If your marketing grew like a dividend-reinvestment plan, would you still let a quarterly target dictate every decision?
Agility requires stacking returning gains faster than the market changes—think compound interest, but for marketing campaigns.
Today we’re going to talk about the Compound Marketing Engine, agentic AI, and why “data-driven” still needs greater adoption among leaders. To help me discuss this topic, I’d like to welcome Chris O’Neill, CEO of GrowthLoop.
About Chris O’Neill
Chris O’Neill is CEO of GrowthLoop and a board director at Gap Inc. (NYSE: GPS). His 25+ year career includes leadership roles at Google Canada, Evernote, and Xero, and board experience at Tim Hortons. As an advisor and investor, his portfolio includes Koho, Plus AI, and Neeva (acquired by Snowflake). Chris lives in Northern California with his wife, two children, and their dog Teddy.
Chris O’Neill on LinkedIn: https://www.linkedin.com/in/croneill/
Resources
GrowthLoop: https://www.growthloop.com https://www.growthloop.com
The Agile Brand podcast is brought to you by TEKsystems. Learn more here: https://www.teksystems.com/versionnextnow
Catch the future of e-commerce at eTail Boston, August 11-14, 2025. Register now: https://bit.ly/etailboston and use code PARTNER20 for 20% off for retailers and brands
Don’t Miss MAICON 2025, October 14-16 in Cleveland – the event bringing together the brights minds and leading voices in AI. Use Code AGILE150 for $150 off registration. Go here to register: https://bit.ly/agile150“
Connect with Greg on LinkedIn: https://www.linkedin.com/in/gregkihlstrom
Don’t miss a thing: get the latest episodes, sign up for our newsletter and more: https://www.theagilebrand.show
Check out The Agile Brand Guide website with articles, insights, and Martechipedia, the wiki for marketing technology: https://www.agilebrandguide.com
The Agile Brand is produced by Missing Link—a Latina-owned strategy-driven, creatively fueled production co-op. From ideation to creation, they craft human connections through intelligent, engaging and informative content. https://www.missinglink.company
Transcript
Greg Kihlstrom (00:00)
If your marketing grew like a dividend reinvestment plan, would you still let a quarterly target dictate every decision? Agility requires stacking returning gains faster than the market changes. Think compound interest, but for marketing campaigns. Today, we’re going to talk about the compound marketing engine, agentic AI, and why being data-driven still needs greater adoption among leaders. To help me discuss this topic, I’d like to welcome Chris O’Neill, CEO of Growth Loop. Chris, welcome to the show.
Chris O’Neill (00:27)
Great to be here, Greg. Thrilled for our conversation.
Greg Kihlstrom (00:29)
Yeah, looking forward to talking about all this with you. Definitely a lot to cover, but I think we’ll get to it all. So before we dive in, though, why don’t we start with you giving a quick background on yourself and your role at Growth Loop.
Chris O’Neill (00:41)
Sure. My career spanned a bunch of leadership roles. I’ve been very fortunate to work at some amazing brands, Google, Evernote, Glean, Xero, and a few others. I’m the CEO of Growth Loop. A company I’ve followed for a while. I had the good fortune of working with her, with Growth Loop co-founders at Google, and have known them for over a decade at this point. really thrilled to be here and really lucky to lead such a great team.
pursuing a meaningful mission here.
Greg Kihlstrom (01:11)
Great. Great. Yeah. So let’s, let’s dive in here. so growth loop recently introduced what you call the industry’s first compound marketing engine. So let’s start there. You know, what exactly does that mean? I know I briefly teased it in the intro, but you know, what exactly does it mean and why should marketers be paying attention to this?
Chris O’Neill (01:30)
Yeah. So compound marketing derived from my fascination with the concept of compound interest. So Albert Einstein famously coined it the eighth wonder of the world. And from a very early age, I became obsessed with investing. This notion of compound interest was really just at the heart of it. So I got to thinking what’s going on in marketing or business more generally that’s preventing the type of gains in growth that we all want to aspire. It often isn’t that we try to get a little bit better. We always try to get better. But what’s missing is the speed, the iteration speed. So the difference between compounding at a weekly basis versus a quarterly basis or a monthly basis is not a little bit. It’s a lot similar to how compound interest works in finance. So really, when we thought about it,
Marketing cycles are too darn slow. There’s manual steps at every step of the cycle is manual. They have to be stitched together manually. And that really holds companies back. So we thought there’s a better way. And that’s really what a compound marketing engine is really all about. It’s applying agentic AI to your data in your data cloud to reduce the distance between I have an idea and insight to impact. That’s what we’re doing.
