AI agents truly are there as ongoing intelligent assistants, and how do we maintain our current staff and then just grow more intelligently versus just always throwing FTEs at everything. And we’ve been, you know, a sixth sense is an AI company. Has been. It has been, you know, since its inception. Our BDR org, which does report into marketing, has changed. Absolutely. It hasn’t gotten smaller, if I’m being very honest, but with our agents working with our BDRs, we are creating more opportunities with fewer human touches, number one. We have our BDRs doing more of the human touches that they should be doing, so they’re on the phones. They’re making social connects. They’re doing all the things that, you know, brings them closer to selling. And in fact, we’re promoting more BDRs to AEs than we have in the company history because of the fact that they are not doing the things that the agent can do, like sending the emails and responding and making sure we hit those, you know, speed to lead SLAs. They’re selling and they’re growing. So it’s been a very interesting year for us and only more to come, I I think, in in the year ahead. And that’s the way to look at it. It’s like, what remains uniquely human and how do we enable the humans to do those things. Like, the reality is what you just described. Those aren’t like your favorite part of the day as a BDR. Like, it’s not the thing you love doing. Most people wanna have that human interaction and and be present and be, like, really listening because you know the AI is also listening and taking the notes and summarizing it. And so you’re locked in. And whether you’re BDR or a doctor, you’re actually just in the moment with the other human. And that to me is, like, a great opportunity. That’s a really interesting point too. Because, again, you start to think in your own head, like, okay. You’re in a meeting. You’re constantly taking notes. We don’t do really so much anymore because of the very point. That AI putting you in the moment kinda thing, that’s that’s a clever tagline for something. We’re gonna put that in the bio. We’ll use that. You’re gonna see somewhere. So now the thing that’s interesting to me also, like, you go on social media and you’re talking about and, like, there’s so much technical jargon around AI. Right? You see RAG. You see Transformer. All and, like, some companies are literally like marketing. You know, we’ve got RAG. We have the best LLMs, all that type of things. So that’s great if you’re a technical, a developer, you know, IT, somebody who’s building AI apps. If you’re in sitting in sales and marketing and we have the advantage of working with folks that are building AI products or using AI, but how do you go about and what do you need to know? Right? Again, because, like, if I’m I’m a sales leader, I am trying to figure out the future of my org. Like, how do you dig through all that? Like, retrieval augmented generation. Do I need to know that? Like, what what do you need to know? And, again, you’re not gonna be able to have look into the future. But to start to evaluate what’s actually a tool, a system that’s going to help you versus something that’s got a bunch of flashy three letter acronyms that aren’t actually doing anything. Yeah. The problem a lot of these AI companies are having still is they historically sell to IT audiences and developers. When I go in as a business user, I don’t need the choice of seven models. Like, I don’t know the difference between Flash and Mini. I just have a prompt or a business case, and I just want you to figure out which model does the work behind the scenes. It’s like when I go into Google and conduct a search, I don’t care what algorithm is being used and if they have seventeen different one. Like Oh, you don’t care if it that it’s a black box? No. Well, I may I may care a little bit about that. But I don’t wanna do that heavy thinking. I just want to have the business case as the marketer, the the salesperson. And so I think that what the average business person, business leader, or practitioner needs to know is what AI is capable of doing in relation to their job. So, again, it goes back to this whole thing. Like, I I built, a custom GPT called jobs GPT. It’s it’s free. Anybody can use it. It’s on smart r x dot a I. Just click on tools and it’s there. All you have to do is give it your title, and it’ll break your job into a series of tasks and actually tell you which of those tasks AI can help you with and how much estimated time you might save. And then you can say, oh, break that into subtasks. It’ll do that. You can do it for your team. You’re spending eighty percent of your time on these five tasks. Let’s just find ways to use generative AI to help you with those five, and we may save you, like, ten hours a week. Great. Like, start there. So I think what happens is too many business leaders think they have to solve, like, the big picture of AI when in reality, all you have to do is give your people three to five use cases where they see value in it, and that starts compounding across the organization. There’s an implication here that we haven’t talked about yet around the fact that so many orgs that are really trying to harness the power of AI and get to using it to simplify their tasks are still dealing with garbage data and CRM. Really difficult sort of, you know, things around just, are my contact data good? Is my opportunity data up to date? Like, how how is the data conversation, you know, a part of this all? So, again, it it goes back to your use cases for the technology. There are a bunch in the marketing, you know, in particular where just doesn’t none of that matters. Like, I always see data as a stumbling block for a lot of organizations