[00:00:00] Welcome to Unpacking the Digital Shelf, where industry leaders share insights, strategies, and stories to help brands win in the ever-changing world of commerce. Hey everyone, Peter Crosby here from the Digital Shelf Institute. The Digital Shelf has been joined by the Agentic Shelf.
[00:00:29] With AI agents no longer just tools in our marketing arsenal, they are emerging as a powerful, influential consumer segment, actively shaping discovery, evaluation, and purchase decisions.
[00:00:42] In this audio version of a keynote from the Digital Shelf Summit, Jason Carmel, Global Creative Data Lead at VML, explains why enterprises need to rethink their marketing strategy to account for consumer machine agents, and how to do so while preserving authentic human connection. Hello. Let's see, that's me. My name is Jason Carmel.
[00:01:06] Well, I work at WPP Enterprise Solutions, VML, which is a part of that, and I'm a creative technologist. I run a creative tech group. So what that means is my job is in essence to take technology that you wouldn't touch with a barge pole, and pick it up, and smell it, and bang it on the table a few times, and then say, let's try this. And then eventually, we see it's safer and able for people to use, and we bring it to our clients, and it's great.
[00:01:36] But we're sort of at the very unsafe cutting edge of tech. And so as you might imagine, for the past probably 12 years, we've been messing around a lot with AI. And so I thought I would craft a narrative about a humanist spin on machine-to-machine marketing. We've heard just a very fantastic debate about the rise of agentic commerce, or maybe not the rise. I'm not sure anymore.
[00:02:06] And what I do know from that debate, and even before that debate intuitively, is that we are all conflicted about this. Just innately. You didn't have to watch that debate to understand that. How we are going to internalize technology and use it for commerce. As consumers, there are good sides and bad sides, and it's hard to sort of, as we ping-pong back and forth, to know whether we are optimists or apocalyptists about this. For sure.
[00:02:36] And the data supports that. Who doesn't love a good bit of data? There's a lot of promise of technology. 67% of people think technology brings people together. 76% of people think technology can make the world a better place. And more than half people think that AI itself can solve our biggest problems, our biggest existential problems. Things like climate change or cancer. That's pretty good. Thumbs up for technology. Now we go to the flip side.
[00:03:05] Which is the threat of technology. And, I mean, none of these are going to be super surprising. 71% of people think tech gets in the way of human connection. That's just probably fair. 80% believe that technology is eroding people's attention span. Sorry, what was I saying? And then 64% of people think that in a few years' time jobs won't exist, that AI will do a lot of this for us. You know what's funny about these two slides? What's funny about AI is that they're not two different people.
[00:03:34] There's not a camp that's saying, yay, technology, and then another camp over here saying, boo, technology. I mean, look at the numbers. I mean, I'm a math person, but you don't have to be to know that statistically they can't be different people who are thinking this. They're the same people having two diametrically opposing thoughts at the same time. That is inherently human, I think. Consumers don't know whether AI is the path to freedom or the chains themselves that are keeping us from being free.
[00:04:04] Which I think is fascinating. That we're sort of at this point in time and we're just, uh, your guess is as good as mine. Um, I think this may get better or worse, again, sort of depending on which part of you is talking, uh, as consumers get agents. When consumers get agents, whatever you're feeling just gets amped up. Um, and consumers don't have agents yet. Really. Or they don't know they do, which is fascinating.
[00:04:34] Um, consumers right now don't knowingly use agents. Consumers may use Rufus. That's not an agent, that's Rufus. He has a name. It's not an agent to them. They don't think that way. Um, and I know what you're thinking right now. Well, I use agents. We just heard lots of people talk about that. Um, lest, uh, you forget, and it is perhaps unkind, but none of you here is people. None of you is people.
[00:05:04] When I say people don't use agents, I'm not counting you at all. You are an anomaly. You're weird. It's true. No normal people use agents knowingly. Um, well, this was pulled from Anthropik a little while ago. Um, and you, it's, it's directionally accurate for every other, um, LLM out there, uh, that has, uh, agentic operations. Um, 49% of agents are used for software development. Nerds like me. Almost half.
