The agent commons: user AI & markets of one (Four Futures Series)

Speaker A: Welcome, um, back to the WPP Media Intelligence podcast and our For Future miniseries. As a quick reminder, in the series we are trying to predict one definitive version of 2035. We're testing four plausible worlds, asking what needs to happen for each to emerge and what they mean for brands, advertisers and agencies. Last week we explored the closed loop, a future where a handful of tech platforms end up controlling the AI agent, the customer relationship, and really almost the entire buying journey. And this week, we're flipping the board. We're moving to the opposite side of the matrix, the agent Commons. But before we get into it, I want to bring on my co host for this, uh, backcasting series, Sam Weston. How's it going, Sam?

Speaker B: I'm back. I was invited back for episode two. I'm.

Speaker A: You didn't screw up too badly.

Speaker B: Thrilled to be here. It's going well. It's been a busy week. I'm sure it's been a very busy week for you.

Speaker A: Yeah. Earnings, man. Oh my goodness, so many earnings. Uh, one of the things that struck me this morning listening to the Warner Brothers discovery was just the impact of not having NBA 20 percentage points headwind on advertising growth, which is massive, uh, and really speaks to fandom, the importance of sports and what people tune in for, what they care about. Are you like, is there something that you will follow? You tune in, you will like, go wherever that is playing or showing or.

Speaker B: I was just gonna say, as a relatively recent Minnesota Timberwolves convert, uh, I have been stunned by the degree to which basketball, uh, content has taken over my life and the degree to which content exists. There are, there's this plethora of, of Minnesota Timberwolves podcast options. And actually just this week I think the Wolves announced a new provider, new deal with dazone. So, um, uh, you know, I'm fully on a, on the American sports fan, uh, journey, uh, through the, through the lens of NBA basketball.

Speaker A: There you go. We'll get into where it becomes so important, but certainly the things that people still show up for in person experiential sports. Disney also announced a partnership with TikTok to bring Disney characters in approved TikTok creator vertical video onto the Disney platform, which will be new as well.

Speaker B: It's, uh, going to be a. Well, I guess after OpenAI Sora didn't pan out. That's a good switch up. Yep.

Speaker A: Yeah, so we'll see where that goes. But, you know, everything in service of what humans care about and how they align their loyalties and how that might impact, uh, where they spend their time and their money going forward. All right, so let's get into it. Uh, before we describe this current world, the Agent Commons, let's go back and, Sam, remind us, like, where we are in the world. We talked about two different axes, um, kind of defining why we picked these four diverging futures. Bring us back into context of where we're sitting today.

Speaker B: Sure. So across all four different futures, we assume that AI becomes significantly more capable, which seems pretty inevitable at this point. I think the only question is how much more capable. Uh, but within that context, we have two axes. The first asks where control lives. Is it concentrated in a few massive institutions or is it distributed across millions of different agents and services? The second asks whose interests govern the system? Is it private and commercial, uh, platforms, um, or is it public, civic and individual interests? We're going to talk about the Agent Commons, uh, today, which is a vision of a fully distributed, user directed future where individuals, rather than platforms or governments, control the agents, their memory, uh, permissions, and how they interact with the market.

Speaker A: Right. So in last week's world, the consumer had to basically step into a platform's walled garden, choose one, align with one, and kind of take what they got associated with that. In this world, every consumer brings their own world with them, their own representative to the market. And then you have this massive market of all these different individuals. So let's paint the picture. Here we are in 2035. AI agents are everywhere. They're fast, capable, deeply personal. Uh, but no single tech platform giant owns the relationship between a person and their agent. So people either assemble them or pick their own from more independent options. Maybe some are run on personal devices, others come from, uh, independent companies or developers, co ops, which would be a really interesting concept, I think. Uh, employers, public institutions. But most importantly, you kind of talked about this. The users control the agent's memory, their identity, their roles. If you want to switch, you can take your history with you. So we talked a little bit about portability last time, but this is really where it comes into its own. And then these agents are acting on the consumer's behalf across search, retail, media, payments. They can compare products, they can test claims, they can negotiate prices, uh, close deals. Right. They're doing the actual purchasing. There has to be, again, a level of trust here. But the core rule is basically the agent works for the person, not for a company paying or assumed to be paying for any kind of preferential placement. I mean, that sounds pretty good. Good, right? You'd live in this future, Sam.