because they do think until we get all the data in order, we can’t do anything with AI. So, no. No. No. No. No. You can be helping with your emails. You can be developing proposals. You can be doing creative. All that can happen. But to do what you’re talking about, we’re actually making a more intelligent sales process and go to market motion. We need that CRM data to be clean. Like, we’re thinking about this as an organization. Now we have ninety thousand subscribers to the marketing institute, and we are nowhere near where we should be from an intelligence standpoint with our own data. And so we’re looking at that right now of, like, okay. How do we get way better at that? So when we’re doing CRM related tasks, I’m preparing for my phone call, I’m automating a nurturing sequence, like, that the data is correct. And that is where again, if you’re a marketer, if you’re a CMO, don’t wait for permission to figure out GenAI until you get the data solved. Keep working on all the things that don’t require it. But when you need to really push deep into this and get all the way across your pipeline and all the way across the buying journey, you gotta be working with the people in your organization who understand the data structure, who can help you get it in order, and can help make sure that it maintains that as you try and scale the use of AI in the organization. Given that the generative bucket is what what everybody’s going fire now, crazy on and everything like that, what’s sneaking up? And maybe it’s continued forms of, you know, using the generative tools. But I have to imagine there’s other technologies out there. Maybe there’s offshoots of generative that are sneaking up. Anything that you’re seeing or hearing about, I mean, everything gets blown up every day anyway. Yeah. But that’s gonna really blow it up yet again, and people are gonna be going back to the drawing board. So last February, I created an AI timeline where I sort of projected out what I thought was gonna happen based on what I knew the AI research labs were working on. And so last year was like large language model advancements was gonna continue moving into reasoning, which is exactly what happened with the o one model from OpenAI. This year was multimodal, explosion and AI agents explosion, which is going to happen as well. And multimodal, what that means is CHAD GPT in its early forms, Gemini, all of them, they were trained on text in, and they were able to do text out. So you could go in and give it a prompt. It had absorbed all the content on the Internet, text content on the Internet, and it was trained on that, and it could do something for you. Now what we’re gonna have, and Google Gemini was the first to do this, is these models aren’t just trained on text. They’re trained on images and video, eventually audio. And so they’re trained on these multimodal data, and they’re able to output this. So we’re now seeing video generation like Veo from Gemini. Google Gemini is amazing. Sora from OpenAI. So video is gonna be a key thing. Vision. So if you haven’t done this yet, I think it’s only in paid accounts for chat g p t. If you go into the voice mode, you know, open up a normal chat, click on voice. In the bottom left is a camera. You can click that, and you can have live video of whatever you’re seeing on your device, and it understands that. And you can actually interact with it as it’s looking at what you’re looking at. And that’s gonna be in the Meta glasses. It’s gonna be in Google glasses. It’s gonna be in Apple’s, whatever they do after the Vision Pro. So, like, vision is gonna be a huge thing. But the thing that’ll probably make the most impact this year is gonna be the advancement in the reasoning capabilities. Now what that means is these models can now go through a chain of thought. They can actually go through a sequence of ten, fifteen, twenty steps. The way to understand this is CHAD GPT up until recently was what’s called system one thinking. You ask it for something, it does it. And the key is how fast can it do it? Milliseconds. It gets back to you. It doesn’t system two thinking is how humans work. You ask a question, and I stop and I process it. Now you don’t know what’s going on in my head, but in my mind, I may be going through ten, fifteen things, experiences, steps, and then I respond to you. That’s what the o one model from OpenAI was all about. It takes its time to respond to you. It could be three seconds. It could be three minutes. And so all of the AI labs are working on that. That has a massive impact on strategy in particular. So anytime you’re doing anything with marketing strategy, sales strategy, customer success strategy, go in and use the o one model and try it, and you will be blown away by how well it can help you think through complex problems. One quick one because, again, with the go to market audience on this podcast, we’ve talked about some of the bigger use cases that everyone is is doing. Right? We’ve talked about BDRs. We’ve talked about content. We’ve talked about design somewhat as well. What are some of your favorite use cases that you’ve heard or done, you know, personally that could maybe folks aren’t thinking about or aren’t top of mind? To me, the thing that’s just often overlooked is how good they are at strategy. So I ran a marketing agency for sixteen years. The hardest thing always was to identify and develop strategists. So people who can connect the dots between seemingly unrelated things, who bring the original ideas to the table that other people aren’t thinking of. So when you’re doing go to market, that is one of the most valuable functions within a team is someone who can actually think through everything. These models are at