In this episode
Paul Meinshausen, co-founder and CEO of Aampe, the agentic decisioning company acquired by MoEngage in 2026, argues that most of what marketers have called personalization was a complicated set of rules applied at scale, and that the control it seemed to give them was an illusion. He and Greg Kihlström cover what “agentic” actually means, why marketing has lagged product teams on personalization, where the marketer’s role moves when agents handle execution, why agentic infrastructure has to work through the CRM and engagement tools brands already run, and why campaign-level averages aren’t precise enough to measure AI.
Key takeaways
- Agentic means giving a system a goal, not instructions. Rules-based software does exactly what it is told; an agentic system works out how to reach an objective within set bounds, the way a chess engine wins without being told which moves to make. It is a continuum, not a binary.
- Personalization has mostly been humans running more complex rules. Those rules hit a complexity ceiling where they stop scaling, and to the customer on the other end they rarely feel personal.
- Marketing lags product on personalization because marketing pushes. Product responds to a customer who asked for something; marketing has to decide, with less data, when and how to reach someone first in an attention-scarce world. That is the harder problem.
- The control marketers feel is an illusion. Meinshausen calls it the “control illusion”: powerful, but not a reality.
- Working with agents mirrors moving from individual contributor to leader. The skill is delegation: setting objectives and guidelines, sharing what the business has learned, letting the system experiment, and governing the patterns that come back.
- The marketer’s workload grows rather than shrinks. There is more work than ever, but it is strategic guidance and governance instead of tactical, hyper-controlled execution.
- Agentic infrastructure has to plug into the existing stack. Aampe builds an agent for each end user, but those agents act through CRM and customer engagement platforms already in place, which is why interoperability and composability matter and why Aampe joined MoEngage.
- Brand consistency is an aspiration, not an achievement to protect. VPs of marketing at larger legacy businesses report campaigns taking two to three months to release across ops, data, and design teams; agents can pull designs from tools like Figma and serve on-brand variants inside guidelines.
- Campaign-level averages aren’t good enough to measure AI. Measure counterfactuals and incrementality at the level of the individual user and the individual message, the way streaming platforms measure content at the scene level rather than asking whether you liked the movie.
Chapters
- 00:00 — Is marketer control over customer experience an illusion?
- 01:53 — Paul Meinshausen’s background: PaySense, Aampe, and the MoEngage acquisition
- 03:32 — Who uses Aampe: Grab, a European tax app, and US consumer brands
- 04:35 — What agentic decisioning actually means
- 07:30 — Why personalization has been complex rules with a ceiling
- 09:29 — Product pull vs. marketing push
- 12:22 — The control illusion and the marketer’s new role
- 15:35 — Governance and consistency when many agents are acting at scale
- 17:03 — Why Aampe joined MoEngage: interoperability and composable systems
- 23:50 — Where performance metrics mislead when evaluating AI
- 26:30 — What we’ll be talking about a year from now
- 27:49 — How Paul stays agile
What does “agentic” actually mean in marketing?
Meinshausen defines agentic as having agency in pursuit of a goal: you specify the objective, and the system figures out how to achieve it within its bounds. That contrasts with the deterministic, rules-based software of the personal-computing era, where a human writes the instructions as narrowly as possible. His examples run from game-playing systems like AlphaGo to Waymo, which handles the tactics of driving but doesn’t pick your destination, to algorithmic trading. The bounds and parameters are what define the limits, and agency is a continuum rather than an on/off switch.
Why personalization was rules in disguise
Most marketing personalization has been a step inside a workflow that makes the output slightly more specific to the recipient. In practice that means enabling humans to run more complex rules, until they hit a complexity ceiling where the rules no longer scale. Meinshausen’s view is that from the customer’s side, the result usually doesn’t feel personal at all, which makes the term somewhat counterproductive.
Why marketing lags product teams
Product experiences respond to a customer who has opened the app and asked for something, so the search box can feel genuinely responsive. Marketing works in the other direction: it has to proactively reach a person who didn’t ask, in a world where countless things compete for their attention. Marketing has historically had less data to work with and is solving a harder problem, since anticipating what someone might want is more difficult than doing what they asked.