Speaker B: I think that's the dream scenario for anyone with a home server or a 3D printer. Uh, and strong opinions about digital sovereignty, which I don't know if I know many people like that. Um,

Speaker A: I never talk about 3D printing. I don't know what you're talking about.

Speaker B: I think the real challenge, it sounds amazing. I think the challenge is making it work for everyone else. Um, distributed systems sound great in theory, but end up being a lot of work for people in practice, which is why we choose to live in towns and cities with electricity and plumbing and taxes and not in the woods carrying our own water and chopping our own wood. Um, people say they want control, but generally they want control without constant configuration. Uh, so for this future to sort of scale and arrive, you know, user controlled agents need to become just as seamless and reliable as handing your life over to a single end to end platform.

Speaker A: We see the same thing across privacy. Right. People often say they want privacy as an example, and some are happy to buy into a system that makes that very easy, whether that's Apple or someone else. But they often then aren't willing to pay for that privilege separately or do the work to create that security and privacy for themselves, you know. You know, not a lot of uptake on paid subscription platform like social media platforms that promise that kind of thing. So this to me is going to be the linchpin, I think, as to whether this future arrives. Can it be made dead simple and cheap for people to create these independent, personalized and autonomous agents? All right, so that raises a question. If people control this memory, identity, permissions, where do those things actually live? Right. Most consumers probably don't want to manage these on a day to day basis through code. My husband and I will often talk about things, he's a developer. I just, you know, like open the terminal and do all these things. I'm like, I'm not, no, it's not me, I'm not doing that. So. And I think there are more billions of people probably like me than like him. So is there a consumer facing product or protocol? What are we talking about here?

Speaker B: I think for us to get to this kind of future, it requires us, uh, to have something like a personal intelligence bunker or a household data vault that is really usable, um, and embraced by consumers. You would need a protected environment where you can safely store your identity, your preferences, your permissions, household policy, and the uh, memory that your agents need in order to represent you out in machine land. The key idea is that your agent wouldn't own your memory Your memory and all of your data would live in your own bunker. And you authorize models and agents to utilize it. They don't get to take it with them.

Speaker A: Yeah, I mean, you and I have had this conversation a couple of times over the, over the years. Uh, and the funny thing is it sounds really remarkably similar to how Infosum, um, works for advertisers today. Right. Data bunkers within platforms and places where there's audience data and it can maintain all of the privacy rights and security and everything that we need to make modern marketing function today.

Speaker B: Right, right. You could easily imagine a personal or household version of what we do for companies, um, being incredibly useful, maybe even essential in enabling a decentralized, federated future for consumers. Your data stays inside your protected environment. Agents use it under very specific rules. And, um, brands can submit an offer or respond to, uh, a need without receiving your entire personal history.

Speaker A: Right. So you can imagine there's a dashboard for people's relationship with the digital world. They can talk about current needs, brand preferences, uh, things around, you know, the importance of durability versus price, uh, versus fandom. How many things can you have branded with Timberwolves logos? Right. Like that could be an added bonus. You get to dial up all of these things in a personal dashboard, uh, which makes a lot of sense.

Speaker B: Right, Right. Ideally you, you have a solution that doesn't require constant supervision, uh, but gives people enough control through standing instructions and the defaults that they set so that agents can act inside those BO boundaries and really just come back to you when you have, uh, a consequential decision to make. Um, I think that's probably what it requires to scale the agent commons world for the normies.

Speaker A: Yeah. So it's easily manipulated by humans and not an onerous process, but very machine readable and portable and secure and everything else. Uh, okay, so sounds nice. Now we have to work backwards since this is backcasting. Uh, so we need to talk about what would need to be true in order for this world to take place versus any of the other possible scenarios in the multiverse. So the first one to me is that independent systems, and here we're, uh, primarily talking about models, but probably beyond that too. But independent systems have to remain competitive. If closed models, expensive paid models, are just far and away so much more proficient and therefore open source or independent models aren't a good antagonist or buffer to those, then I don't think this works. Right, Right.