Greg Kihlstrom (02:43)
Nice, nice. And so we’ll I want to talk a little bit about how that works with with agentic and stuff like that in a second. But I mean, first, I mean, this is really possible because things move so quickly. Right. I mean, this is, you know, we have access to data. You know, big data was like the thing what like 15 years ago now or something like that. So everybody’s been stockpiling all this stuff and data lakes and lake houses and all that kind of stuff. And now we actually have the ability to move quickly. But yeah.
Is that kind of the genesis of this is just that that need for the for the speed of marketing?
Chris O’Neill (03:16)
Yeah, well, I just happen to believe the more agile you are as a marketer or business, you’re going to win. You’re going to take more shots on goal. You’re going to take better shots on goal. You’re going to be able to learn from previous efforts. But it really does start with the data. Like this is all possible because of the rise of data clouds. Part of the challenge is the fragmentation of data all over the place. So you got to of stitch things together. It’s a very good thing. And it’s very obvious to us that and either through serendipity or luck and maybe a little bit of intelligence that things are going to be in the data cloud. They’re going to start and end in the data cloud. All right. So that’s very much a part of this is getting your hands on data, having a very clear data strategy, having a semantic layer on data so that you can do important things. In this case, lifecycle marketing and really personalizing the journey, which has been the Holy grail for many years, of course, and we’ve really fallen short.
for over, for decades, really.
Greg Kihlstrom (04:13)
Yeah.
So could you walk us through, you know, either a real case or, you know, a hypothetical example where, you know, we’ve got several AI agents handing off tasks and this idea of compounding over time. And, you know, how do the marketers also factor into this this scenario?
Chris O’Neill (04:32)
Yeah, so it’s very important that marketers are in the loop indeed. So I think of agents as teammates and the very beginning is really understanding the data. So agents are good at that. So understanding the schema, understanding what’s in the data itself, what worked in the past in order to suggest experiments, starting with who to talk to, who to target, right? So the very first agent, well, there’s a coordination agent that basically wakes up and says, what is the person asking? So we very much believe in outcome back work, meaning what are you trying to accomplish? You’re trying to reduce churn, you’re trying to increase lifetime value, you’re trying to, I don’t know, grow a specific category, whatever it is. You then turn that into specific ideas by first looking at the data to say, hey, what have we learned in the past in order to say, okay, what does that look like in terms of an audience? Previously, just a little bit of background, the before to now.
It was, ⁓ know, metaphorically, like the marketing teams would line up outside ⁓ the data teams door, like a bread line in the depression, asking for them to fulfill their needs to say, have an idea who should I target? Let’s run some SQL and bounce back and forth.
Greg Kihlstrom (05:42)
That line, yeah.
Chris O’Neill (05:45)
There’s
gotta be a better way. And that’s where we started to be clear, really being precise with democratizing that data so the marketing teams can do that themselves. No lines needed. Okay, so we can now do that by actually having agents do that work too. Not only do you have to translate insight from a marketing person to an audience, they do that instantly. But now we’re surfacing proactively, surfacing some suggestions which are literally served up to you, which then are able to be activated through a journey or orchestrated through different channels and then actually executed across channels, whether that’s SMS, push, email, a paid ad, you name it, hundreds of different surfaces, campaigns run, results run, read back into the data cloud, and then you lather, rinse and repeat.
This is the notion of growth loop, the name of our company. But at each of those steps, we have agents. Now, some of the agents are better and more fully developed than others. And we’ll be clear about that. But that’s very much our vision that that end to end happens with agents at every step, but a human also in the loop to sign off and inject creativity and spontaneity into the mix to make sure it’s on brand and make sure it’s really resonating with the audiences.
Greg Kihlstrom (07:02)
Yeah, so it sounds like, I mean, it would make sense that it works best when it’s, you know, full funnel, you know, start to finish or I guess it’s it’s hopefully it never finishes, but you know, full full life cycle and omni channel and all that. Are there places where, you know, if you start seeing momentum first, so, know, in in in that funnel, you know, are there are there certain parts where you start to see the results more quickly?
Chris O’Neill (07:27)
Yeah, it is in the audience area. And it happens to be where we have the deepest level of domain expertise. So perhaps that’s not coincidental. One of the agents that is happening and developing far more quickly than I would have anticipated is on the image creation. So somewhere in that loop, you basically have to say, what are we going to say? What words? What content are we going to put? And even now, what video or image?