[00:05:34] 27% are used for business task automation. That's a very specific flavor of nerd. That's a data nerd. And then if you go all the way down to the long tail, you're not really sure if anybody's using it for consumer-focused, uh, orientation. Uh, now once that happens, because I do believe it will happen, consumers will get agents. We're gonna have this, this dance called machine the machine, where consumers will have their machines.
[00:06:01] And, um, us marketing folks, we already have our machines. And they're gonna do a lot of the talking. And the humans are gonna be over there and way over there. And we may or may not be involved. It's weird to say out loud. Um, so how do we prep for that? Like, is it, are we sort of condemning ourselves to a world where, you know, we just sort of sit by and sort of check our phones while, while the agents do all the work?
[00:06:27] Or is there still a human element to this that we can kind of hold on to? So I thought I'd give you my opinion on this as a technologist. Um, I love this quote. Um, it's from J.C. Reese, J, J.C. Reese Anthes, a scholar at Stanford. Um, and, uh, and they say, AI is a new social entity we are coming to coexist with, where we no longer apply the same principles, practices, habits, and behaviors that we have with past technology. Uh, if, anybody you recognize that person? It's not a person.
[00:06:56] It's a trick question. That's Tilly Norwood. That's the first AI-generated actress. Um, and they, uh, it, do I say it? She? It has an agent and is actively pursuing roles. I don't know. Um, but we haven't, we're going to have a new relationship with technology as well. And so I think in my dumb little head, how do I start to cope with this?
[00:07:24] How do I, how do I figure out, especially as a marketer, how do I figure out how my consumers are going to be, um, internalizing this? And how do I prepare for that? And so I go to where I'm sure a lot of you go immediately when you're confronted with a marketing problem, which is philosophy. Uh, there's a left turn that none of you saw coming. Um, and the good news is I'm not going to require any Greek. You don't have to put on a toga. The philosophy that I am talking about very specifically, uh, is called post-humanism. And this is, uh, it's philosophy.
[00:07:54] You could go to a, a, um, a college and study this in the philosophy, uh, department. Um, post-humanism, humanism was actually relatively recent. Uh, it started, the first coined phrase of it was in the 70s, but it really took off actually with the computer, uh, um, generation. So 80s onward is where it really took off. And, and lots of philosophers are writing about this now. And essentially post-humanism is sort of the, the, um, the concept is right there in the end.
[00:08:21] It is we're no longer in a human-centric world. I mean, humans are still there, um, but now humans are sort of coexisting with these other entities outside of them, uh, in a very real and meaningful sense. Now, for post-humanism in philosophy, the stuff that we're not going to study, they talk about things like, um, how are humans adapting with, with animals? Or how are humans adapting with nature? And all of that's fascinating and completely irrelevant.
[00:08:46] Um, where I think post-humanism gets interesting for us is there's a huge wing of post-humanists who are talking about how humans adapt with technology. How they coexist with it. And one of the big principles in post-humanism is this concept of the cyborg. Now, when most of you think about cyborgs, you think you go apocalyptic. You think, uh, Terminator, things like that, right? Um, but in the world of philosophy, cyborg is actually something quite different.
[00:09:16] Um, and it's way less bleak. Um, this is a cyborg, uh, to Hollywood. These are cyborgs to, to post-humanists. So, let's unpack this for a minute here. Um, by show of hands, who here doesn't have their cell phone? Yeah. I'm talking to you now on stage. And my cell phone is in my back pocket. Like, I'm expecting a text. Sorry, guys, I gotta take this.
[00:09:46] There's no need for that. Uh, really. But we have evolved with this relationship with this piece of technology where it is essentially a part of us. Think, now think, uh, I mean, obviously we have our cell phones here. But think in the past 24 hours, when was the last time your cell phone was more than six feet away from you? So, what a philosopher in post-humanism would argue is, listen, the fact that the phone is here and not attached to my face is irrelevant.