Where the marketer’s authority moves when agents execute
Meinshausen compares the shift to becoming a manager for the first time. Individual contributors control every word of their work; new leaders who can’t let go of that feeling tend to fail. Successful leaders set guidelines, share what the business has learned, send the team out to experiment, and govern the results iteratively. Marketers working with agents face the same transition: they can’t dictate each action without defeating the purpose, but they own strategic guidance, governance, and managing the patterns that emerge across many agents.
Is brand consistency at risk when agents take over?
Meinshausen argues brand consistency, like control, is more aspiration than achievement today. Campaigns pass through business, operations, data, and design teams, and he hears VPs of marketing at legacy businesses say a new campaign can take two to three months to release. Approaching agents from loss aversion misses the point when competitors are getting much better. Agents can work more fluidly across design systems; Aampe integrates with Figma so agents can pull designs and serve variants that stay within brand guidelines.
How should marketers measure AI performance?
Most marketing teams measure at the campaign level, averaging the effect of one send across a million people over a week. Meinshausen says that isn’t precise enough. Agentic systems call for measuring counterfactuals, controls, and incrementality at the level of the individual user and message. He compares it to how streaming platforms evaluate content scene by scene rather than asking whether viewers liked the whole film.
FAQ
What is agentic decisioning in marketing? It is an approach where an AI system is given a goal and works out how to reach it for each customer, rather than following rules a marketer wrote in advance. Aampe, now part of MoEngage, deploys an agent for each individual end user.
Why hasn’t traditional personalization felt personal to customers? According to Paul Meinshausen, it has mostly been humans running increasingly complex rules, which hit a complexity ceiling and stop scaling. The result is segment logic that rarely feels individual to the person receiving it.
What role does the marketer play when AI agents handle execution? The marketer shifts from controlling each tactical action to setting objectives, guidelines, and governance, much like a new manager learning to delegate. Meinshausen says this creates more work than ever, just not mundane, hyper-controlled work.
Why did Aampe join MoEngage? Aampe’s agents act through the CRM and customer engagement tools brands already use, so its functionality was closely connected to those platforms. Joining MoEngage, which already operated at far larger scale, was a faster path than building that infrastructure itself.
How should marketers measure the impact of AI? Move beyond campaign-level averages to measuring counterfactuals and incrementality for individual users and individual messages. Precision in measurement is what lets marketers calibrate agentic systems effectively.
About Paul Meinshausen
Paul Meinshausen is Co-founder and CEO of Aampe, an agentic AI decisioning company acquired by MoEngage in 2026, where he now leads Agentic Decisioning. Aampe’s approach deploys a dedicated agent for each individual user, built on reinforcement learning infrastructure with causal inference at the user level, and runs in production across consumer brands including Grab, Swiggy, ZenBusiness, and Taxfix. Previously he co-founded PaySense, a consumer fintech acquired by PayU in 2019, and served as its Chief Data Officer. His background spans data science, computational behavioral science, and building engineering and data teams. He has been an active angel investor since 2018.
Paul Meinshausen on LinkedIn: https://www.linkedin.com/in/paulmeinshausen/
———- Resources ———-
Aampe by MoEngage: https://www.moengage.com
MoEngage is an Agentic Customer Engagement Platform trusted by over 1,350 global consumer brands and combines deep customer analytics, agentic marketing workflows through its Merlin AI Agents, and omnichannel orchestration across messaging and surfaces, helping marketing and product teams deliver experiences that feel personal at any scale. Learn more at moengage.com
The hardest part of martech right now is telling which tools are actually earning their keep. That’s the question running through the MarTech Summit Boston agenda on October 27. We’re attending MarTech Summit Boston on October 27 as a media partner. If you want to join, the code AGILE15 saves you 15%. https://aglbrnd.co/r/c70c250b8d52dff3
Need to evaluate martech, AI, or CX platforms and don’t know where to start? The Agile Brand Guide List is now in beta. We’re building the most useful, free-to-use, and up-to-date way to assess what truly matters when evaluating platforms. See more here: https://aglbrnd.co/r/f76be3a33cc491bb
Reach your customers with Reddit. Spend $500 in ad spend, get $500 back in ad credit! Learn more: https://aglbrnd.co/r/491818c79fb1873f
Chaser is the only Slack-native project management platform that helps teams turn messages into tracked tasks, automate follow-ups, and maintain team-wide visibility, without adopting another tool. Now integrated with Claude and other GenAI tools. Learn more at trychaser.com and use code AGILEBRAND for a 3-month free trial (normal trial is 14 days).
