Latest / Entrepreneur Odyssey / AI's New Category: Solving Marketing Silos & Privacy
Transcript
- Tejneet Selhi: Hello everybody and welcome to today's podcast. Welcome to the entrepreneur odyssey, Clay Sharman. How are you? Clay: I'm doing great, it's great to be here, thanks for having me. Tejneet Selhi: And Clay is the founder of Crateo, which is great company. So Clay, why don't you tell us a little bit about your background and why you started the business. Clay: Sure. So I spent most of my career in the kind of infrastructure software world solving government commercial problems where, you know, they had legacy systems and ⁓ the goal was to modernize for government ⁓ systems and build more efficiencies. But along the way I got into some machine learning and AI kind of foundational logic in the software side. I took about a two year detour into heavy, very deep Intel community based solutions where we looking at how to use machine learning to really capture inside threats before they took action and created problems. And so ⁓ that's where learned the framework for taking an alias down to an identity ⁓ based on lot of different sort of infographic, kind of graphic. ⁓ things like identity graph type things, ⁓ affiliations, direct messaging, social media content, looking at all that stuff, and then essentially distilling what the nature of that relationship of the alias to the identity to the action is. ⁓ so I use that ⁓ the framework for ⁓ at anonymous traffic from a commercial marketing standpoint, not nearly so nefarious. And actually, as it turns out, you know, very helpful. It's a huge problem for marketers and it always has been so. I felt like in kind of one sweeping move, I could solve a real problem in marketing that would not just help the marketers, but it would also help us as consumers get better messaging faster. so that was kind of the genesis of it was what could I help marketers do better when it came to building with their anonymous traffic. And that is what essentially Cradio's job is, is to ⁓ help do better and help online. experiences be safer and more effective. Tejneet Selhi: Amazing. And how does Kreative do that? What's the mechanism you use? Clay: Yeah, so in our case, what I really wanted to do was build a foundational logic at the base layer. So the AI in Cradio is in fact native AI and it is simply a logic based system that looks at how you and I or anybody goes to a website and whether they raise their hand and say, my name's Clay or not, my... Models logic is looking at how am I interacting with that content? How am I interacting with the brand? The messaging ⁓ much as we can sort of distill down. So if you think of it like ⁓ Analytics does a nice job of telling the marketers what is happening with their website But they'll never tell you who is doing it and so we wanted to go a step further and help define what that individual's behavior actually means to the marketing campaign budget and your strategy and all of that. And so we use AI to score every individual that we can identify. And then what we look to do is basically group them in potential audiences, like who's most likely to take a next best action. And also at the same time, we want to look at the people who are not likely to take an action. And so we can move them out of consideration. That way the marketer is not spending money needlessly. And, you know, so the whole intent is can we of build this funnel from the top? Yes. But also shrink the funnel at the bottom. And so at the end of the day, we're really looking for that sweet spot of who really is interested in what you're trying to tell them and how can you build a better relationship? So AI does that for us because it's able to react in real time, dynamically adjust scoring and give marketers real intelligence ⁓ what just happened in the last. 12 hours, 24 hours, ⁓ them with the pop-ups that say, before you leave, we've found that trying to pressure people into a decision when they're not ready is a failed campaign strategy. So what we really wanna do is give a way to identify who might be interested to hear back from you within a respectable amount of time, rather than just try to hammer them with buy now, buy now, buy now. So it's a nuanced. Tejneet Selhi: Yeah. Clay: model that does require technology like AI to help establish what that friend zone is for these potential buyers. Tejneet Selhi: Nice, I like the friendzone analogy. nobody wants to waste time, do they, on leads are very unlikely to convert. So think it's ⁓ a piece of technology. ⁓ And you came from a sales background. ⁓ So did feel, what made you down this path? Because obviously this is more of a marketing tool. Clay: Yeah. Well, I think at the end of the day, we're all consumers and in the digital world, you can't escape onslaught of advertising and marketing. know, ⁓ and I will have this conversation and I'm pretty sure that when I pick up my phone later, Facebook will have thrown something on about maybe the software we're talking through or something along the lines of a podcast I might enjoy. It's always listening. So ⁓ can't escape it. And so, you know, my motivation is a little bit. Selfish. simply hated getting bad email spams. I didn't understand how in a world where we had access to data, we couldn't figure out what to do with the data. so, you know, essentially I'm on a one-man mission to destroy spam ⁓ bad messaging. ⁓ so, ⁓ know, I had to