Zeta Global’s David Steinberg on AI, Data & the Future of Marketing Technology

Well, Matt, I'm a huge fan of yours, too. Thanks for having me. It's great to see you again.

Well, first of all, you know, you know how it goes, Matt. You're an entrepreneur, too, right? How do you become an entrepreneur? You're unemployable, right? I couldn't get a job out of school. But, you know, it's sort of like I do this thing I call triangulation, where I just, I read a lot. Call it five to six hours a day. I'm reading different publications, white papers, scientific journals, so on and so forth. And I try to triangulate sometimes incredibly disparate data points that lead me into believing that certain things will happen. So, with Zeta, it was 2017. We had bought a bunch of, you know, customer acquisition and ESP assets. And we were sort of running these eight disparate platforms. But we had started the beginning of building the sort of data ecosystem. Now, it's pivoted a lot to the data cloud, but we had a lot of data. And what I was looking at at the time, Matt, was there was no way to process the amount of data we were ingesting in time to make an even somewhat intelligent decision. Meaning, it could take hours to make a decision that would be best served made in a millisecond. And that's when I started to explore and learn about artificial intelligence and sort of it came from reading about different things about AI. It really started with natural language processing. How do you ingest information, synthesize it, and so on and so forth. But, you know, you just have to be willing to be open to incredibly new ideas and you are, too, right? I've seen it with you in your career. You have to be incredibly open to new ideas and, you know, a lot of people ask me how did I get from telecom to, you know, AI data-driven marketing. I had founded four telecom companies, sold three, taken one public. When the fourth one got purchased, I found out from the private equity firm that had bought it, I had signed this incredibly comprehensive non-compete in the wireless space that, quite frankly, was probably one of a hundred documents I signed without reading the day we went public. I didn't make that mistake of not reading everything this time. But, you know, I sort of was like there and I remembered how incredibly complex it was for my last company to run its marketing. It needed 15 to 17 different vendors. And the vision was not, I mean, listen, in 2008, the vision was not data and AI, you know, we would have been way too ahead. It was just how do we put everything a marketer needs in one place and eliminate all the different point solutions. And on our journey to that sort of North Star vision, we figured out that without a data cloud and without AI as foundational to the platform, native to the application layer, we couldn't deliver on that North Star vision. So, we sort of happened into data and AI in 2017 as a part of continuing with that very early founding vision.

Yeah, so we help very large enterprises to meaningfully better manage their customer acquisition, customer retention, and customer monetization. So, you have acquisition full, I mean, literally from the moment you meet to keeping them from churning. Like literally. And what we have found, Matt, is that flywheel starts with acquisition. Or it starts with retention. But the data that informs acquisition then informs retention, which informs monetization, which better informs acquisition. And if you look at it, recently there was a Forrester study that just published that for every dollar spent through the Zeta marketing platform, we return between 600 and 700% on ad spend. So, yeah. I mean, it's really what I think one of the reasons that even going into 2026, where we are now, right? We've projected our fourth consecutive 30-plus percent growth rate year.

Yeah, I mean, I would start by saying that the biggest difference between us and our competitors, we are the disruptor, not the disrupted. Like, we're the guy, like last year, if you look at the marketing ecosystem, it grew about 10%. We grew 30%. Right? So, we're, by definition, you're taking meaningful market share from your competitors. And if you look at it, when we started programming with AI in 2017, we launched the Zeta marketing platform in 2021 with an entirely new architecture that put AI and data as native to the application layer, which, by the way, removes latency and by definition drives massively higher return on ad spend. Right? So, you know, I think, you know, I think that, I mean, let me also say that I think there's going to be major winners, major losers in the large language model ecosystem, right? I'm old enough. I was around when the internet was going to kill Walmart, was going to kill JP Morgan Chase, was going to kill FedEx. And, you know, what we found were companies that adopted the internet massively accelerated. Ones that didn't are dead. Same thing with cloud computing. Right? When cloud computing first came out, remember that? All enterprise software is going away. You're going to have little apps that sit on top of you like Snowflake, Databricks, or AWS. The exact opposite happened, right? I think when you look at what's going to happen, companies that embrace artificial intelligence like us, and I'm sure you saw at CES this year. I think I saw you there. We announced one of OpenAI's earliest enterprise-grade partnerships, where we've built Athena, which, as you know, we just launched the other day. We've built Athena in partnership with OpenAI. We can get to her in just a second. But I think the biggest differentiations between the winners and losers are, A, who has first-party proprietary data that doesn't feed into the large language model. So, just to re-architect our model, we set up a consumer data platform for all of our clients. We had 603 enterprise clients, including 51% of the Fortune 100, at the end of last quarter. Okay? If you look at that, we ingest all of their first-party data into the CDP, we merge it with the 552 million opted-in individuals in our data cloud where we have an average of 5 to 7,000 data elements per person, and our first-party tracking pixel is sitting on trillions of pages of content. And all of that is feeding into a proprietary algorithm inside of the consumer data platform.