[00:10:15] That's a rounding error. You're a cyborg. So, lean into it, is what they would say. That's what I think is really interesting about the post-humanists, is they don't view it as a, oh, God, uh, idea. It is something to actually be celebrated. It's something to be acknowledged, for sure, but actually, uh, leaned into and understood. For, uh, post-humanists, technology is not an other. We are this technology.
[00:10:42] Um, this is, uh, uh, one of my favorite quotes from a post-humanist philosophy. Christopher, um, the machine is not an it to be animated, worshipped, and dominated. The machine is us. Our process is an aspect of our embodiment. Now, this was written by, um, um, someone named Donna Horway, um, in her essay, uh, Cyborg Manifesto. I absolutely recommend that you do not read this. It's readily available.
[00:11:11] It is. It's 40 pages long. You could knock it out. But it's 40 dense pages written by a philosopher. And it gets weird fast. For sure. Um, but the parts about, uh, technology, I think, are, uh, and this was written in the 80s, too. The parts about, uh, being a cyborg were absolutely prescient for what we're dealing with with agent-to-agent marketing.
[00:11:36] Um, consumers are absolutely getting used to this idea of having technology as a wingman. We literally, 30% of adults use dating apps. They literally have technology acting as their wingman. And it's, uh, and all of this stuff is, is currently prompt dependent. We pull in ad blockers to essentially give agency to technology to determine what we see and don't see.
[00:12:05] We, for fashion boxes, if any of you use, like, the stitch fixes of the world and things like that, we give agency to them to decide how we dress. Just phenomenal. For LinkedIn, and I'm sure everyone here uses LinkedIn, you are giving agency to a technology to define your career. Like, think about that. That's bizarre. That's being a cyborg. Um, and we're starting to get to a point where agents are evolving to become prompt independent.
[00:12:34] So, if you want to date somebody, you have to ask the thing to find me somebody to date. Um, but we're getting to a point now where we're no longer having to ask. And, of course, ground zero is search. Um, and I'll blow through these because you've seen these and you will see these stats all over the place. But you know that's, that's a low estimate for zero-click search is 60%. Low estimate. Forty percent of consumers are using AI, um, to research product recommendations.
[00:13:03] And 49% trust in AI to recommend products on their behalf. And that's where the big difference is, uh, for us. That's where it starts to become super real. Um, because you all have seen this. You know how this works. Um, AI, a consumer, type in a question. What is the correct internal temperature for chicken? And it comes back with a response. The safe minimum internal temperature for poultry is 165 degrees.
[00:13:31] Now, you as, uh, as, um, marketers know how this works. It's pulling from a few high relevant sites. It's cross-referencing. Make sure it gets the same answer from everybody. And then it's printing you the answer. This is fact-based. And so, originally, this was all the AI search did. All, almost all of the queries that prompted an AI result was fact-based. There, no one's ever going to move Atlanta. So if you ask how far away Atlanta is from something else, it'll know.
[00:13:59] And that's where you'd find most of the AI responses. Uh, I mean, maybe not for, I mean, geologically recently. Uh, in the past year, we've seen a huge inflection for things that require opinions. Should I attend the Digital Self Summit this year? Yes, you should. Well, that's an opinion. Like, what, like, the AI, and a lot of, uh, Salsify people are like, oh, thank God. Oh. Like, you can unclench. You did really well.
[00:14:29] Um, uh, um, but this is an opinion. Like, this isn't right for everybody. Um, and maybe there's a universe where it says, nah, you can skip it this year, I guess. But no, it gave an opinion on, on this. And that's kind of the world, um, where we are starting to see ourselves as, where these agents are starting to recommend things that are beyond fact. So, before it was, here's how you cook dinner for your kids so you don't kill them. Um, now it is, what should you, where should you go? What is worthy of your time?
[00:14:58] How should you dress? Um, what cars should you buy? And we're entering this era of AI will buy that. Um, now we heard the debate go back and forth, uh, quite a bit on the specifics of it. But I think we can all agree that directionally, this is where we're heading. I think it sounds worse than it is. Because the way it sounds is we have, uh, two agents in the middle, kind of, who know everything and are going back and forth.