Speaker B: I don't think that open, uh, models have to be the absolute best, uh, on every Benchmark, but they have to be good enough that people are no longer tempted to give up their entire digital lives to a single company in order to get best in class frontier capabilities.

Speaker A: Yeah. All right. Condition, uh, two, personal computing becomes powerful enough and this ties into, I mean we're talking about having people having their own data bunkers that comes with some physical gear. Right. So we're talking about edge devices being good. You want this to run across your entire home network. So whether that's, you know, some kind of physical compute data server, are phones and laptops going to be enough at that point? Are they strong enough to run all this edge computing locally? I'm also thinking I live in a single family home, but there are gonna be lots of people in big cities that are gonna need to be able to rent from a server rack in an apartment building probably too. Um, but basically we have to have enough computing sitting at the edge and on device to make this work.

Speaker B: And can consumers even compete for access to chips against the main providers? Right, but yeah, I mean, uh, if you don't have personal computing capacity to run this intelligence and you're deploying, you know, using cloud services or you know, your agents just dependent on another company's infrastructure and their economics and their rules,

Speaker A: I think this probably gets helped actually by what we're seeing today, which is pushes across Europe and the US potentially to limit big data centers, move to solar power. I mean, this is a world where every again, home or apartment complex is, is doing local energy generation and local, you know, compute, um, which is a very different model to what we have today. All right, nodding your head. That's very useful in radio and podcasting sounds.

Speaker B: Sorry, I'm making affirmative noises.

Speaker A: Condition 3. Memory becomes portable, but we need meaningful portability. So not just of data, but memory, context, everything that makes the agent useful. Right.

Speaker B: That's the breakthrough that everything depends on. If you can't move your model, then none of this works.

Speaker A: Right.

Speaker B: Uh, an agent that you can't leave is just another platform that you're locked into.

Speaker A: And what do you think is the drive like? Is market competition enough of a driver to push that forward or does that have to be regulation, you think? In some.

Speaker B: I think it's. I think it's both. I think you need both. I think that current and dominant providers have a very strong incentive to keep memory locked in because that's what makes agents valuable and is already making them valuable and is only going to make them more valuable and what will make leaving them incredibly painful for people. But Regulation by itself probably just gives you portability on paper that no normal person can really use. So you're going to need market competition to get a, A product that's so seamless that it can scale. Um, a rival agent should be able to make a pitch to consumers, like, bring your memory preferences and permissions and we'll have you up and running in 10 minutes.

Speaker A: Like, Port your phone number. That became a thing. You didn't used to be able to keep your phone number in the US when you changed mobile service providers. Uh, okay. Condition four.

Speaker B: You'll never get my New York number out of, out of, you know, out of my phone. There you go.

Speaker A: Condition for shared protocols emerge. So here we're talking about independent agents needing common standards for identity, permissions, product data, payments in particular, negotiation, uh, readdress when something goes wrong.

Speaker B: Right. Every, every agent needs to be able to speak to every market without requiring the user to become, you know, a systems engineer.

Speaker A: The payments piece reminds me of one of the questions we asked in the Advertising 2030 report around micro payments. Actually, like, ideally. So we asked whether micropayments would be a substantial revenue stream for publishers in 2030, and the majority of our experts said, no way. No one's made it work yet. Uh, but a big piece of that was people not signing on. It's, it's not. Without a critical mass, it's very difficult to get something like that to work. Um, which is also why I was interested in the news. This week I saw Cloudflare announcing, um, digital Wallet for agents, to my understanding, could potentially allow them to pay for content like news articles and publisher content. Um, so maybe we are now headed more in that direction, but this is definitely going to be an unlock.

Speaker B: We'll see.

Speaker A: You're still not convinced?

Speaker B: I'm skeptical. I mean, if we've learned anything from the last 20 years, it's that, uh, for anything to scale, it needs to be so easy that people don't have to think about it.

Speaker A: Yeah, but it's not you managing. You just say, I want to read about this topic and your agent is the one negotiating and navigating.

Speaker B: Totally. That could be exactly what does make it so easy that people don't have to think about it.

Speaker A: Keeping publishers alive. Yay. Okay. Ah.

Speaker B: AI to the rescue. Who knew?