No, I mean, that's the thing I actually, they're most concerned with.

In fact, it's what we hear from every one of them. In addition to that, our data cloud has never been fed into a large language model and never will be. Second, are you creating return on investment for your clients or are you a major cost center? At Zeta, we're a revenue center. We're returning $6 to $7 for every dollar our client spends. And then three, Matt, you know how hard it is to sell into a Fortune 500 company, right? You've got procurement, you've got data security, data privacy, the CIO, CTO, legal, accounting, then you have to do SOC 2 audits, and everything has to be delivered. I just don't think the large language model companies are going to build expertise in that. I think what they're going to do is partner with companies like Zeta that embrace them. And, by the way, they'll become a meaningful component of our cost structure. If you look at our first adoption, Matt, of AWS, which I think we did four or five years ago. And we had 26 hosting centers at one point. And we work with everybody. We work with Azure, we work with Oracle, we work with Google, we work with AWS. We do the most with AWS. But everybody said, you're crazy, you can't do that, the compute will be too expensive. We tried it in Europe and we loved it. We've now adopted it globally. In the last four years, actually the last three years, not only has AWS continued to drop as a percentage of revenue, but we've added 500 basis points to operating margin, and we've grown the business on a compound growth rate of greater than 30% top line, greater than 50% EBITDA, and greater than 75% free cash flow, while simultaneously making AWS one of our larger global expenses. I think you're going to see a similar thing with the large language models. We'll adopt them, we will pay them a fair rate for their software, which is very, you know, very valuable, but I think we'll continue to grow at a faster rate than that cost structure will grow, and I think we'll continue to expand operating margin.

Let me, let me start with the second part. I'm going to use invasion, I like that. I like, I like invasion. Athena's goal is to solve our clients' biggest problems. It's really that simple, right? And if you think about it, like almost all software companies, anybody who's watching this has a piece of software that they've paid for that they use 10% of. You're paying for 100% of it, but you use 10% of it, right? So, what I say is we have built a 747 jet for a platform and a data cloud, and our clients don't know how to fly a Cessna. Athena is our voice-enabled, fully conversational co-pilot for the platform. So, instead of using the keyboard, which I think, if you think about, and I don't want to get too existential here, but if you think about humanity, right? We've been around for, you know, hundreds of thousands of years, right? We're the last second at midnight from the time of the Earth's existence. And for our entire existence, we communicated via voice. We didn't even create a keyboard until the 1950s. We didn't have a mouse until Xerox created it. You know, like you have all these different things that came out, but humans like to communicate with the voice. And the idea really germinated from, like, as you want to remove the friction between the individual and the AI.

Exactly. They know how to explain their problem and ask how to fix it. I'm sure you saw the other day when we launched Athena, we did it as a birth announcement. We sent hundreds of boxes out with a, like, beautifully done thank you, you know, like birth announcement. And then we did the birth of Athena instead of the birth of Venus, you know, the old Botticelli painting. They had her clothes, by the way. And then we did baby pictures of Athena and really sort of rolled it out because she's a person. Like, Athena is meant to be your friend. She's meant to help you navigate the platform. So I thought a birth announcement versus we're making her just generally available was a cool idea. It seemed to have been very, very well received. But when you think about it, the journey started years ago to answer your first question second. We created a program called Zoe, and Zoe was our first voice-enabled platform. She was very clunky. She was, you remember the old sort of walkie-talkies, you'd say, 'Hi David, out.' And you'd, like, have to, like, really stop and, like, whereas you can interrupt Athena mid-thought. Like, she's fully conversational. But we integrated and launched Zoe a few years ago. We actually found something super interesting, Matt. Companies that adopted Zoe spent between 250% and 275% more on the Zeta marketing platform than clients that didn't. So when I came to the team with the idea for this voice-enabled co-pilot that could run your entire life, but starting with marketing, everybody was like, 'This is a good thing to invest in because we had already seen the ROI from clients that were using Zoe.' And we went into full build mode. I would tell you from the germination of the idea for Zoe until her going generally available was approximately 10 months.

Oh, and by the way, you know, everybody talks about the SaaS apocalypse. We're using Claude, Anthropic, to build her.