[00:15:26] And the consumer is sitting on one side, kind of bored. I guess the AI will figure out what I need and get it to me. And then the marketer on the other side is kind of bored. Well, machine will just show the right stuff and we'll just hang out and read the reports. Um, I don't think it's going to be quite that bleak. I don't think it's going to be bleak at all. Because we're entering this, um, era where, um, where I think we start to treat people more as humans. Uh, this is, uh, um, a MasterCard product that was released, uh, last year. I think it's AgentPay.
[00:15:55] It's where they're getting their credit card systems situated to actually take payments from agents. And the way, uh, they view it, and I think it's accurate, is people are no longer going to be buying necessarily a thing and then another thing and another thing. They're going to buy, uh, groups of things that answer problems. I have a birthday party for my kid, so sort it out. Why are you asking me if I need candles? Of course I need candles.
[00:16:21] So it becomes this kind of group, uh, uh, process as opposed to individual links that require individual clicking, that require individual purchasing. Um, that's how we talk as humans to each other. If I ask you where you want to go to dinner tonight, you don't give me a list of six links. That's just not human. So I think we're actually getting more and more human in our behavior on this, which is kind of interesting. We're adding this context layer on top of it. We call it sort of an attention layer as well, you know, potato, potato.
[00:16:50] But in essence, it's the same thing. It's adding this sort of human, uh, element of human language to all of these, um, this, this sort of bag of word querying that we used to do. I think it's cool. This behavior is actually very human. Now, five years or so from now, things get even weirder. We have building blocks already there. We have LLMs that can sort of act as people. They can pretend to be us. We have agentic commerce, the hooks for that.
[00:17:19] Uh, or we have independent agents able to make decisions and then we have agentic commerce. Um, so we are living in a place where, um, uh, you can give yourself a surprise gift. It's weird. 50% of Gen Z, uh, people in a study, uh, last year said they wanted to buy themselves a holiday gift. Great. What if you surprised yourself? That's kind of fun.
[00:17:43] Um, we also, we did a survey last, uh, last year, uh, e-commerce survey, uh, shopper report. And, uh, we asked people, uh, just sort of a throwaway. Every time you have a survey, you want to throw a weird one in there. So we're like, hey, would you be interested in, in buying after you died? Uh, 40% of people said yes.
[00:18:08] Now there was, there was, uh, and fair to you, there's laughing and scoffing in the audience. But 40% of people. So look to your left and look to your right. And if you don't want to do it, at least one person on the side of you does. It's too big to ignore. Now I'm not saying you need to go back to your respective organizations and say, guess what guys? We're selling to dead people. Just, you know, put on a pot of coffee. Because it's going to get weird. I don't think we're there.
[00:18:37] But what I do think is that we are getting to a point where we're going to have multiple agents doing multiple things for us at the same time. Um, and I'm going to quote from the, um, uh, Cyborg Manifesto again. Machines have made thoroughly ambiguous the difference between natural and artificial mind and body self-developing and existential design. We're going to have a customer who's going to have agents for their work at their workplace. They're going to have agents for how they spend their time outside of work. They're going to have agents to define how their legacy is after they die.
[00:19:07] They're going to have agents to do their finance. Uh, uh, dozens if not more agents for every single person. Uh, and I love this quote. We are responsible for our boundaries. We are they. This is the best part of it is that the consumer gets to define what they stuff into the agents. The consumer gets to put all their data in there, put all of their intention in there. We no longer have to guess because the agent knows.
[00:19:32] Um, and as marketers, we kind of have to make peace with this, that the eight, that the consumers are going to take control of their stuff and stuff it in agents for us to then partner with. Uh, do you remember? This kind of dates me. Uh, um, uh, but y'all remember personas? Uh, yeah, they were fun, weren't they? You get like a peanut buttered average consumer and you'd be like, meet Donna.