Start building your own apps with Replit and get $20 off. Learn more: https://aglbrnd.co/r/93531742a7625a20
Enjoyed the show? Tell us more and give us a rating so others can find the show: https://aglbrnd.co/r/faaed112fc9887f3
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 check out Martechipedia, the wiki for marketing technology: https://aglbrnd.co/r/35ded3ccfb6716ba
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://aglbrnd.co/r/f1782ab12211f765
Transcript
[00:00:00] Greg Kihlström: Hi, I’m Greg Kihlström, host of The Agile Brand, and here’s a question for you: What if the control marketers think they have over customer experience was never really control at all, just rules applied at scale? Agility here isn’t about moving faster, it’s about being able to hand execution to systems that adapt in real time without losing the thread of who your brand is. Today we’re going to talk about agentic marketing and what it actually asks of enterprise leaders. We’re gonna explore what one-to-one personalization requires that a rules engine can’t deliver, no matter how many marketers you throw at it, where the marketer’s authority moves when execution is delegated and which decisions shouldn’t move at all, and what counts as credible proof when a vendor shows you an AI performance number.
To help me discuss this topic, I’d like to welcome Paul Meinshausen, CEO at Amp by MoEngage. Paul, welcome to the show.
[00:01:53] Paul Meinshausen: Thank you, Greg. Excited to be here.
[00:01:56] Greg Kihlström: Yeah, looking forward to diving in here. Before we do, though, why don’t you give a little background on your role, at, at Amp and, which was recently ac- acquired by MoEngage.
[00:02:06] Paul Meinshausen: Yeah. Cool. I, co-founded and, was, CEO of Amp. and most of my career has kind of been at the intersection of cognitive science, and thinking about how we make decisions as humans, and how machines make decisions, machine learning, AI, various forms of non-human learning and, and operating. and the last 12 years I’ve spent, primarily as that manifests in consumer technology. I, previously co-founded, and was chief data officer for a, a consumer fintech business in India called PaySense. we started that in 2015, and we’ve scaled that to be one of the largest, lending, the
[00:02:51] Paul Meinshausen: non- non-banking finance company kind of lenders in India, and that was acquired in 2019. and then in 2020 w- I started Amp with two other neuroscience, kind of cognitive science, machine learning computer scientists. and we started the company because we saw a pretty significant change happening in the way we were building and developing software and we thought that for consumer businesses specifically, that the way we built infrastructure and the way we built software for the applications that they provide to their con- customers were, were changing pretty significantly. And so that there was lots to be built there and, and we started Amp, to solve for those problems, which I’m sure we’ll talk about today.
[00:03:32] Greg Kihlström: Yeah. Yeah, definitely. And, I guess one more quick question before we dive into the, the topics here. Who are the customers, like who’s a typical customer you work with, and, and what’s the focus there?
[00:03:43] Paul Meinshausen: Right. So Amp is used by any business that is, effectively, it, it is communicating and interacting with its customers at scale. so you could think of it as any, ultimately, a consumer business or really a, even a B2B business, again, that’s operating at scale. so some, some of our customers in Southeast Asia include Grab, which is a super app, in Europe in a tax business, which like basically like the TurboTax of Europe and Germany. in, in US we, we span spectrum of entertainment and ride hailing and on demand and e- ecommerce, meditation apps, fitness apps, really, yeah. Again, any business-
[00:04:25] Greg Kihlström: So wide, wide range then. [laughs] Yeah.
