Pod Insights MarTech Podcast ™ // Marketing + Technology = Business Growth How AI orchestration will shape the next wave of software innovation About this episode The AI arms race is a head fake. Tom Chavez, Founding General Partner at super{set}, shares his expertise as a serial entrepreneur who has built companies acquired by Salesforce and Microsoft. He explains how marketers can leverage synthetic data to maximize efficiency with smaller, high-quality datasets rather than massive volumes of dirty information. Tom also reveals how AI orchestration can transform marketing workflows by automating repetitive tasks while augmenting human creativity rather than replacing it. See Privacy Policy at https://art19.com/privacy and California Privacy Notice at https://art19.com/privacy#do-not-sell-my-info.
Transcript 0:00 Speaker 00 The Martech Podcast is a proud member of the I Hear Everything Podcast Network. Looking to launch or scale your podcast? I Hear Everything delivers podcast production, growth and monetization solutions that transform your words into profit. Ready to give your brand a voice? Then visit iHereEverything.com.
0:25 Speaker 01 From advertising to software as a service to data, across all of our programs and clients, we've seen a 55% to 65% open rate.
0:36 Speaker 02 Getting brands authentically integrated into content performs better than TV advertising.
0:42 Speaker 03 Typical life span of an article is about 24 to 36 hours.
0:46 Speaker 01 If we're reaching out to the right person with the right message and a clear call to action, then it's just a matter of timing.
0:53 Speaker 00 Welcome to the Martech Podcast, a member of the I Hear Everything Podcast Network. In this podcast, you'll hear the stories of world-class marketers that use technology to drive business results and achieve career success. Here's a host of the Martech Podcast, Benjamin Shapiro.
1:15 Speaker 04 Recent research from McKinsey Digital identified 63 AI use cases in marketing that they estimate will generate $4.4 trillion in economic impact annually. It's no secret that AI's future impact is staggering, but today's reality fuels the illusion that every business must universally adapt or die. Here's the real real. Us marketers are drowning in AI hype today. Tools promise everything, but they don't deliver meaningful returns. So how can you separate out real innovation from empty promises? I'm Benjamin Shapiro and joining me today is Tom Chavez, a co-founding general partner at Superset. Tom has built and sold startups to some of the biggest tech companies in the world.
2:00 Speaker 04 And now he runs a venture studio that funds data-driven AI applications to deliver practical business results. And today, Tom and I are going to delve into what's Real and what's not when it comes to AI's impact on marketing. Tom, welcome to the Martech Podcast.
2:16 Speaker 05 Good to be here. Thanks for having me, Ben.
2:18 Speaker 04 Excited to have you on the show. It's an honor and a privilege. I said before, you've done some great things, started startups, sold them to big companies, and now you're passing it forward in your venture studio. You said something smart to me when we first met was that you described the AI arms race as a head fake. So how do you view the landscape of AI software innovation today?
2:41 Speaker 05 I find it incredibly exciting, right? So it was funny that I was at a party recently and there's a curmudgeon there who's talking about, oh, this AI stuff, it's such a bunch of nonsense. It's just endless hype as far as the eye can see. And I found myself putting the brakes on this and letting this person know, no, I've been at this for a while and I've seen fads and I've seen things that were never meant to be, right? In tech, we get very excited about new possibilities. That's why we love what we do. And some of those things never come to pass. AI is very real, very deeply rooted, so certainly not a fad. On the other hand, and I think the reason I think of it as a head fake is that in this particular case, the hype is just hopelessly out of control, right? So we can hold this idea of the promise of AI
3:26 Speaker 05 without subscribing to all of the breathless assertions about artificial generalized intelligence, you know, the AI fairies swooping down from the sky to take care of everything, to do all of our thinking for us, to do all of this gut work that we don't want to handle anymore. I think that's wrongheaded. So my concern as everybody rushes into AI is that we've just taken more balanced view. I think there's an equilibrium point here in the middle where it's possible to harness the potential of AI to radically transform what we're doing in the marketing realm. But with a view towards what I think of more as augmented intelligence. So AI, but let's have the A stand for augmented intelligence. And what I have in mind is this idea of putting the machines to work
4:10 Speaker 05 to make what modern marketers do cooler, better, faster, stronger. That's the premise.
