Nick Craig, Head of GTM at Rokt mParticle, joined Marcus Johnson and Principal Analyst Yory Wurmser on EMARKETER's Behind the Numbers podcast to talk about where AI agent are being built in the marketing funnel, why the technology itself stopped being the differentiator, and what settles a campaign's outcome before a single ad ever runs. A few things from the conversation worth unpacking.
AI stopped being the differentiator
AI is getting cheaper and easier to access by the month, and Nick's point was that it now shows up inside every marketing tool, whether that's creative generation, copy assistance, or campaign orchestration. The catch is that those downstream tools are only as smart as whatever slice of data they're trained on. A segment built inside an email platform only knows what that email platform has seen. Yory made the same point from the innovation side: AI has democratized what products get built, but none of that downstream work matters if the upstream data feeding it isn't right.
Which raises the obvious question: what does getting the upstream data right actually require?
Garbage in, garbage out still applies
Nick's answer was blunt. People move in real time, not in scheduled batches, and a lot of marketing data infrastructure still hasn't caught up to that. He pointed to identity resolution and data hygiene as the unglamorous work that decides whether an agent can act on what a customer is doing right now, or only on what they did in last week's export. Get that layer wrong and, in his words, it's garbage in, garbage out no matter how good the model sitting on top of it is. When that foundation is actually in place, the payoff shows up in how audiences get built.
Just tell the agent what you want
Instead of manually building a segment, a marketer can describe the audience in plain language, or go a step further and hand the agent a business goal instead of a segment definition. Nick's example was telling an agent "I'm trying to drive premium subscriptions, help me make money for my business" and letting it propose the segmentation strategy from the full data set. Marcus tied this back to a broader habit shift: the same way search went from five typed words to full paragraphs once AI search took over, marketers are still catching up to how much they can now hand an agent in one request. None of that changes Nick's advice on where to start.
Use case first, agent second
Nick's watched the same mistake happen twice, once with data collection over a decade ago and now with AI: companies reach for the technology first and ask what they can build with it, instead of starting with the outcome they actually want. His fix is to reverse the order. Define the use case, understand the goal, then let AI serve as the enabler, with testing and iteration running the whole time rather than sitting at the end as a final step. Yory's addition was that the upside isn't only speed. It's what that speed frees marketers to do next, including putting the same data to work for teams outside marketing entirely.
Different entry points, same conclusion. When you're evaluating AI tools, the model matters less than what it's trained on and how many teams across the company can actually put it to use.
Watch the full episode below or listen on Apple Podcasts, Spotify, Pandora, Stitcher, or YouTube.
Marcus Johnson 00:00
When a consumer reaches checkout, they're no longer browsing, they're buying. It's a moment of peak intent, attention, and engagement. That's where Rokt comes in. Rokt helps brands reach customers at the moment that matters most, delivering relevant offers and content that feel like a natural part of the transaction experience, not an interruption. Learn more at rokt.com Hey, gang. It's Monday, July 6th. Nick, Yory, analysts, welcome to Behind the Numbers, the e-marketing podcast made possible by Rokt. I'm Marcus. And joining me for today's conversation we have two folks. One of them is our principal analyst living in New Jersey, it's Yory Wurmser.
Yory Wurmser 00:00
Hey, Marcus. How's it going?
Marcus Johnson 00:00
Hey, fella. Very good. How are you?
Yory Wurmser 00:00
I'm doing great. A little hot, but, Yeah. Otherwise, though, I'm good.
Marcus Johnson 01:00
Ah, I can relate. We had the heat wave across the pond, just recently, and so, I can sympathize greatly. We're also joined by a special guest living in the Bay Area, so I imagine it's not, that much cooler. Head of go-to-market at Rokt, and past call is Nick Craig joins the show. Welcome.
Nicholas Craig 01:00
Thanks so much for having me, Marcus.
Marcus Johnson 01:00
Yes, sir. Thank you for being here. Anytime we have a special guest on from outside the company, we start with a speed intro. All right, gents. So, I've got two questions for Nick, one for Yory. We start with Nick. What do you do, sir, in a sentence?
Nicholas Craig 01:00
So I currently lead our go-to-market teams at Rokt mParticle, which is where we help our customers maximize the value of their customer data. We do things like powering marketing performance, improving customer engagement, and really driving overall business growth.