move from the ⁓ side of selling that solution. had to move to the execution side of, well, this is a problem I would like solved for myself. probably others will like it too. And so that's how I kind of moved into the, don't see it as sales and marketing at the end of the day. I'm still kind of looking at it as I'm just trying to get rid of something that might make the online experience better and hopefully safer at the same time. Like try to solve two major problems at the same time. figure, you know, you could do worse than trying to solve irritation and privacy, you know, simultaneously. So. I'm satisfied that, know, I'm at least on a decent path. I'm not just trying out here to make a buck. I'm trying to improve ⁓ like essentially the Wild West and bring some order to it. Tejneet Selhi: No, absolutely, that's great. And since you started the business, how do you see that the industry has changed? Or not even necessarily since you started the business, but really, let's say in the last 10 to 15 years. Clay: So yeah, in this case, marketing has been what I would call, and this is not intended to be an insult to anybody. think we marketers have dealt with what they had to deal with as best they could. It's essentially been a slow moving iceberg. have not changed essentially since 1994 when the first online marketing message went out to try to drive traffic to websites. It was always, if we send enough messages to enough people, you know, a small number will respond. And so it's high volume, low reward model became the way of doing business, especially in an anonymous world, you really had no choice. so, you know, what I've seen is platforms are getting better at driving some intelligence and insight into that anonymous traffic world. But where they're failing is none of those platforms share their intent with one another. I'm interested in solving my problem, you're interested in your problem. And so there's been no way to look across horizontally. And so, you know, the industry really needed was ⁓ ⁓ over top of that that could be agnostic and look at the intent of each platform with a goal to ⁓ improve outcomes of each. And so really strengthen that marketing stack. That took a major turn, as you can imagine, in 2022 when generative AI was essentially introduced commercially. And then the idea of wrapping everything in an AI kind of paper became very vogue. But what we didn't anticipate, myself included, because even though I built a version of Cradio that was generative AI and did have native capabilities, I certainly was not prepared for 12 months later. Tejneet Selhi: Yes. Clay: this notion of agentic AI to suddenly storm onto the scene, absolutely lay waste to generative AI as a functional automation practice. I don't think any of us saw that coming. so, right. And so, you know, in the last 18 months alone, the industry has taken such a hard turn for the better, I think. But it has forced a lot of pivots or, you know, Tejneet Selhi: No. Clay: exposed a lot of gaps in capabilities because some of these legacy tools that are really good and enterprise level, they simply can't pivot and add a Gentic technology into foundation that was not built with AI at the base. And so ⁓ a lot of challenges to ⁓ at this kind of speed when you're an established enterprise tool. And ⁓ where I'm seeing a lot of the current challenges is the problems. remain the same because some tools simply can't adapt fast enough to get the job done. So silos still exist in the marketing stack. And so that's really where we're sort of trying to essentially create a new category almost. Tejneet Selhi: And that's really interesting because I think if we're working in silos and not sharing data, then we're not smart and as sort of specific about who we want to target ⁓ and how. So in of security then, how ⁓ are you of breaking down ⁓ those barriers you're trying to speak to a wide range of tools? ⁓ Clay: Yeah. So that's, mean, that obviously is not just the best question. It's also the hardest question that we had to tackle first. First and foremost, we had to make the commitment internally as a company that we were not going to deal with third party data in any way, shape or form. And so if we couldn't get consent based first party data, we don't deal with it. And so we don't, we don't look at, kind of the, ⁓ lead generating mills that will just throw you names of people that they buy up. Everything we do is fundamentally associated with our client. And so our customers have to allow us access to their website in order for us to begin the search because everybody who goes to a website, whether they're anonymous or not, has voluntarily gone there and thus makes them what we would call first party. privacy laws agree with that. If you voluntarily go someplace, that's first party. And then it becomes our job to distill that anonymity down, protect the person's privacy while also still allowing personalization to take root. so I think the fastest way we do it is regardless the platform that we work with that's siloed, we ⁓ can't so that the machine, our logic is built on, if I can't draw a direct line, from any information we get directly to our customer, don't use the data. And so that's one way. And the other thing that we filter out is if recognize somebody under the age of 18, they're automatically expelled from the database. We do collect personal information like addresses. We ⁓ immediately expel So