4:16 Speaker 04 I hear you. I agree and I disagree. I agree with the principle of augmented reality being the reality. I'm thinking about the stages of prepping for this interview. I'm using artificial intelligence to mine data sources that I know are important to marketers to figure out what is the most relevant so I can frame my questions. I did research on you, right? I analyzed my own tone. We created our copy for our thumbnails in advance and our titles for YouTube. We do all these different things in advance to get ready. It's amazing. It makes my whole job easier and it augments the entire process. It is an external brain.
4:55 Speaker 05 It's awesome.
4:56 Speaker 04 Your argument for me to take the counterpoint is, well, but it's not going to be an AI fairy that drops down from above and hopefully not the Terminator to come and kill us all. But I do think that there are stages. There is an evolution here where we get to a point where actually that does become the case. And we are just so in the first inning of what's probably a 25-year rollout of the sort of manifestation of AI in the same way that mobile technology is still rolling out. And the first wave of iPhone apps were groundbreaking, amazing, but Uber didn't come around till 10 years after the iPhone. So I hear you in the sense of like, it's not going to be life-changing, but I also want to say not going to be life-changing and take over everything now. But what's
5:40 Speaker 04 stopping it from getting to that point over the life cycle of the technology, which seems to be moving very quickly?
5:47 Speaker 05 So if we're talking about 25 years, Ben, I'm all the way with you. It really comes down to the time horizon we're using to structure this conversation. I think certainly in 25 years, Agentic AI will be doing the lion's share of the work that we think of in the context of modern marketing. No question. Is it going to happen next year? Is it going to happen within five years? I'm out of the business of trying to predict exactly when these things happen. But I do like to just in this moment, let's recall you and I have been at this for a while. In the late 90s, I think it was 99, at the height of the dot com bubble, there was this concept that Rogers and Peppers were putting forth called one-to-one marketing. The reality is- If you've been in the boiler room of digital marketing, there's smidges of one-to-one marketing here and there,
6:31 Speaker 05 but it's really been one-to-many marketing, right? We do a lot of segmentation and cohort analysis and so on.
6:36 Speaker 04 Hang on, hang on.
6:37 Speaker 05 Yeah.
6:38 Speaker 04 But aren't we getting to one-to-one marketing now? Like, weren't they just wrong about the timeframe?
6:42 Speaker 05 We finally are. And instructively, it's 26 years later, right? So if we're taking the long horizon, absolutely, these things unfold. By the way, I also want to say with AI, what used to happen in nine months is unfolding in nine days now. So that's why I stay out of the business of trying to predict exactly when. What I am proposing is that in the next one to five years, there will be transformative possibilities in the use of AI to remove so much of the scut work that modern marketers deal with, right? As they try to synthesize data from different sources and cobble it together into a picture of what actually worked with the last marketing campaign. They launched new techniques for experimentation, personalization, targeting, synthetic data. I'm really excited about those possibilities,
7:27 Speaker 05 but it's not going to blow up the entire apparatus, I don't think, in the next one, two, or even three years.
7:34 Speaker 04 We've both been around for a little while, but not forever. But we've seen a couple of the cycles here. And I think it was, let's not go back to the invention of the telephone and the newspaper, but we saw sort of the peak of television advertising, the rise of the connection of the world with digital marketing, the mobile marketing revolution. And this seems to be the next evolution. Are there common stages that you've seen from some of the other changes in technology that have affected marketers?