Marcus Johnson 01:00
Very nice. And second question to both of you, to Nick first. What's been your favorite band to see live?
Nicholas Craig 01:00
Oh, Marcus, this is, this is a tough question. I would say there is a band, though, a band that I've probably seen over the past 30 years, always has a unique performance, keeps me kind of coming back. Ooh. I'm gonna go with Red Hot Chili Peppers.
Marcus Johnson 01:00
Yes.
Nicholas Craig 02:00
And it's great band. In particular, their bassist, Flea, their guitarist, John Frusciante, always feed off each other. It's always natural. It's a band you can see over and over again, and every single time it's unique.
Marcus Johnson 02:00
Yeah. That's amazing. One of my favorite bands growing up. Just, oh, I loved listening to them. Have you seen the documentary on Netflix? I haven't got to it yet.
Nicholas Craig 02:00
I haven't
Marcus Johnson 02:00
Got to it either. It's fantastic. Okay. Yeah, I'm looking forward to seeing that. What a great choice. All right, Yory, best of luck.
Yory Wurmser 02:00
Yeah, no, I mean, that's, that's a great choice. I, I grew up listening to, to them as well, but I'm going with Brandi Carlile. I actually haven't seen her recently. Mm. But I do like seeing her because she puts on an amazing show, just goes on and on, all types of covers. Really great stuff.
Marcus Johnson 03:00
Very nice. Very nice. I had The Fray, which is not my favorite band, but in terms of my favorite show, I saw them in New York when I lived there. I forget where they were, but they, they did, like, an, an acoustic set. Maybe it was the Beacon Theatre. They did an acoustic set in the middle, and it was amazing. It was like a set change almost, like I was watching a theater production. And they had a full set at the beginning, and then, like, somewhere in the middle they broke it down and had, like, five acoustic songs and a very intimate set on stage, and then went big again for the finale, and it was just so creative. Great band too. All right, gents. Very good. Well, they're the two folks we have joining us for today's episode, and the real topic we're gonna be discussing, the agent is only as good as what it sits on. Nick, everyone has a podcast, it seems, and now everyone, has or at least is building an agent, as, you, you guys know very well. So first question to you is when you look at what the industry is calling the AI revolution in marketing technology, what pattern do you notice about where all these agents are actually being built?
Nicholas Craig 04:00
Yeah, good question, Marcus. And first off, I'll, I'll open by saying it is a really exciting time to be a marketer these days, and in particular for everything that AI is unlocking. And we are indeed. We're going through a revolution, and you can see it how every single platform is thinking, "How can I add AI onto the core components to really help out?" But with AI, it's a little bit being kind of commoditized. You know, it's, it's very easily accessible. It's becoming more inexpensive, and it's no longer becoming that overarching kind of differentiator. And you see it, the pattern you kinda spoke about is you see it all the time talked about with your downstream marketing tool sets. And I think downstream AI is real and has genuine utility, but it is becoming very creative, crowded. So that can be everything from creative generation, copy assistance, campaign orchestration. I think the core problem is downstream agents, while extremely valuable, they tend to be limited by the quality of inputs that they receive.
Marcus Johnson 04:00
Mm-hmm.
Nicholas Craig 05:00
So if you're in a downstream platform at this point, you're creating a segment, don't forget that AI layer within that segment, it can only be trained on that limited data set. And so I would say this for kinda marketers within this kinda pattern, it's always critically important to understand where your data is located because that's the data the agents are trained on. And so again, if you're sending a limited data set to a email service provider, to a demand side platform, that's all that's gonna be involved within the training. And so I would say that's why we're excited in introducing Agenta capabilities really upstream where the foundational data set is actually captured. It's somewhere we've operated for close to 13 years, and, and ultimately I think there's enormous amount of value of being able to have an agent make recommendations, help drive campaign creation when it has the full set of data, from real-time behaviors to anything on the historical basis
Yory Wurmser 05:00
Yeah, I mean, I, I agree. And I, I think one of the things that AI enables, right, has, it's democratized what can be done, what products can be built.
Marcus Johnson 05:00
Mm-hmm.
Yory Wurmser 06:00
And the result is that a lot of, you're seeing a lot of innovation on those downstream marketing, behaviors, creative or, audience segmentation, things that are built on top of that, that upstream data. And if you don't have that, get that upstream data right, then you can't do any of this stuff downstream.