there is no address. mailer companies can't use us because we're not going to allow that type of personalization. you know, to the extent that we can avoid no phone numbers, you know, to the extent that we can avoid that unless it's a business phone number that we can lock down. And so, you know, our whole goal is really fundamentally, is there a direct affiliation to our brand that we're supporting? And how can we even tighten that snare a little bit more to make sure that We're thinking of age-related consumers and being respectful of underage buyers. Yes, 16-year-olds are interested in stuff, but I'm simply not interested from a company standpoint in sharing 16-year-old data with marketing companies. So we don't target anybody who's not an adult. Tejneet Selhi: I'm sure that faced many challenges starting business and throughout your career. What one of the biggest ones you've had to in this and how have you been able to overcome ⁓ that? Clay: Well, yeah, I mean, that's the question that probably threatens and haunts us all as entrepreneurs, right? At all points of the day. First off, think you at any you cannot think to yourself, by Jove, I've done it. You can't be ⁓ that you've arrived some state that allows stability because the nature of entrepreneurship ⁓ is instability, uncertainty. And what is our greatest weapon in that battle is the ability to pivot within a relative time. We can act faster than our competitors. So have to be flexible in that sense ⁓ you're approaching these things. And I think the challenge that we try to foresee, and I'm sure everybody tries to foresee is, ⁓ know, what landmines are out there waiting for us that we could step around? And the answer is there are some, but there are some that... are deeper that you simply can't see and you've got to be ready to, you know, adjust. So, I the challenge in our case was how do we build ⁓ real that actually does something automated that other competitors can't How can we make the data, ⁓ as you pointed earlier, more accurate? you ⁓ know, there's nothing... The data sources that we pull are similar to the data sources anybody would pull from at that kind of volume. So our question was, what can we do with this data that has a more logic-based approach that's going to give a better outcome? So that was a challenge. And then making sure that the data that we're using and seeing is accurate because it's recent. So recency is a major player in understanding buyer intent. ⁓ Six-month-old data is useless in terms of what we're trying to accomplish. It's useful for... Tejneet Selhi: Yeah. Clay: you know, companies that might be doing email based newsletters, that it's probably okay. But in our case, we're trying to help marketers get to these people in a timely manner and build relationships with them. So you can't have this aged out stuff, even if it's accurate, if it's aged out, it's not recent, thus accuracy for our model fails. And then the challenge of, in our case yet again, choosing to use AI as our foundational technology, it forced us into being up to date with all of the shifts in AI technology at any point and trying to build some type of engine that would allow us to adopt new technologies relatively quickly without having to go back all the way to formula and create a completely new so in a technology landscape like today, ⁓ challenge that faces us all is, and I'm certain we're gonna get to this, is ⁓ waiting around the next corner from a technology standpoint and how do we prepare for that? know, I think amongst of that stuff, as you, as you know, you've got to think about the team that you're, you're going to battle with and are they the right people? And so there's a fundamental logic that I did not subscribe to at the outset ⁓ that now do wholeheartedly adopt. And I would say every entrepreneur should learn this. And it's a challenging lessons lesson to internalize is you, if you hire the wrong person, you have to get rid of them fast. You cannot carry them. So I mean, the wrong team can take you down the wrong path. And thus you have to be able to recognize that opportunity at the same way you're looking at your technology. So you should always be thinking what's best for the end outcome, you know, the company or the product, whatever. ⁓ who you have aligned with. Are they the right fit? You know, is it too soon to have a sales team? Is it too late? You know, at some point you just have to kind of be ready to make hard decisions about all of that. And so, you know, mean, there's, there's books and books on the challenges of entrepreneurship for a reason. They're right. Tejneet Selhi: There is, yeah. mentioned one thing there. So in terms of a sales team, when was it right for Credo to ⁓ on having a sales team? Clay: Yeah, so in our case, the goal was to prove market fit first with a generative AI version one. And so we bring in a small sales group in order to ⁓ onboard ⁓ kind early adopter mindset companies and work with them to get the feedback mechanism in place. You know, what's working, what's not working, what would you like to see ⁓ and what's for us to roll out. So that sales group worked with us early on and they were sort of I want to have a long-term relationship with them because I didn't believe ⁓ I the network at the time to hire the right person to take me from zero to a hundred thousand was going to be the person to go to a million. And so ⁓ I to be flexible and also the