8:05 Speaker 05 Well, I think you have your thumb on so many of the most important ones, right? There was the movement to mobile. I'm thinking of my friend Greg Stewart, who started the Mobile Marketing Association, and now it's the Modern Marketers Association, because as mobile just became part of marketing, right? Once upon a time, and you were there for it, there was a thing called online banking. Now it's just banking. So what part of marketing for the modern marketer is not mobile or multi-channel, multi-surface and so on? But that was one big revolution. I think the movement to big data, we used to use that word. We don't use it anymore. Big data in the cloud, that was equally transformative. Right in my first startup, it used to take us literally two days to compute something equivalent
8:45 Speaker 05 to the calculation of a segment, an audience segment in our second company that we used to do in about seven minutes. Right. So I've experienced those transformations myself. I think cloud and data was equally tectonic. AI is going to be at least as huge. And as you said a minute ago, we're in the first inning for sure. So when I say it's a head fake, I'm not a luddite about this stuff. I think it's really happening and it's happening fast. I just reject the idea that all of the work that clever, creative marketers do is going to be immediately supplanted by machines. I think the claim on the table here is that the near term possibility is really about augmenting what marketers do, again, to make it cooler, better, faster, stronger.
9:29 Speaker 04 The only argument I have for expediting the pace, like when we think about these sort of generational technology changes, 15 to 25 year rollout periods, and there's this incredible hype up front. Al Gore created the internet, whoever we're going to credit for it. And then all of a sudden, everybody's excited and there's websites and there's, oh, look, the first blogging platform and AOL. And then the shine goes off the apple and then it gets more into a slow, steady pace of innovation over time. And people don't think about it and then it just kind of gets normalized into the way that we do things. Mobile is the one that I think that we saw most recently. Nobody thinks of their mobile strategy. They think of their strategy and they have to have
10:13 Speaker 04 a mobile component to it because that's how the world works now. With artificial intelligence, there is a notion of the AI does the work for you and gets smarter and you can build faster. And so I guess the only question I have for you is when you compare these 15 to 25 year segments, do you think the artificial intelligence technology wave happens faster? Does it consolidate faster? Do we peak hype and drop and then go steady? Or is it just straight up and to the right?
10:39 Speaker 05 Look, in my second company, it wasn't just in the calculations and computations we were doing, things that used to take two days suddenly are happening in seven minutes. The way the markets moved, right, what used to take years would happen in quarters. As I moved from my first company in the early aughts to the second one in the 2010s. So we're dealing with that here today. And I don't know, is it 5X, is it 10X? I don't know exactly, but it is up and to the right, to your point, Ben. It's exponential, it's way faster. In our studio, I work with our teams and I always aim to be realistic, but a little bit unreasonable too. Where I say, listen, if that thing that you thought you could do in two weeks, if you can do it in two days, just know that your competitors are getting it done in two days.
11:22 Speaker 05 The speed at which we, using these new vibe coding platforms and the like, the way software is generated is fundamentally transformed. And I was, by the way, six to nine months ago, I was a little skeptical. And then I wrapped my head around it, got closer to these developments. Holy guacamole, it's not a state of mind. So is it 10X faster, is it 50X faster? I don't know, but it's a hell of a lot faster.
11:45 Speaker 04 I broke my leg and I spent the last two weeks on my back with ice packs on my leg, in bed with my laptop, and I became an email deliverability expert. I had no idea why, when I had every excuse in the world, that all I wanted to do was build infrastructure for my company. But somehow I got locked into talking to AI about how email deliverability works and how our infrastructure should change. And I basically went under the hood, ripped out the engine, rebuilt it, popped it back in. And now we've solved all of our deliverability issues. I'm not an engineer, I'm not an email deliverability expert. And so to me, what's interesting for marketers, and maybe this goes beyond just our craft,
12:25 Speaker 04 is your ability to execute in areas where you are not a specialist is now possible. I guess that's the definition of vibe coding, but maybe it's vibe marketing. You can do so much more with very little knowledge of what you're actually doing.