Marcus Johnson 06:00
Mm-hmm. It seems as though very easy for folks to fall in love with the output of something, particularly the output format of that thing. If it gives me something quick, or it's relatively what I wanted and in a decent format, who cares where it comes from? But that shouldn't be the case. Nick, how, how do you communicate the importance of what's going on behind the scenes and the importance of the data that's supporting the model?
Nicholas Craig 07:00
Yeah. We always talk, when we talk to marketers, that you always need to take a step back and make sure that your foundation is in place. Mm-hmm. And foundation can mean so many different things, but ultimately we're talking about how are customers capturing data in real time, which is critical to focus. Mm-hmm. So we always have this saying that humans, we oper- we operate in real time. We don't operate in batches. And so it's critically important- True that we're capturing this real time information and we're able, able to create impactful campaigns downstream. So while the exciting stuff is actually executing the campaign, you really have to make sure that your data's in align for impact upstream.
Marcus Johnson 07:00
Mm-hmm.
Yory Wurmser 07:00
Yeah, no, I mean, so much of the advantage of agents is speed and how fast you can optimize, whatever, you know, workflows or your, your campaigns. And if you, if you have that ability to, where everything is connected and everything is, you know, well defined on the base level, then you can move a lot quicker too.
Marcus Johnson 07:00
Before the recording, we, we were discussing what to talk about. We, thinking of this idea of, where marketing campaign is actually won or lost, and this idea of, yeah, I guess, Nick, what, what settles the outcome before anyone runs a single ad?
Nicholas Craig 08:00
Yeah, I would say marketing campaigns performance-wise, the reality is it's dictated at multiple levels. Mm-hmm. Whether it's the upstream customer data platform, that foundational data layer that we're talking about, or if it's the email service platform, the demand side platform, both are really critical to success. But I think it's always going back to that original question, that it's really important upstream to understand where is the data available and what are those data points. Mm-hmm. I'd mentioned this, but as a marketer asking, "Am I capturing data in real time? Do I have the proper ability to bring data into centralized profiles?" Something we call identity resolution, is that actually occurring? And then of course, data hygiene. Do you have the capacity or are you taking on action to clean up the data? 'Cause again, as marketers, we all know the age-old saying again, garbage in, garbage out. Mm-hmm. So if everything's built on poor quality of data, it's gonna have a direct impact on what those downstream outcomes are eventually gonna be being.
Yory Wurmser 08:00
Mm-hmm.
Nicholas Craig 08:00
So again, I would argue that while a campaign can be won and lost at many levels, it's really truly hard to find that long-term success without that proper foundation of data currently in place- Mm-hmm.. Without making sure you're taking into account all of the customer context.
Marcus Johnson 08:00
Mm-hmm.
Yory Wurmser 09:00
Ri- what I'm, what I'd like going back to as well is just, like, what you can now do with all this increased power. R- r- you know, the, the more you can see insights from the data, you can now im- you know, implement that super fast, in various workflows upstream. And I, you know, I, I concentrate a lot on that upstream level, but it, but it's true that, you know, all those capabilities rely on, on, you know, a more basic level.
Nicholas Craig 10:00
Mm-hmm. And we saw this for years, Jordy. It's like, for instance, a lot of these marketing actions, you would often have to bring in a data engineering team. You would bring in an analytics team. And again, not that collaboration is not key to success, but naturally, when you bring in so many different teams, you're gonna find natural bottlenecks. And so marketers, really why it's an exciting time, they are fully enabled at this point with the data available to take on many of those actions themselves, and really essentially minimize that time to value. No longer do you have to wait weeks to kick off a campaign to implement a new strategy. We're not even talking about days. We're at the point with the tool sets in place that in hours you can actually develop and execute on something.
Marcus Johnson 10:00
Mm-hmm. This will be a huge part of this. W- are there any other things, Nick, that, that this, this technology is allowing marketers to do today that just wasn't possible before?
Nicholas Craig 10:00
Yeah, there's a ton. The biggest thing I think this is most exciting, again, going back to the upstream conversation, is the ability to create segments. So, you know, if you think about you have an LLM that's in place within your platform, you now have the ability to say, let's assume an example. You're a premium subscription-based company. Instead of actually creating that audience in a manual capacity, you can go through and actually just with natural language create that segment. Say I wanna target users who have visited my website, but they have yet to sign up for a premium subscription.