[00:04:26] Paul Meinshausen: [laughs] Yeah. Basically, most businesses are interacting with their customers digitally through some product interface, and so, we serve those businesses.
[00:04:35] Greg Kihlström: Great. Great. So yeah, let, let’s dive in here, and I wanna start by having some definitions here ’cause I think, the, the buzzword of the, of the last few months in the AI world is agentic, right? So it, it’s getting attached to a lot of different thing– it used to be just it was AI powered and all that a couple years ago, now, now everything seems to be agentic. So I wanna, I wanna differentiate between kinda slapping that term on things with what you’re doing in, real agentic decisioning, the, is, is the term used, I believe. Can you talk a little bit about, what, what is agentic decisioning? What is it actually deciding? And, give, give us a little background here.
[00:05:21] Paul Meinshausen: Yeah. So the way I think of a- agentic is, it is sort of, almost like the noun-verb form of, of having agency. and when you think about having agency, you have agency in pursuit of something, some goal, some objective. and what having agency means is that you have the ability to figure out how to achieve that goal or objective. now the, the bounds and the parameters within which you operate are really the key thing that determines the limits of what the kind of systems we’re talking about. but in essence, most software that we’ve developed over, in the age of personal computing has been what you could think of as rules-based software, where a human programs the software to do, something
[00:06:07] Paul Meinshausen: that we would say is deterministic, as in you say X and it does X. And you write the instructions as specifically and as concisely and as narrowly and as As possible. agentic means more, more approaching and in the direction of how we interact with humans, where I can give you a goal and you can work towards that goal however you want to get toward it. The first agentic systems, I would say, actually w-were really built to design to play games. So you can think of like AlphaGo and he original Blue, right? And so you say, “I want you to win the game of chess,” but you don’t say which moves to move, right? more recently, in the last couple of years, we’ve introduced to fully autonomous driving. Well, what do we mean by fully autonomous? Really, we mean in terms of the tactics of driving. So we have a Waymo in San Francisco, here in San Francisco, and it can take you from point A to point B. It doesn’t choose what your point B is. it doesn’t… But, but it follows the rules of traffic, and you don’t have to tell it when to turn, when to slow down, things like that.
[00:07:10] Greg Kihlström: Yeah. Yeah.
[00:07:10] Paul Meinshausen: yeah. So, algorithmic trading is another example. You give it an objective. It’s deciding how to buy and sell stocks in order to maximize whatever your, goal or objective is. So that’s what we mean by agentic. Now, but there’s a lot of nuance under the hood, and it’s a continuum. It’s not like a binary thing.
[00:07:30] Greg Kihlström: Well, yeah, and that, I think that, that gets to, what I, what I touched on in, in the intro as well and, and certainly a, a central thesis behind Amp is, marketers have, we’ve been talking about personalization and one-to-one personalization for a very, very long time. We’ve been talking about doing it, for a very long time. But really what has ended up, for the most part, being done, let’s say pre-agentic, is just a very, a very complicated set of rules. To, to your point, it’s like s- a marketer trying to figure out how to win a chess game move by move, to use that analogy, and hoping that the customer does exactly, makes the countermove at the
[00:08:15] Greg Kihlström: right time and, and so on and so forth. So, is, is– First of all, is, is that a safe, kinda characterization of what, quote unquote, “personalization” has been? And, what does that mean for marketers who have maybe believed that they were controlling more than they were?
[00:08:33] Paul Meinshausen: It is a safe characterization. and when you get under the hood, you start to realize that marketers have what you could think of as workflows and processes. And personalization has been sort of, narrowly conceived as some th- some step within that workflow that makes the outcome that a h- a human consumer or customer would see slightly more specific to them. And as you described it, right, it’s, really so far it’s just been about enabling humans to run more complex rules. It’s just at a certain point, there’s a ceiling to the complexity of those rules. You could think of it as like a complexity ceiling where they just don’t scale well. And, from a end cons- customer perspective, a person in the world, they don’t feel very personal
[00:09:18] Paul Meinshausen: at all. and so we use that term, somewhat to, somewhat to counterproductively, I would say, even for marketing.