the highest level of any strategist I ever developed at my agency, probably already beyond most of their capabilities. Now you have to be a domain expert yourself to know if what it’s giving you is good. But I would say that I don’t know if I had to put a percentage on it. Fifty percent or so of my use of these tools is strategy, and it’s talking to it about strategy. So, like, our online education, as I’m reimagining what that looks like, I am talking to ChatGPT about that. Like, thinking through pricing models, I’m thinking through distribution channels, I’m thinking through partnership, all of that lives within threads with ChatGPT. And I actually built a co CEO that’s trained on our business model and revenue model. And I talked to co CEO about these things because it knows our business model, and it’ll actually bring that into the conversation. Well, based on our, you know, North Star, this actually would fit really well. I think strategy is the most underappreciated thing that these things already do really well that most people don’t think to use it for that. That’s good to know. And you talk about the o one model. And, again, we’re talking about endorsing products and so forth. But do you think that from a reasoning perspective, is that the model that’s I mean, it’s most accessible. Yeah. It’s they were first to market with that form of a model. Google Gemini will have that soon, very soon, I would expect. Anthropic Cloud will have it. But to get the best sense of what reasoning is gonna enable, o one would be the one to play with. They’ve already announced o three is coming. They had to skip o two because of, IP issues. But think of o one as kind of like so today, we have GPT four, the current model, or four five or whatever it is. This is like GPT one or GPT two of reasoning. It’s early, but it’s gonna improve really, really fast. And at some point this year, probably first half of the year, we will have the o three model. I think that they think it will be PhD level and beyond at strategy, and I don’t doubt that at all. All strategy. Any strategy you need to build in business, it will likely be PhD level. And data analysis too, they’ll do that. So we’re gonna switch gears for a second. We have a question that we ask all of our guests. K. It has nothing to do with AI. And it’s what is the most ridiculous thing you’ve been asked to do in your career? And it doesn’t have to be bad. It could be good in your business. I actually have one. I don’t know if I should tell it on this one. Tell it. Oh, that means it’s good. That means it’s good. So I won’t give you the example because you lock them in that you you’d lie. I’ve never seen anyone lock in that fast on that question. So you are locked in. Officially the most ridiculous. Yes. It’ll overshadow the whole podcast episode. I’m hesitant to even do this. No. Do it. Okay. Do it, Paul. So I ran a marketing agency that was also a PR firm. My background was in PR. I ran, junior golf tournaments, high profile junior golf events, and had some very high profile clients in health care and insurance and banking and the traditional stuff. And so one day, as early in the the life of the agency, I get a call from a adult magazine, let’s say, and club. I will not name the brand. And I thought it was a joke, but it was a New York number. And I was like, okay. Okay. So as I’m driving from a meeting with a junior golf client, I call my office, and I said, can you guys just go on this? This is the only time I’m ever gonna ask you to do this, but I need you to go to this website and see if this is even a real thing. Like, are they putting a location in our city? And and so I just hear all these giggles. I’m like, oh, yeah. You know, they they are. And I was like, so I I take this initial meeting as I called my wife. I was like, I have no intention to take this client, but I’m just see this conversation through. They wanted me to be the public face of this brand in this city as the outside agency. And I later realized it was because they were in, like, a a political battle with the city, and they needed, like, a clean rep to do this. That meeting ended really fast. And, so it could have been. Yeah. So they wanted me to do, like, a grand opening of a club, and I was like, you do not have the right person for this. So hands down the craziest thing I was ever asked to do in my career, and it did not go very far. I think you win. And, actually, that could open up, like, where AI works in that industry, but then we would be in some serious trouble. Yeah. You don’t even wanna go there. I know. I mean, this was, like, two thousand seven, probably. Like, it was a long time ago. Wow. Yep. Well, you win. I think you won. There’s more to that story, but I’m not gonna tell them. No. That’s no. I think that’s that was just the right amount of story. Yeah. On that note, Paul, you’ve been great. Thank you so much. I think folks listening will get a lot of way in it. What my takeaway from this is is leaning way more on the strategy side Yeah. Of what AI can do. I think the other thing is don’t overthink it. Like, pick the few use cases, get value, and then just stack those use cases one at a time and make them a part of your workflow, and it just becomes a natural part of what you do. Don’t don’t overthink it. Don’t take months trying to solve this. Just find use cases and go. Thanks so much, Paul. Thanks for having me. You’ve been listening to Revenue Makers. Do you have a revenue project you were asked to execute that had wild success? Share your story with us at six cents dot com slash revenue, and we might just ask you to come on the show. And if you don’t wanna miss the next episode, be sure to follow along on your favorite podcast app.