Marcus Johnson 10:00
Mm-hmm.
Nicholas Craig 11:00
Where I think it gets really cool is taking a step back, again, assuming the data is all available within your one tool set, now make it a more broad ask. So ask to create a segment to drive subscriptions. And in this case, you're giving the flexibility to agent to look through all the data, all of the previous results, and start to figure out, hey, here are two to three different segmentation ideas that you can make. And then even taking it, I would say, one step further is you're a marketing VP, a director of marketing, why not go into the agent? We actually have people doing this within the current state, and say, "Listen, I'm a marketing VP. I'm trying to drive premium subscriptions. This is how we make money. Please, as the agent, help me make money for my business." Mm-hmm. And this way you're giving kind of the full flexibility of the agent to really think through the strategy, and it's doing so on all the data that's available within your customer data platform. That's when these agents, we've talked about it for a little bit, but this is finally when these agents are becoming true strategic partners of these marketers hand-in-hand.
Marcus Johnson 12:00
Mm-hmm. You know, it's interesting, though, 'cause it's so, it is such a shift. It's like such a behavioral shift from how people have been doing things before, similar to, like, people would Google something, and they'd add just the least amount of words possible. You know, one, two, three, maybe five. And now with AI search, with large language models, people are putting in, you know, full context paragraphs, images, so much, more information than they did before. But it's not easy. It's, it's hard to get from one place to a- especially when you've doing, been Googling a certain way, searching online a certain way for so long. What are your thoughts here in terms of, the, the, the capability's obviously there, but getting people to kind of shift mindset in terms of how they did it before and unlocking the opportunity of, of what's in front of them today?
Yory Wurmser 12:00
Yeah, I mean, it, it is a mind shift. I think part, part of the, aftereffect of the consumer embrace of the, these technologies is, is that a lot of these marketers are consumers as w- as well.
Marcus Johnson 12:00
Mm-hmm.
Yory Wurmser 13:00
I think they, they.. I think there's gonna be a natural embrace of using, you know, more complex input, st- you know, inputs into these platforms to create more fine-tuned audiences, but also to respond more precisely to the information you're getting from the sell side, from publishers, that they're gonna give you a lot better information on, you know, where you can, place ads, where you can, you know, type of messages that, that might resonate. And if you, you know, eventually that'll all be automated, but, you know, someone who's working in marketing now can really create very precise types of audiences and campaign goals that it can use to, and find, you know, very precise ways to reach the audience.
Nicholas Craig 13:00
Mm-hmm. I would also add probably, too, like, the feedback loop is just much tighter now, right? Historically, as a marketer- Mm-hmm.. You have a campaign, you set it up in a manual fashion, you execute it, you review the results, you go back to point one at this point. You know, going back to these kinda upfront toolsets like a customer data platform, we are in real time ingesting the behavioral information. So we're starting to understand which advertisements are resonating with users with the data, and being able to action on that very quickly when those audiences. It was, again, just a very manual process beforehand. I'd like to think that we're at the state that we can get better at that-
Marcus Johnson 13:00
Mm-hmm..
Nicholas Craig 14:00
With what consumers actually see.
Marcus Johnson 14:00
Mm-hmm. Let's end by talking about,.. So for the folks listening, what's, what's one thing about how they evaluate AI in their marketing stack, that you would want them to be thinking about at this point, Nick?
Nicholas Craig 14:00
Yeah. So I think it kinda goes back to the original point. Again, AI very powerful. It is kinda commoditized at this point in that it's within every single platform.
Marcus Johnson 14:00
Mm-hmm.
Nicholas Craig 15:00
But I would argue the part that is still very problematic, it's hard to solve. It's not as if no one's solving it. It goes back to that customer context. So I, I think at the end of the day, the winners, whether it's the software company winners, whether it's the marketers working with them, it's not gonna be the company or the platform with the most AI, better AI than anybody else. It's gonna be the companies with the deepest understanding of your customers, and that's all gonna be enabled on that kind of customer context. Mm-hmm. So I'd say that, Marcus, and going back to your original question, I would ensure as a marketer when you're doing your evaluation, you're certainly taking a look at the AI models that were within it, but again, going back, make sure you're taking a look what is the underlying data with which the AI is trained on. Mm-hmm. Is it at the right place within your stack to make those proper marketing recommendations? Mm-hmm. That's probably one. The other one I always say, and this is a key one that sometimes gets overlooked, is how useful is the AI within the evaluation for your entire company? You know, I would look at tools.