[00:09:29] Greg Kihlström: Yeah. Yeah, I mean, it’s kinda like when, it can get more specific, but it, that depends on c-customers taking more individual actions, right? so in other words, if there were 100 different gates that they pass through, then you could probably… But, but it all, it all kinda boils back to we have to try to anticipate and build a rule in place versus a-an agent is, is taking more actions on its own, right?
[00:09:59] Paul Meinshausen: Well, so this is a great, th- y- this is a great sort of useful, place to create a, a distinction. in fact, what you described is, right, it, in some sense, personalization requires the participation of the customer, and that’s absolutely true, right? Even a human can’t know another human if that other human is not interacting with them so a classic distinction within technology for businesses is between what you could think of as product teams and marketing teams. And product teams have been about building products that customers can come to, in a sense, as an alternative to their physical real world. So instead of going to Target, I can open the Target mobile app, okay? Now, what we’ve seen is within those applications, those have gotten far more specific and meaningful for customers and, and f- have felt more personal in the sense that you can guide that application, and it will increasingly and appropriately respond to you. So just think of the search box now. that is a powerful technology where you can just try
[00:11:08] Paul Meinshausen: to express yourself, and it’ll come back with a set of things that are– you’re looking for. Marketing, I would say, y- it’s almost like push and pull. So that’s product, and that’s sort of that the customer is pulling it, as in they’re telling you, “I wanna find something,” and you’re just responding to them. They’re opening the app. They’re deciding to do that. What marketers are trying to do is, the other direction. In a world that is extremely attention scarce, as in, like, there’s a million things trying to get any per- person’s, attention, the marketer’s job is to figure out how to proactively communicate on behalf of the business to that person. So they have to make that choice, and that is where you could sort of say marketing is far behind where product is in terms of its ability to be personal. A-and partly that’s because it gets less data. Now we’ve started to use more and more of that product data on the marketing side, and partly because it’s, it’s a, it’s a s- it’s a harder challenge actually. It’s easier to do what somebody asks you to do. It’s harder to think what that person might want and preemptively prompt them, send them a notification, send them an email, kind of provoke them to think about your problem in a way that’s appropriate for them and effective for them.
[00:12:22] Greg Kihlström: Some of this goes back to the control thing is I know it’s hard to give up the idea that, well, but shouldn’t we just send more emails or text messages or, blanket everybody with this offer because we’ll, we have the best, chance of getting a response or, what- whatever those things are. But if execution is moving to agents, what’s, what’s the, what’s the role of the marketer in, in that when, again, they, they may have to give up some … the feeling, the feeling of control. But you know what, what’s, what’s still really lies in the hands of the marketer in that, in that, relationship?