The models are getting so good, so fast, so that that part of the loop is really starting to elevate quickly. We had that on our roadmap sometime next year. In startup land, that means pretty much never, right? There’s so much to do right now, but we’re really pleased with what we’re seeing around the ability of these models, multiple of these models, to translate what we’re trying to accomplish, outcome orientation, into a creative brief, creative brief into copy into actual images and now with VO3 and other things like it, like full motion and video, it’s pretty astounding. So it’s exciting to be in this business.
Greg Kihlstrom (08:31)
So, know, agentic AI, mean, it’s, you know, it’s it’s the thing right now, the kind of the shiny object right now, so to speak. And yet, you know, as you mentioned, lots of actual practical applications for it. What do you think over the next, you know, let’s say, I know 12 months is like forever in the future at this point. But, you know, over the next six to 12 months, what kinds of like agentic workflows are going to be?
Generally, you know accepted kind of the norms and what are some things that you know, maybe CMOs skip for now?
Chris O’Neill (09:01)
Yeah, I think the full, the full end to end will take that full time. I think we’ll start to see people pulling it all the way through in that time horizon. We’re doing that more manual, automated way, but not necessarily with the, the, ⁓ assistance of AI at every step right now, to be clear. So, but I think the full loop will be in realm. A couple other things that I I’m paying attention to is really the intersection of machine learning and more specifically reinforcement learning and AI.
Machine learning is propensity modeling and really predictive based upon previous patterns, right? And then reinforcement learning to say, hey, what are you trying to solve for? There’s a lot of innovation in that realm that’s just starting to take hold. It’s really more promise at this stage than quantifiable reality, although that’s happening quickly. One possible iteration of that is we’re talking about agents to manage workflows for the marketers.
What we see is the opposite to be true too. In other words, an agent for the actual consumers that will actually help personalize in a way that’s highly relevant. And then these agents will talk to one another to start to deliver fundamentally better experiences. That’s something that’s important. I alluded to the image stuff that’s happening at a breakneck pace. I wouldn’t have expected that. That’s happening. Another one we’re paying attention to and quite excited is basically simulated data where you’re basically proxying real life profiles of humans so that you can actually test with synthetic audiences, synthetic panels to say, hey, we think this collection of activities, the combination of who and what and when, will have a desired impact against an objective we care about, but we can run it against synthetic data. And you know, it’s getting pretty good. So we’re really excited about that as well. those are the things we’re paying attention to.
Boy, boy, in my life, I’ve not seen the pace of innovation, the pace of change, and that’s what makes it exciting. It makes it daunting for people to to stay in touch with it all, but we’re doing our best, and that’s part of the relationship we have with marketers and the data teams that we are so fortunate to work with.
Greg Kihlstrom (11:08)
Yeah, yeah, definitely. Yeah, well, we’ll have to have you back on the show and talk about the synthetic stuff, because that’s that’s definitely a top of mind for me right now for for a few things. But yeah, you know, definitely everything that you mentioned, ⁓ you know, lots lots of exciting stuff there. You did mention, you know, the importance of keeping humans in the loop as well. And, you know, I think that’s that’s an important thing to underscore, too, because we’re all talking about, you know,
personalizing, personalized journeys. And, you I’m excited about the, the predictive and the machine learning plus gen AI that lets us actually do like the segment of one stuff that we’ve been talking about for quite a bit too. But how do you look at, you know, putting the guardrails in place with the agents to make sure that, know, it’s yes, customers want personalized experiences, but we still want that brand control. also don’t want our customers to find it creepy, right? know, yeah, how do you how do you recommend finding that balance?
Chris O’Neill (12:07)
Well, humans do need to be in the loop, as I said, but it’s not altogether different from suppressions that we currently do in the platform today. So there’s suppressions for regulatory reasons, there’s suppressions for all sorts of privacy to comply. We’ve been built with enterprise in mind from the very beginning. So we have those. And also suppressions as mundane as, hey, Greg’s already purchased something, let’s suppress that. That offering of the same thing he just purchased. That’s the benefit of having all the transactions and all the data.
in one place. I don’t think it’d be very different than that really, provided there’s a human in the loop. And I think part of what’s happening with again, content creation and the suggestions is that they can be, they can pick up the essence of a brand. They won’t get it all the way right. The agents aren’t quite there yet. That’s where humans will need to override it. I think ultimately in the short term, who knows where that goes in the long term, but it really is guard rails in terms of adherence suppression of certain things, again, just like what we do today with the platform. So I don’t think it’s going to be too far too big a leap at that point.