[00:19:54] Donna, and then, and then you would always have that one person who was a little over, uh, emphatic about things and they would get a cardboard cut out of Donna and bring them to meetings. And you'd like report, you'd have an idea and they go, well, that's a good idea, Jason, but did you ask Donna? And we knew that was wrong. Like we knew that was a metaphor that we took too literally. Because as soon as you have data like that, it's out of date.
[00:20:18] Um, and now what we have with, with agents is the ability for the, uh, data to always be up to date. The intentions are always up to date. And all we need to do is be there for it and, and, uh, and converse with it in a way that it expects. And I say all we need to do is if that's trivial. And we all know it's not. And that's why we're all here, probably. Um, and so, um, we need to prepare as well.
[00:20:45] Um, and I will suggest three ways we should be preparing. None of this will be new to you, but maybe it'll crystallize underneath this whole weird philosophy thing I'm doing. I don't know, I'm just as curious as you are. Um, the first thing we need to, uh, prepare for is content. And by, uh, prepare for I mean create. And by create I mean a lot of it. Um, I was at, uh, Adobe Summit, uh, last month. And there was a Adobe guy on stage who said, uh, um, they expected a five times increase in content.
[00:21:15] Needed to keep pace with, uh, consumer agent. Five times as much creative as you are producing now. You will need to keep up with all of these agents that live around your consumers. Five times as much. It's bananas. So if you don't have a good, uh, content production plan, if you don't have a good syndication plan, I would say, now's a really good time to start.
[00:21:41] Um, we also have to consider this idea of dual optimization, which I know you're, you're all very familiar with. But this idea of this context layer, this human, what the human wants, the human intention. And then what they, uh, and then this visibility layer of what the, the machine can see and pull in. And then we have to add all of that stuff to all of our systems. All of our front end interfaces. Our pricing. How we, uh, convey trust. Trust what our governance is. And very specifically, content mapping.
[00:22:10] Because if you have five times the amount of content and you have no structure, you're cooked. As the kids would say. Um, and so structure is the second place where you can prepare. And structure is hard, as we all know. And there are a lot of questions that agentic, um, infers with structural, uh, content. So how will you manage the flow of information coming into your system? I have information from my customer and from my customer's agent. Um, how do I sort that out?
[00:22:40] Um, how will that translate into culturally relevant content going out of your system? How do you map assets and campaigns? Um, how do you reflect multiple actors in your CRM? If you have a CRM system, uh, do you have two people? One for the consumer and one for the agent? Do you merge them somehow? Do you link them? You know how we always used to make fun of the B2B people? Oh, those poor slobs doing B2B. There but for the grace of God.
[00:23:08] Um, it was always so much harder for B2B people because it wasn't just one person they were typically selling to. There was the, the CEO and the CMO and the CIO. And then the people actually did the work. You'd have to get them on board. And like, in what order you do that? Who do you send what to at what time? And they had this, they were always looking around. They were a little wild about the eyes because they always needed way more content. And they didn't know who to send to when. And they are going to be our prophets. Because they've already lived this.
[00:23:34] As we figure out how to syndicate our content both for machines and for humans in what order so that they, um, they build and synergize rather than contradict. Um, uh, B2B people are going to be your saints. Um, as far as flow goes, I really love this slide. A colleague of mine, um, Molly Schoenthal, uh, built it. I, I feel like this slide gets re, rebuilt every six years since the beginning of e-commerce.
[00:24:04] Um, and now this is sort of a, an agentic, uh, operating model that my colleagues use, uh, I think quite effectively. Uh, and the idea is, um, you have brand in the middle. So this is sort of, uh, uh, a loose interpretation of structure. It's thematic, obviously. Um, but you have, uh, readable signals that come in. You internalize those. You send them out to create content. And then that content lives through culture and gets, uh, sort of, uh, recycled back through culture back into your systems.
[00:24:31] And then for, uh, and that's the system as it works, whether we want it to or not. The key for us is to make sure your brand lives in the middle. Of all the three questions that were, uh, that were, uh, proffered in the debate last round, uh, that was the only one that I thought had, um, zero debate. Your brand will never be more important. Never will your brand be more important. Because it's the only thing standing in between these two machine layers, uh, to keep the humanity there.