freedom to ⁓ thank them their service without any animosity, have a clear set of goals, and then we would move on. And so. once we proved the market fit, we hit a collision of the technology, Agentech coming into the world. So we had to kind of hit pause on onboarding new customers while we developed a version two ⁓ could take all the strength of what Agentech automation can do and enhance what we were able to offer. so we shifted focus and that intentional because I wasn't looking to build a company that could onboard 50,000 customers. What I was looking for was a company that would support, you know, mature or mostly mature marketing groups that understood the problems that they faced in a competitive market were looking to ⁓ and maximize their current spend infrastructure without having to spend a ton of money to rebuild or get rid of a bunch of contracts for marketing platforms. So I, ⁓ know, our sales strategy shifted from put as many pilot customers on and build this feedback loop too. We want to enterprise capabilities at a commercial off the shelf price. ⁓ so, you now our strategy is 250 customers that are, you know, curated that we can really support versus high volume, lower cost. Because, so I could be a little bit more clinical in looking at the sales group. Tejneet Selhi: So. Clay: And then we leveraged more about channel sales, partnerships, agency partnership relationships, people that could empower them ⁓ one to sell to many on our behalf. And so that became our of our strategy going forward. It also reduced our overhead. Yeah, I mean, from a sales perspective, we were able to manage overhead much better with ⁓ capital. Tejneet Selhi: Nice, very smart. Absolutely, absolutely. Now obviously you are developing, well you have an AI tool. other AI tools ⁓ you use as a business? Clay: Well, like everybody, I love the features capabilities of straight up ⁓ technology. So perplexity and chat GPT serve very well for research ⁓ you know, helping curate messaging, pressure test our ideas against, you know, what we think are AI ⁓ capabilities look across general industry standards. so, ⁓ you the, the trap that we try to avoid, and I recommend that everybody try to avoid this, we don't assume that because chat GPT or perplexity gave us an answer, it's the right answer. We have to then at that point, we have to verify and validate because at end of the day, we're putting our name next to that posted answer. And so it's not, you don't get the point back and say, well, that machine over there told me to do this, you know, at the end of the day, right? what we do, I certainly am a major proponent of right? Tejneet Selhi: No. Clay: I've worked for companies that touted efficiency but used none of it in-house and I always found that to be ironic and always swore that if I became a technology company, we would be technology first driven. And so we absolutely use it to help pressure test ideas. Definitely do research. Tejneet Selhi: And how do you stress test? you use different LLMs to do the stress testing or? ⁓ Clay: Yeah, I pit one against the other, usually, to see ⁓ the differences in the responses are. And I try to build what I would call word problems, where think I've got at least preconceived notion of what I think from my own experience the answer might be. then I serve what I is the third human independent variable in my own experiment. So I use the two different LLMs and then my own supposed assumptive response and then you know, Don't always win that battle I'm happy to report but it does allow me to always have a Human interface into the exchange with these two LLMs because then I can pressure test with questions. Why did you ⁓ go route? This model came up with that model. Why do you think those two? Why are you guys diverged here? When ⁓ as as I know you found you know, you looked across the same research And so I can force the questions a little bit internally on each one's, you know, and so I don't have to stress test it better than that, but to pit them against each other and then me be a referee, but I also get to be sort of the judge at the tail end as to how did they stack up against what I know from my own experience is probably an okay answer too. And so that's how it worked. And then, you know, like all things, I usually wind up getting a better pressure tested result. that I can feel more confident going to market with. Tejneet Selhi: Nice, good. And then what about AI tools ⁓ ⁓ business? Do use any kind of automation emails ⁓ or ⁓ things that? What's, yeah. Clay: Yeah. mean, so this is, this is again, not a knock on the industry. we've, we've used a lot of the AI SDR platforms ⁓ four in the last 18 months, four different ones, with ⁓ varying results of success, all of which I would consider a little disappointing. and, so what I found and what I think the market would agree with is, you know, AI sounds the same when it starts talking about going to market with SDR-based scripts. almost every case, the subject lines are a little more generic than a human because there's no intuitive nuance that they understand how to add in. They just look at what they think is a compelling keyword. so, know, and adjustments don't exist yet in today's nuanced response from an LLM. And so what I found was across all four of these SDR companies, the script always sounded exactly the same. And then it was up to me to try to find out what could I put into it to make it sound