Greg Kihlstrom (13:09)
Yeah, I mean, I think it makes sense that it also a lot of that needs to be able to be automated because of the speed that we’re taught. know, like it’s even if we wanted a human to be involved in some of this stuff, the, you know, the speed at which things need to happen, know, humans need to be involved in setting those those guardrails up. But then we need to be able to trust the machines to automate it or else we can’t move in real time, near real time, stuff like that. Right.
Chris O’Neill (13:38)
Well, precisely, right? And this is the issue. The point isn’t it, it’s not just that things are manual. They actually don’t scale. There’s a reason why there’s limited number of segments that people usually carry around. There’s a reason why there’s a limited number of stimuli or creatives. It’s like, you have to go get approval and all that stuff. And really the reality is that leads to lowest common denominator kind of thinking and execution.
So it’s not only about iteration and fast, know, higher velocity. It’s like, it can scale. So you can do thousands of things, right? We do need to use algorithms to help with this, much the same way as Netflix makes suggestions to you for the next show you wanna watch or Spotify with the next song, et cetera. You know, those algorithms with the underlying data and propensity models and new AI.
LLM are getting astoundingly good and it’d be as it would be as absurd as Thinking that there’s a human behind there making a recommendation for every single one of the shows That’s not the way the world works The permutations are literally measured almost towards infinity infinity, right? When you think about all the different permutations, so it has to be Automated there has to be guardrails which can be consistently applied and then you have to let the machines and the algorithms do their work.
Greg Kihlstrom (14:57)
Yeah, well, because I mean, to other thing that you briefly touched on as well is we’re on, mean, you know, MasterCard, Visa, PayPal, unveiled shopping agents, right for consumers. you know, that’s we’re not only on the cusp like it’s kind of here. So how do you move so quickly? You know, when when I have an agent shopping on my my behalf, it’s less about that the brand color is right, then I still get what I need. But I still want everything the way that I want it. So it has to move even more quickly, which means, you know, again, we have to we have to be able to get the data quicker and make all these decisions even quicker. Right. Yeah.
Chris O’Neill (15:36)
Yeah. So I serve on the board at Gap. So I thinking about these problems through apparel quite deeply. And what’s going on is equivalent to what happened when Google disrupted Yahoo. Yahoo used to be this directory. And along comes Google. It’s like, no, this is what I want. And here it is. It’s not some directory. It goes like understands intent. Something similar and I think far more profound is happening. It’s not like I want a pair of blue jeans. It’s like, pretend for a minute I’m a
I don’t know, a teenager going to Coachella, right? I have something very specific. I pay attention to certain influencers. I’m going to Coachella, right? I want it to match with my cowboy hat. I mean, I could go on for a long time and making this fictitious example up and it’s kind of ludicrous, but you get the point, right? Those are not a pair of jeans. Those are something that does something far different. I’m looking for something far different that cannot easily be discerned and boiled down into a simple taxonomy.
It really is about exploding the variance of what the product is. That’s what the agent is going to do. And it’s going to really serve me as a consumer. And the brands that actually tap into that effectively are the ones that are going to win. And that’s not easy to do. It’s hard to do. It requires significant investments in data, in teams, in algorithms, and the tech stack that allows it all to happen seamlessly.
Greg Kihlstrom (16:53)
Yeah, and I think the other the other part of that is just becoming you know, I mentioned at the at the very top of the show that this data driven decision making, everybody talks about it, everybody says it’s important. And yet, you know, I work with some very large companies. It’s really hard to do that, right? It’s hard to let go of kind of the human intuition and all those kinds of things sometimes when the data says one thing and stuff.
How is the average enterprise doing as far as this stuff goes? Like, we still like, well, you know, have a lot of work to do still and being more data driven or, you know, where’s the average enterprise these days?
Chris O’Neill (17:31)
I’d say we’re in the early innings, I’d say average enterprise is very poor at this. And there’s a lot of good and bad reasons for that. Again, it does start with basically the underlying tech stack and systems, the fragmented nature of the data, the lack of investment in the data, not just putting it in one place like in a cloud or fewer places. It’s about investing the semantic layer to impart meaning in that layer. starts there.
And then it is, okay, how do you modernize that? So it’s not just these one size fits none platforms that really promise everything, but actually in the end are very slow, cumbersome and costly as heck. And that don’t ultimately allow you the flexibility to move at the pace. That’s the sort of what we were aiming to do better than. there’s modern tech stacks that ⁓ are very composable. You can mix and match them. You’re not locking in all that good stuff.
Historically, companies come in and they buy the belief. They say, this tool is going to solve world hunger, cold fusion, and every marketing needs you. Except for it doesn’t. That person moves on or is invited to leave, and then another person comes in with another tool, and all of a sudden you have all these tools.