[00:24:57] So all that little blue dot of brand houses so much humanity. So much. Um, it's where you need to make sure that, uh, you're telling the right narratives. That you're creating new narratives. That your products are differentiated. Uh, that your consumers are, questions are being answered the right way. Um, uh, it is a mighty blue dot.
[00:25:22] Um, and then the third place that I would say you need to start preparing for this, uh, future of ours, uh, is people. Uh, um, tomorrow there's gonna be actually a whole session on, on this that, um, that Lauren's doing. She's way smarter than I am. Especially, uh, with things involving people. Um, so I would desperately encourage you to attend that. Uh, but here's, um, uh, here's my take. I think Martech is gonna require, uh, new roles. Or maybe not new roles, but sort of evolving roles.
[00:25:50] And just like, uh, this idea of a cyborg, we're gonna have a merging of, of roles. Um, so the first is, um, so e-commerce people are gonna need to merge with this concept of detectives. Like, who did what when? Where were you on the night of October 7th? Like, these are questions that e-commerce people are gonna have to look at their analytics and figure out. Who did this? Was it a machine or man or neither or both? Um, they need to figure that out. Now, the good news here is that, uh, e-commerce people have been doing this forever.
[00:26:20] This is not a stretch for e-commerce people. E-commerce people have been reading Agatha Christie since day one. Um, and so I don't view this as a huge stretch for you people. I think you'll be fine. Um, second role, the sort of, uh, merging collective is UX and lab tech. And first, a little note on UX. Um, a lot of, there was that, uh, idea of AI stealing jobs. Um, and a lot of people are, I think, saying that AI will take a lot of UX jobs.
[00:26:49] Um, and that is pure idiocy. Um, anytime I hear people saying, are we using AI to do our UX, I'm like, oh, kittens. Because they have no idea what's coming. No idea what's coming with respect to all of the content coming in and coming out and being organized. Um, so find these castaway UX people. Um, pay their weight in gold and put them in your organization. And then make sure that they can become lab techs, too.
[00:27:19] Because everything that they need to do, everything that they proffer, that they suspect, will need to be verified in experimentation. Now, a lot of UX folks are, come from the, the land of user-centered design. So experimentation, again, will not be new for them. They'll be very comfortable with this. Um, but it will become an increasingly large part of their vocabulary. So prepare them for that. Um, get them comfortable with the technology that allows them to do it.
[00:27:45] And then the third, uh, sort of merging role is a coder and a writer. So you have one narrative for your brand. Your brand isn't a different thing to everybody. Your brand is focused and centralized. Um, but the coder has to make sure that that is machine-readable and visible to that agent, but also human-readable to the human.
[00:28:12] And so the coder has to take what's written in human, and they have to rewrite it into code so that it becomes visible in the right way to all of these machines. Um, now remember when I said e-commerce and detectives, you guys are good. UX and lab tech is pretty second nature. Coders are in trouble. We don't write well for humans. We barely write well for other coders. You should see our comments section. It's filthy.
[00:28:42] It's, uh, almost but not quite completely useless. And we need to be better at that. Um, uh, and whether that means having a writer-pair program with your coder, if you can imagine that. I'd love to be a fly on the wall in that conversation. Um, or just making sure that your coders now understand the narrative that they're expected, um, to, uh, propagate throughout the, the machinosphere. It's going to be important. So look, the duality is not going away.
[00:29:12] Um, but I would argue it is very human. Um, remember it's not two people arguing that, um, machines are, or in trouble or machines are going to save us. It's the same human. Um, and so as we as humans, for once, um, need to prepare our own organizations and make sure that we're setting up our, our, our organizations for the best possible outcome for our customers. And I think, uh, I know we can do this.
[00:29:41] So thank you for your time. Thanks to Jason for allowing us to share this content with our audience. DigitalShelfInstitute.org is the home for diving into the best practices of the digital and agentic shelves. Become a member while you're there. Thanks for being part of our community.