Marcus Johnson 15:00
Mm.
Nicholas Craig 15:00
Certainly, there's gonna be the primary benefit for a marketer at that point, but ones that can be used by ancillary groups as well, whether that's an engineering or analytics team. You know, for 10 years in the data space, we talked about this all the time. Data, it's not an individual sport. It's a team sport. I think it's the same thing, honestly, with AI, that the more users within your company that can use that combined tool set together can find value, you're gonna start to see really the results amplify for your organization.
Marcus Johnson 16:00
Mm-hmm. Yeah, that's a really good one. People rushing to use AI in different corners of the company, and not so much thinking about how can you kind of harmoniously make this all work together at some point as well and, and benefit from, the kind of combined strategy. Uri, how about for you?
Yory Wurmser 16:00
I mean, it's, it's a very similar type of, thought, and I, I think people think when they think about agents, they're thinking a lot about efficiency, and it, it does, there is so many ways you can make workflows more efficient and speed them up. But I think they need to spend more, more people need to think about what does that enable, and what can you do with that. And part of that is also, you know, extending some of this marketing data to other parts of the organization, as Nick said, for product or other things. And parts of it is just, thinking, you know, how do you approach advertising when you can have such, you know, much more fine-tuned, analysis. And also, you also have more power to, to link data and things like that just through the power of AI.
Marcus Johnson 16:00
Mm-hmm.
Yory Wurmser 17:00
So there are all these types of marketing functions that this opens up and you may not have thought of. And so spending more time thinking about what you can, might be able to do with this, and less than just simple efficiency.
Marcus Johnson 17:00
Mm-hmm. Yeah, Nick, I got quickly, kinda piggybacking off, off Yory's point there. When it is, it is true a lot of the time people think about AI technol- technology in general, particularly AI, they're, they're leading with efficiency. What to you is one of the main kind of starting points that you think people should be thinking about first, if you could help kind of reset the industry around a certain, point, starting point?
Nicholas Craig 17:00
Yeah. It's really interesting. It- it's the same thing we've approached with data collection, again, for over a decade at this point, is that all too often you find companies or brands jumping in and thinking about, "Well, what data are we gonna collect?" Mm-hmm. And then eventually from there, "What are the use cases we wanna do?"
Yory Wurmser 17:00
Right.
Nicholas Craig 18:00
It's very much backwards, and it's- Yeah.. The same exact thing for AI. So it's the same recommendation here is that AI can be extremely powerful, but it can also be problematic if you start with that technology. So the same recommendation, taking a step back, understanding the business, understanding the potential goals that you have, and then AI really is just the enabler at this point. Mm-hmm. You know, as complex as it might have, it is the enabler for you to drive a premium subscription, for you to get maximizing customer lifetime value, to get those repeat purchasers. So again, I always say start with the use cases, start with the value, then let AI take hold as kind of the enabler to reach it, and then of course test and reiterate is always the big one. Mm-hmm. Test out how it worked. Have that kind of vicious cycle that's directly in place to continue reiterating very fastly. Yeah. And I think to your point earlier, like, that is the benefit of AI is that you can do things very fast, so you can test and reiterate over and over again without it having to take weeks or months at that point.
Marcus Johnson 18:00
Mm-hmm. An excellent point to, to end on. That's all we've got time for unfortunately for this episode, but thank you so, so much to my guests for hanging out with me today. Thank you first to Yory.
Nicholas Craig 18:00
Always great to be here.
Marcus Johnson 18:00
Yes, indeed. Thank you to Nick.
Nicholas Craig 18:00
This was great. Thank you both.
Marcus Johnson 19:00
It's a pleasure. Thank you for joining the show. Thank you to the production crew. We've got John and Luigi, I believe, hanging out, helping us out with this one. Thank you of course to everyone listening in to Buy and the Marketing Podcast made possible by ROKT. Suzy will be here on Wednesday, on the Reimagine Retail show talking to the CMO of Saatva, the mattress people, and I'll be back on Friday speaking about why Fang and the Magnificent Seven have given way to the new AI giants acronym Mangoes, and how AI's biggest players are choosing their lanes.
%20(1).png)
.png)