12:42 Speaker 05 I'm curious to know what systems or platforms you are using, but it sounds like you're doing some vibe coding and you're experimenting with the lovable and cursor and similar systems. I think vibe marketing is coming soon to a theater near you. It's unfolding now. My concern for the marketers who are letting somebody else do that, right? I love to ask people when I bump into them like, how many times per day are you using an L L M to do anything? To get an answer that satisfies curiosity or to get your job done, to rewrite some of your copy, to write some code, to get some data? How many times are you using it? And the folks that scare me the most are the ones who kind of fumble about and say, oh, still, you know, I'm not quite on it yet. Or I have some guys who do that for me like, oh, my God, dead man walking. You've got to be on these systems.
13:28 Speaker 04 I use it for everything. I craft emails. I identify my tone. I use it to research my guests, every single part of my world. I'm using it to diagnose whether the level of pain I'm having for my ankle is normal. So I don't have to harass my doctor. Like, I use it for everything. And the scary thing is, I don't know if it's right. I just changed all my email marketing infrastructure. Boy, I hope a consolidated D mark is the way to go and I don't blow up all of my deliverability for all of my podcasts, but I guess we'll find out. So talk to me a little bit about some of the practical applications. How can marketers integrate A I into their existing workflows to maximize their efficiency?
14:10 Speaker 05 I think that where we live today, the highest, best use of these kinds of A I techniques is in dull, repetitive, high stakes tasks. So in the context of marketing, and you mentioned it, listen, Nobel Prize winners and Pulitzer Prize winning authors are using A I to make their copy cooler, better. That's what the Nobel Prize winners are doing. By the way, I'm using it a lot myself in the same way. Use the A I to write smarter, better copy. I think that there's a lot of opportunities in the analytic realm now to do what you might call A I for B I, A I for business intelligence. Marketers are exhausted by all of these damn dashboards. Right, hunting and pecking in yet another dashboard.
14:55 Speaker 05 I'm not in Looker, Tableau, all of those systems were awesome for a long, long period of time. But what I see is a lot of exhaustion out there among marketers who have to hunt and peck through 25 screens to get the answer that they need. It's happening now, we're building it in my shop. A marketer can go now to what we call a chat with data interface and ask how many women between the ages of 25 and 35 who buy $250 of groceries or more per month use my coupon to buy my product in April. Boom, right away. No dashboards, just give me the answer, right? That's what we love about, I like to point out, LLMs are obviously a technology transformation, but I also think it's a revolution in the interaction mechanism now that we have with
15:40 Speaker 05 machines. I don't want to go and poke all around your website. By the way, this is dangerous ground for publishers because too many people are just going straight to the answer engine. They don't want to go to a new source anymore, they just want the answer. So there we are again, all of us using, I think marketers have the opportunity to use these kinds of techniques to change what they do in the realm of analytics, targeting and personalization. I don't know why we don't pay more attention to this, Ben, but whenever it was, three, four years ago, my chronology is always a little off, Meta's sole regulator, a company called Apple, decided that they were going to start asking all of us, do you want this app to track your movements around the internet?
16:19 Speaker 04 I heard about that.
16:20 Speaker 05 And so in one fell swoop, 70% of us said no. 70% of Meta's data used for targeting and personalization was vaporized and their market cap nearly collapsed. It was a really scary thing. And there were a number of people who were foretelling the end of Meta. Well, I wasn't in the room, but I'm pretty sure there's a guy over there named Jan Le Con, who's the head of AI, probably went to Zuck and said, listen, man, I think we're okay. Believe it or not, 30% of the data we used to have is plenty good. Let's put the machines to work and create synthetic data with the same statistical properties as the broader data set and use that data for personalization and targeting. It worked.