[00:13:07] Paul Meinshausen: A whole lot what you’re referring to, I like to think of as the control illusion. It’s the illusion of control. it’s not a fact, it’s not a reality. It’s, but it’s a powerful illusion. fortunately, we can learn from a sort of similar experience and, we learn it as humans, in organizations as we move into positions of leadership. So it’s, it’s a common problem that almost anyone will recognize. When you’re a individual contributor, you decide the nature of your work and you can control the quality. You feel you have the control of the quality in the sense you can, “I’m gonna write this thing and it’s gonna be really good,” in your opinion, but you know every word of the thing you’ve written. Now, now when you begin to move into a position of leadership,
[00:13:58] Paul Meinshausen: now you have other people who are being guided by you. You’re providing in kind of guidance, maybe instruction. You’re kind of setting targets, and those people are helping achieve those. Now, the first problem that people run into when they become leaders is the problem of delegation. If they don’t know how to give up some- if you give up the feeling that they control and kind of delegate, they tend to fail. What allows them to be successful is when they can sort of trust, and then it becomes a question of how do you set the guidelines, how do you set … And how do you kind of strategically govern the work that your team is doing? So you don’t hire 10 people and say, “Go do whatever you want.” You say, “Here’s what we’ve learned. Here’s what our business is about.” You provide some education and some training, and then you send them out to learn, experiment, run things,
[00:14:43] Paul Meinshausen: come back and report to you, and then it’s sort of this iterative process. and that’s very similar to what, any marketer now is having to do with agentic software and with, AI agents, is realize that they can’t tell that agent exactly what to do because that would be kind of defeating the purpose-
[00:15:02] Greg Kihlström: Yeah
[00:15:02] Paul Meinshausen: … and would prevent it from scaling. but, but they very much have a role to play in terms of strategic guidance and governance and, creatively managing the patterns that emerge from many, many, many agents or many, many team members doing things. Um-
[00:15:19] Greg Kihlström: Yeah, yeah
[00:15:20] Paul Meinshausen: … and like help them learn from each other and you’re, you’re introducing more external knowledge. So there’s so much to do. In fact, there’s more work than ever. It’s just not mundane, tactical, hyper-controlled work.
[00:15:35] Greg Kihlström: Yeah, yeah. Well, and I, I love, I love the analogy of that because, I, I mean, I do think, we are, we are moving more with, with more agentic. I mean, with the pre-agentic, let’s say, it was very dependent on, you submit a prompt or something like that and you have to wait, and it’s a very, step-by-step kind of thing. But, with, I think with agentic, supporting, it is a lot, there’s a lot of similarities to what you’re saying of, of managing a- another teammate, for instance. So that’s, yeah, I, I, I like that, that way of thinking. I guess maybe a good follow-on to that is just as we think about this at, at scale as well. So, MoEngage acquired Amp as, as, as
[00:16:20] Greg Kihlström: we, as you mentioned. This, this allows you to access a, a platform serving over 1,300 brands. Obviously, some of those very large brands with lots of team members working. there, certainly there’s a lot of… I think after the maybe AI honeymoon, a lot of companies did a lot of pilot projects and things like that. There’s a lot of focus on operationalizing AI, their, their AI adoptions and, and things like that. So, with lots of team members with their own agents kind of, doing things, like how, how do you make sure that there’s consistency within a brand and, and that kind of governance and, and guardrails?
[00:17:03] Paul Meinshausen: So here, here’s what I would say. First, let’s just think about the nature of the acquisition for a minute, and then we can reflect on that in terms of the, the, the, the question around how do you achieve something like brand consistency.
[00:17:19] Greg Kihlström: Yeah.
[00:17:19] Paul Meinshausen: Okay. So first, when… I, I would say think about this from the perspective of how technology evolves and develops, especially when it’s rapidly devol- evolving and when it’s making big jumps. So it’s not just kind of like changing things at the margin, but like whole fundamental things are changing. Whatever is changing is alm- is, is always a part of a much broader system, and consumer software that is provisioned for an end user and that is being used by a business to manage all of its processes is an incredibly complex thing. It’s full of lots of different pieces. So you look at any market map in marketing, right, and you’ll sort of see millions or, is it thousands or hundreds of different companies, different technologies.
[00:18:05] Paul Meinshausen: that I think is inevitable and it sort of, it’s, it’s almost like m- it just mirrors what we see in the natural world, where you just see an evolution of lots of things that fit into these niches. Because they’re increasingly specializing and solving very specific problems. But every once in a while, you start to sort of see, an evolution in one space create a change in another space that then moves over. So just think again of autonomous driving. We- A company like Waymo is sort of started out of a company like Waze, which is solving for maps. So you can sort of see in order to solve autonomous driving, I have to solve many different things. I have to solve the reinforcement learning element of driving, but I ha- in order to drive, I have to have a map. 20 years ago, we didn’t even have good maps. You could– I ca- I can still remember growing up with, I– what was it? MapMyQuest or m- um-
[00:18:57] Greg Kihlström: Oh, MapQuest. Yeah, yeah. I r- yeah.