Greg Kihlstrom (18:44)
It’s always the tool that’s going to solve it.
Chris O’Neill (18:46)
And then later on top of that, like the longest poll of all change ever, and this is certainly true with AI when people have deep fear about their jobs, this is going to disrupt, is, okay, yeah, think I can do better. Now that is a common thing, even though there is incredible, historically, I know, look at lot of these different companies have propensity models that say, you know, whether it’s a supply chain or assortment or marketing that are quantifiably better, and yet the human reaches in and says, no, I can do it. And again, for a whole bunch of reasons.
maybe trying to justify their existence or job, or they sometimes do think it’s better, except for they’re usually not right. So it is about doing compare and contrast, right? Sometimes there are situations, but it is smart to do that. But it is having the courage to actually trust the machine first, right? Humans in the loop, but trust the machine. It’s going to do a better job. And then also go on the change journey, right? It’s not about, going to, you know, there will be some jobs which will indeed be disrupted, of course.
But the bigger story is how do you actually use these tools to reinvent the actual workflow? That’s what we get excited about. Like, you know, look at all these disruptions that have happened have changed the consumption of media. This is true of AI. We’re consuming media in different ways, information in different ways, JetTPT, cloud, you name it. But the bigger opportunity I think is that it changes the work. It’s a supply side thing. It’s like how marketing gets created, how work goes from here to here. And you have these agent swarms with MCP and agent to agent protocols, which allow you exchange context in very rich, nuanced ways, the similar way to what TCPIP did back in the day with the internet. with the packets, right? It’s repeating itself with bigger stakes at a faster pace with, I think, more transformative potential. I haven’t even talked about business models and all this good stuff where you can start to use outcome-oriented business models. And we’re playing with some pretty powerful stuff all at the same time at a pace we’ve never seen before. it’s really, really, really fun. But I have some empathy for these companies because there is a lot, there’s a lot, both technical data and human.
Greg Kihlstrom (20:47)
Yeah, and it’s you know, it’s it’s giving up a little or it’s it feels like giving up a little bit of the control. But I mean, to your point, I, I there’s plenty of talk about A.I. being biased and all that. But there’s, you know, humans have plenty of biases as well. And, you know, cognitive, you know, anchor bias and all those kinds of things. And so, you know, I think the the partnership idea makes a lot more sense than, you know, it’s it’s us or them or something like that. Right. It’s it’s that that must be hard as a leader to to let go of to an extent to say, OK, we’re going to we’re going to let the data lead us. But it’s not really giving up on being a leader. Right. It’s there is some kind of, you know, middle ground. Right.
Chris O’Neill (21:29)
Quite the opposite. I don’t think it’s giving up. It’s actually being a better leader, right? You just have different team members and they’re called agents. even more, think individual contributors are actually now managers too. Why do I say that? Well, they’re gonna have like ⁓ metaphorically a thousand interns called agents. And what is good management entail? Well, good management means you need to set good goals, right? You need to set expectations. What do you expect? Like these agents.
If they don’t have context, they’re useless. But if they don’t know what they’re solving for, they’re also useless. Well, guess what? Agents need feedback. They’re not going to be right right away. They need to get feedback mechanisms, et cetera. So these are all hallmarks of leaders or managers, more specifically. So I think it’s quite the opposite, maybe counterintuitive to say, everyone’s going to become a manager. They’re just going be managing these things called agents. together, this is what it’s all about. I actually think of these agents as like glue people.
They’re going to glue together or create these agent swarms, like these workflows that don’t require manual stitching together. They’ll happen automatically.
Greg Kihlstrom (22:33)
Yeah, love it. Love it. Yeah. Well, Chris, thanks so much for joining today. One last question before we wrap up here. Yeah. What do you do to stay agile in your role and how do you find a way to do it consistently?
Chris O’Neill (22:45)
Well, I try to ride my bike every once in a while to say say reasonably fit but you know, I like to experiment with all these different tools. have a college just started college-aged son and a high school-aged daughter. I like to look at the world through their eyes. They teach me stuff all the time. It’s amazing how they’re using AI. So I learned from them and then I make it okay to for the team to experiment and fail with these things. That’s how I try to do it. I don’t pretend to keep up with it all. It’s I think that’s, it’s really, really difficult. So those are the things I try to do and try to have a sense of humility about it all. But but boy, it’s an interesting time and a fun time to be in business with all these all these tools and models that are are shaping our world that break neck pace.