17:04 Speaker 04 Let's dive into synthetic data because I've had a couple of interviews where we've touched on it and haven't really gone deep. I get the general principle, which is you take an existing data set and you kind of extrapolate out what it might look like if the data set were bigger. But how is it accurate when you're taking 100 million Facebook users and you're saying, okay, well, what would this look like if there were a billion Facebook users? They're assuming what the other 900 million, whatever the number is that's the delta, that they know what's going to happen without the actual interactions happen. It seems like it is predicting the future. Walk me through how synthetic data actually works.
17:43 Speaker 05 These techniques, by the way, predated LLMs, and we should be clear about that. Some clever statisticians around Stanford and other places developed a set of techniques. Back in the day, there was something called boosting. Same idea. I'm going to take a small sample of data set. I'm going to recycle that data. I'm going to boost it in a way that maintains the statistical properties of the core data, the same means and modes and moments of a statistical distribution, right? So by emulating the behavior of that data as if it were a larger data set, but in a way that retains the core statistical properties of the seed, you could get more data to feed. And this is where it became so powerful in the context of AI, because we know AI is hungry for data. You can take
18:27 Speaker 05 that much larger data set and not predict the motion of individual users. We're not at Minority Report yet, but you're able to predict with very high efficacy.
18:37 Speaker 04 It scares me. You said yet.
18:38 Speaker 05 Yeah, it's probably coming. You could take me via my participation in a cohort of other users who look, walk, quack, talk like me, and give me more of what I want in that little moment of truth when there's something happening on a screen. And low and behold, the combination of these tools for processing of very large data pipelines, coupled with the statistical properties that were invented a bit ago, this kind of grand synthesis of all of these techniques now to create the synthetic data that saved Meta's ass, as far as I'm concerned, right? NH plus is the AI system that they use now for targeting and personalization. By the way, we use it in some of our companies inside my studio.
19:22 Speaker 05 It's magic. You can train those algos on here are users who converted. Here are users who bought my product. There's no way to decompile, and that's what's scary and frustrating for some marketers. I can't tell you why, because it's deep, deep in the third layer of a multi-layer neural network, and nobody knows why. Jeffrey Hinton, the inventor of all of these kinds of systems, will also tell you, we don't know why. I can't explain it, but it just effing works. And it's working at gigantic scale. Meta's numbers are evidence of that. So when it comes to modern marketers, my point to them is, listen, this is the game. You can run, but you can't hide. This is the game. Meta, Google, with those kinds of AI-based systems for targeting and personalization, it works.
20:06 Speaker 05 Do you really want to show up at a gunfight with a really sharp pencil? You're like, don't play along. Play the game. You've got first-party data, and it doesn't have to be as huge as you think it needed to be. In my last company, Ben, we had a very prominent digital marketing brand who came to us hanging their heads and said, well, we only have about 250,000 records in our loyalty database. And at that time, because we didn't have all of these new techniques and all of this new compute power, yeah, you had to go out to the open internet and buy a bunch of third-party data, which they didn't have a lot of confidence in. But you had to boost the signal from what felt like this meager data set to go and do personalization at scale. Well, now, 250,000 high-quality, like this is the triumph of small data now over big data.
20:51 Speaker 05 Small data that has a lot of fidelity and accuracy to it is high-octane fuel for the modern marketer if you're going to play this game.
21:01 Speaker 04 I guess that's really the takeaway is that small quality data can be synthesized to be more powerful than big, dirty data.
21:09 Speaker 05 And listen, I'll do a mea culpa here with you. And in my last project, I used to sit on stages and I would crow about the 3.5 billion monthly uniques that my company's infrastructure was seeing and managing monthly. That's a huge number. What I didn't want to talk about was that a good 20% of it was bots. I don't know, another 15, 20% of it was fraud. And then a lot of it was just, as you said, just dirty, unreliable data targeting behind the targeting of those uniques. Like, we were barely hitting the side of the bar with some of the uses of that data at that time. It's not like that anymore.