[00:18:59] Paul Meinshausen: It-
[00:18:59] Greg Kihlström: I remember printing out MapQuest maps. Yeah.
[00:19:02] Paul Meinshausen: Yeah, exactly. And then you’d go out in the real world, and you’d realize, oh, it turns out there’s construction or the road has changed or something like that. So we had to solve mapping, and then in order to– we also not just solve algorithmically, we had to solve the problem of imagery and, and that’s why on top of Waymo, they have these cars. So then you have different companies merging. And in marketing technology and in B- B2B technology in general, there’s so many different suites of things that are happening, and very few businesses, especially who are operating at scale, have the luxury or the time or the interest in just kinda like sweeping through and changing everything. So if, a functionality is being offered, it has to be integrated into the landscape of all the other tools that they’re working with. This is where terms like interoperability and modularity and, kind of composable systems come in. and so AMP, when we set out to design AMP, we set out to solve the problem of how we could pro- make, infrastructure agentic in the sense of how could we develop an agent for every single end user of a business. Now, that agent has to solve the problem of learning about that person and communicating with them, but they do that through all the other tools that are already being used. So you could think of like CRM systems, customer engagement platforms. Those provide a lot of infrastructure around the delivery and message maintenance and email servers and all that stuff. And so we set out to build this specific new functionality that we call the agentic infrastructure. we grew across three continents, with customers in over thirty-five countries. but at a certain point, we also recognized that as we grew, that our functionality was very closely conne- connected to other functionalities like traditional CRM and CMP. And so we could build out that software so that we could provide it to our customers, or we could, partner with and work with other providers, that are already at a scale far larger than AMP was. and so there was this opportunity where just knowing the MoEngage founders and knowing that business, we just saw a lot– we, we sort of have a similar vision for where the industry was going, and so we said, “All right, let’s come together.” so now you can think, all right, well, how do these technologies fit? And let’s take the just the, the more simple task of saying how do I ensure brand consistency? In fact, this is a notoriously hard problem today. You think about it. In order to run marketing campaigns, you have multiple teams. You have the marketing operations team who’s kinda deciding how to navigate and orchestrate emails across different lines and all this stuff. You got a data team that’s deciding, and preserving, making sure data’s flowing here and there. Coming to brand, you have a design team, a brand team who’s saying, “Here’s a good image. Here’s a good tagline. Here’s something that resonates with our business and values.” Oftentimes, that work is being done in a design system like Figma or something else. So keeping all of those in sync is actually just a really hard problem, which is why many bigger legacy business, for instance, I’ll, I’ll constantly hear VPs of marketing say that it takes them, up to two to three months to release a new campaign ’cause there’s so many steps between first the, the operations team. The business team has to say, “Here’s our objective,” and then the business team says, “All right, here’s how we’re gonna operationalize it,” and the design team says, “Here’s how we’re gonna, create a design.” So in a sense, brand consistency is so similar to control. It’s a standard that’s sort of more illusion than anything else. It’s an aspiration. It’s not something we’ve achieved and that we’re protecting. and so yeah, we could get worse, but we could also get a lot better. And the situation most businesses are in or they’re facing is that their competitors are gonna get a lot better. So if they approach the problem from a fear and a risk sort of loss aversion mindset, they’re gonna be like, “Oh, I don’t wanna be any worse than I am today.” Really where their minds should be is, “I need to be a lot better than I am today.” And so brand consistency is one of those places where, agents can operate a lot more fluidly across these design systems. So for example, AMP, we’ve got integrations with Figma so that designers can design in Figma, and then agents can pull those designs more fluidly. marketing teams and copyright teams can generate many possible versions of those designs that are, within brand guidelines, and then those can be served to end users, just sort of as a natural part of the course of a, a daily message going out so it’s an exciting sort of problem to solve for, and it’s a, it’s a, a big part of, why, why we think agentic infrastructure exists.