more like a Cradio specific thing. But personalization is challenging. So the AISDR thing, I think it's going to become really cool. I just don't think it's there yet. In fact, I think it's further behind than the copywriting and the research capabilities for sure. because it's trying to mimic ⁓ relationship building at some scale and speed, and that's hard. ⁓ ⁓ didn't want to get into that world for that reason. Behavioral understanding is one thing. Behavioral assumptive scale reasoning, we're nowhere near that capability to get to a bottom line. So I do use it for scheduling, and we have automated email. of outreach through SendGrid, which claims to use some AI in its back, kind of back end. So at that point, it's a trust exercise for me. If you tell me you're using AI something, my assumption is it's a little bit more maybe quality, consistent quality, not necessarily more accuracy, of course, because wrong is wrong, but it might have a more consistent grammar, you know, or because. Tejneet Selhi: you Clay: That's because I think it does do copy well. It's just not exciting copy. And that's what's missing in an SDR script from an AI perspective is it doesn't sound exciting yet. Or if it does, it's an attempt at sounding exciting, which comes across as disingenuous. so, you know, so we, but yes, I'm technology first. So we try it at the end of the day. don't think the personalization capability from an internal standpoint is, is, is there yet, but back to where I think we're going to wind up getting to this conversation. I certainly think it's going to get better. Tejneet Selhi: Yeah, I mean with all of this technology it's only ever going to get better isn't it? So yeah that does lead us then to you ⁓ to take the business in the next six to twelve months ⁓ this ever-changing landscape. ⁓ Clay: Yeah. Yes, so I think I hinted at the category building, which is what we're really trying to do is a an absolute missing layer ⁓ technology understanding in the in the marketing space. And frankly, I could probably put that layer across other verticals as well. But I'm picking marketing because it's massive. It's got a problem I can solve. So that layer ⁓ you know, kind of I'm evolving to what I would call Tejneet Selhi: Mm-hmm. Clay: ⁓ artificial general intelligence. not, didn't coin that term, AGI exists, but it doesn't exist in the marketing space. So this layer that I'm looking to create is essentially AGI for marketing. And you know, what that simply means is agnostic platform that can understand ⁓ intent of any marketing platform's goals and then bring all that information into a single platform with a single intelligence that looks across and says, I know what you're trying to do here, but this is counter to what this platform is trying to do. And over here, they're trying to do this. So we're going to take all of this and I'm going to give you the best recommendation for your next best action, taking all of this into consideration. And so, you know, I'm trying to build the ultimate platform recommendation engine that can understand your entire marketing stack, your corporate growth goals, unify those two and give you actual recommendations on how to move forward. by campaign channel or campaign message or audience, audience being the most important of those three, by the way, without on budget or better yet, ⁓ your existing budget and giving you much better return. And so ⁓ the end goal for me. And think that that's, so, ⁓ know, heavy market sort of acknowledgement, penetration and and then ultimately, market winning, right? I want to own that market. So we have to have speed. So six to 12 is a good model. In 12 months, I need to be pretty far down that road of market ownership, even if I've not crossed the finish line, you know? Tejneet Selhi: nice. Well, I think they're ambitious, but hopefully very achievable goals. And I'm excited to see the journey. Clay: Yeah, I mean, I hope that I give you something excited to see, you know, in these coming months. I think we will. I think what we've built is a great stepping stone towards flight, you know, whereas marketing's always sort of tried to run. I think we've developed the engine that's going to allow us to glide and fly. And so I think it's going to change things. Tejneet Selhi: Very cool, very cool. Well, maybe you can come back on towards the end of the year and let us know how you're getting on. Clay: Fantastic, I accept that. Tejneet Selhi: Excellent, well I'll hold you to it then, Clay. been such a pleasure to have you on today's podcast. ⁓ there any final message you would like to leave ⁓ the audience today? Clay: ⁓ No, think this is an exciting time for entrepreneurs, especially I think younger entrepreneurs are a great spot because you're seeing the horizon of what this technology could be and in terms of seeing what's been before you, ⁓ have an opportunity to build a map and a roadmap that really can accelerate and excite kind of... what's possible. And so I would say it's a great time to be an entrepreneur and technology, if we use it right as the tool that I know it can be, going to really be an journey for the next generation of entrepreneurs. Tejneet Selhi: Absolutely, absolutely. It's a very, very exciting time. I have to concur with you there. Well, thank you so much for coming on Entrepreneur Odyssey. It's been a real pleasure. And in for our next episode soon. Take care, everyone. ⁓ Clay: Likewise.