21:45 Speaker 04 Tom, I want to move on to your current role, understanding the funding phenomenon when it comes to artificial intelligence. Does AI integration create a competitive advantage in marketing? Like, how much of this is a hype cycle? And how much of it is companies that are actually saying they use AI or invest in AI or a built AI actually increasing their valuations?
22:08 Speaker 05 I think there are a number of companies who are talking the talk, and God bless them, they probably should. And they're seeing the results in their equity valuations as a result. Yank back the covers. And as you said, we're in the first inning still. So lots of pilots, nobody's moving effectively yet, to our knowledge, from pilot to production. And that's okay. That's exactly how these kinds of technology cycles evolve. The marketers who aren't taking shots are the ones who screw worry me the most. Take some shots, invest in some experiments, and be zen-like about the fact that not all of them are going to work. But, oh my God, you got to get off the couch and start experimenting.
22:46 Speaker 04 I wonder about the positioning. The marketer in me, sitting here saying, you have to position yourself to be an AI-led, whatever you are. We are an AI-led podcast production company. Now, I'm using ChatGPT and Claude and zapping data around to ask different queries. I'm not inventing anything proprietary. I'm just using the tools that are at my and everybody else's disposal. Does that mean that we're AI-led because it's integrated in what we do? Well, if it comes down to the valuation of our company, you bet your ass I'm saying we're AI-led because I think that actually means something right now. What does actual AI-led mean?
23:23 Speaker 05 Well, you can hate the game or you can hate the player. I urge everybody to hate the game, right? I mean, I think you're doing it exactly right. You are an AI-led studio and that's what you do. Now, how do you pay that off? So you can take a few liberties, I guess, Ben, is the point at the beginning. But you have to pay it off with real outcomes and real results through systematic investments in these kinds of techniques over time. There's going to be a big washout, right? We've all seen the trough. You know, we entered the back end of the hype cycle and we entered the trough of disillusionment where the bubble bursts and everybody says, okay, you were high and/or you were lying. That's going to happen, but there will be a number of serious contenders and serious players who emerge from all of that.
24:05 Speaker 05 And they're going to win because they weren't afraid to take shots, in my view. And they were just learning from those experiments and the ones that failed, at least as much from the ones that worked. And they're doing it faster than everybody else. So that's sort of the way I look at it. I'm excited by all of it, right? I think sometimes this technology stuff exhausts people like, oh my God, you guys all get all worked up about these things and then so much of it doesn't work. Yeah, yeah, yeah, but listen, the pieces that do work are really consequential. So not a good time to be a curmudgeon. I think it's full of possibility these times we live in.
24:40 Speaker 04 I think the metaphor is you miss all the shots you don't take. I might be butchering that, but it sounds like it should ring true in this case where if you're not leading with AI and flouting how you use it and how it is transforming your business, you are missing a great opportunity to, A, position yourself to be a leader, as somebody, an expert. And also, there's a valuation wave that is very much tied to the hype cycle around AI. And that wraps up this episode of the Martech Podcast. Thanks to Tom Chavez, the co-founder and general partner at Superset for joining us. If you'd like to contact Tom, you could find a link to his LinkedIn profile, which is in our show notes, or on martechpod.com. Or you could visit his company website, which is superset.com.
25:23 Speaker 04 You can also find Tom on his podcast, which is Closed Session. If you haven't subscribed yet and you want a daily stream of marketing and technology knowledge in your podcast feed, hit the subscribe button. In your podcast app or on YouTube and we'll be back in your feed next week. Okay, that's it for today, but until next time, my advice is to just focus on keeping your customers happy.
26:08 Speaker 00 Thanks for listening to the MarTech Podcast and I Hear Everything production. Looking to launch or scale a podcast like this one for your brand, then visit iHearEverything.com.
Previous AI-Led Social: Going from listening to interaction (copy) (copy) (copy) Next Big AI Platforms vs. Specialized Tools Page updated November 7, 2026 · transcript and summary are machine-generated