[00:23:50] Greg Kihlström: And so, as, as we wrap up here, one, one last challenge to, to throw your way. We talked about a, a few things, including operationalizing, but let’s talk about measurement. And, certainly I think as much as marketers are under pressure to, again, operationalize some of these AI investments, they’re also under a lot of pressure to prove ROI and, and just continue to improve performance of, of marketing just in general. So, what, what are your thoughts on, where, where are marketers, maybe misled by performance market metrics when they’re evaluating AI, and what’s, what, what’s a b- Maybe the, a better way of, of framing this and, and thinking, thinking about this
[00:24:36] Paul Meinshausen: Yeah. So a, a general theme of agentic is it gets more precise, more specific, more personal to that end user. the measurement and metrics have to catch up with that. So what we have today, mostly m- marketing teams operate on, is often really aggregate rough average metrics. So you measure at the level of a campaign. What is a campaign? It’s a single email you sent out in a week or over the course of a week to a million different people, and you’re measuring the aggregate average effectiveness of that campaign. that’s really not good enough. m- more efficacy comes from more precision, and so what you need is much more precise measurements of counterfactuals, of controls,
[00:25:21] Paul Meinshausen: of what incrementality means at the level of the individual user, at the level of the individual message, et cetera. And so that’s a- another task that is, taken on by the agents that enables that strategic marketing manager to, again, measure more precisely, and calibrate their systems more effectively. and again, just to use a simple analogy, y- I think Hollywood used to measure the effectiveness of movies kind of like, did you like that movie?
[00:25:50] Paul Meinshausen: We all know that a movie is composed of lots of different pieces. There’s the cinematography, there’s the actor, there’s this actor, there’s this part of the movie, there’s that part of the movie. Now, platforms like Netflix and others, they are much more hyper-specific. They are able to calibrate their production of new movies at the scene level, right, at the lighting level. All very, very precise sort of granular measurement. and that’s what we want in, in marketing, is to be able to not measure aggregate performance of a single campaign over a week, but really how does this work for this much more s- sort of, narrow set of users, right down to the ul- ultimately to Greg, to Paul, to individuals.
[00:26:30] Greg Kihlström: Yeah. Yeah. Love it. Well, Paul, thanks so much for joining today and, and sharing your insights. I got two questions for you as we wrap up here. First one, if we were having this interview one year from today, what is one thing we would definitely be talking about?
[00:26:45] Paul Meinshausen: Well, thank you for having me. This has been re- really fun, and I think provocative, and I like to think of these as really helping me step back from my work and sort of saying, “Wait, what is the bigger picture here? Where are we going to? How do we communicate?” So, I’d probably be thanking you a year from now and sort of saying, “Here’s, here’s, here’s what we learned,” and, and kind of helping me think about by asking the questions you did. and then I would sort of say, try and almost certainly just, make sure we’re keeping track of how fast things are changing. So I would say the things I’m saying now to me, have been true for, for a little while, but they’re becoming more and more clear much more broadly. But it’s more important than ever as well is to kind of not think that, that we’re done, not think that we’ve caught up and that like we’re all good and we can just slow down. [chuckles] It’s not gonna happen that way. And so sort of to prepare for the next year and then to look back, it would be to sort of say, “Wait a second. How are we making sure that we’re really staying as agile as possible and open-minded to f- the fact that there’s such dramatic change happening in the, in the technology?”
[00:27:49] Greg Kihlström: Yeah. Love it. Love it. And last question for you, what do you do to stay agile in your role, and how do you find a way to do it consistently?
[00:27:58] Paul Meinshausen: Well, I try and talk to as many different people as possible, subject matter experts, industry experts like you. I think I try and get out there a lot. I was out in Dubai a couple weeks ago, like really talking to a lot of marketing and technology leaders there and it was just– it’s so fascinating. You see now I’m back in San Francisco. I’m speaking at a conference in just a, a few minutes actually here in the city so just listening, a lot, kind of sharing ideas and sort of seeing how they come back, building internally as fast as possible, reading widely. I think these are all parts of my, my practice.