Speaker A: Well, yeah, we'll see. Um, another one. Trust. We brought this up last time. Trust in platform agent. Platform agents deteriorates probably right? If people are going to wholesale shift, and again, if it's at all friction additive, then oftentimes I think we need to see a push away from something as well as a pull towards it. So is there something that happens that makes people less willing to accept a platform to owned assistant as a neutral representative?

Speaker B: I think this future never arrives unless there's some um, visible, uh, public or personally felt conflict of interest or betrayal of trust that shifts how people think and ideally you know, on a mass scale. Right. You could imagine this happening uh, if people discover that their assistant has been steering them towards products or ideas that benefit the platform at their expense. But it's also why those platforms right now are so focused on trust donates

Speaker A: to a political party that you uh, aren't affiliated with. Can you imagine?

Speaker B: Can you imagine?

Speaker A: Yeah, I mean things like fraud, misinformation as well as just misaligned priorities. Here I think is what we're watching for. Um, and last condition, I think here we're talking about agents acquiring a duty to the user. So what does that look like? And we talked about redress at the top of this. And how does payment work? Are we talking about some kind of regulation that defines like a fiduciary duty to a uh, consumer? The same way that financial advisors have to agree to something like that to be certified. Do we need a certification process? And then again who's paying for that as well?

Speaker B: But yeah, I think that's another essential requirement for this world too arrive. And I think it's also something that we may see getting discussed sooner than we think as the AI revolution becomes more and more central to policy discussions and probably sooner rather than later political campaign promises. So similar to the conversations about disclosure of AI usage in um, in advertising, you could very well imagine people arguing for a fiduciary obligation in which agents have to disclose their ownership and material commercial incentives and act according to their users stated interests. Um, because if you don't have that, you ultimately end up with the same commercial conflict of interest that you have in the closed loop future just you know, with a more reassuring label.

Speaker A: Okay, so uh, this is possible. Yeah. And some people are going to find this attractive.

Speaker B: I think people are very, will be very excited about this theory. There are a lot of people that want this future and are actively working on it. The challenge is getting it to scale.

Speaker A: So let's bring it to life a little bit more though. Like uh, what does it feel like, feel like on a daily basis we're talking about here a day, a day in the life 2035. And I tell my agent, this is a new agent maybe. So I'm Setting up some of the rules, prioritize reliability over the lowest price, buy locally where the price difference is reasonable, avoid companies that sell my data, default to brands that I already trust and have bought from before, unless there's something like a materially better alternative. And then just set it loose to check sellers and reviews product claims. Uh, and it can also then ask for offers or see what kind of upgrades or whatever, whatever else you could get for that price, negotiating terms, uh, and ultimately buys and handles the logistics.

Speaker B: Mhm. So in that world, consumers are no longer just a targetable profile inside a platform. They're essentially a market of one represented by software by a gatekeeper. So today advertisers buy access to aggregated audiences. But in this world the situation is reversed and essentially you have advertisers answering billions of individual briefs. Each brief could have a completely, would have a completely different combination of price limits, values, service requirements, brand loyalties, willingness to share data. And that would be extremely empowering for individuals and probably operationally terrifying for brands.

Speaker A: Do you mean instead of dealing with several large distributors, I have 8 billion individual customers? What's wrong with that?

Speaker B: Yeah, give her what you wish for. Maybe,

Speaker A: uh, I mean it sounds, it sounds like the Wild west because it's all rules based. I mean it is just hyper distributed in that sense. Unless you have pooling. What we've seen from agents is that they like to coordinate. So I don't know how that works, but maybe you get policies, it's right,

Speaker B: it's the Wild west, but it's a, what's the, is there a term for like a global bazaar in the Wild West? Uh, every agent has its own rules and ranking, preferences and permissions. And then every brand potentially makes different offers.

Speaker A: Right. So every brand could be formulating a bunch of different offers, but they would have the benefit of those being, you know, hyper relevant to each individual consumer and driving better loyalty probably and satisfaction off the back of that. All right, let's talk about what advertising becomes in this world. In the closed loop, advertising increasingly became the price brands paid or the system that brands had to opt into to participate in a platform's decision architecture. And here the agent sits within individual audience members, household members, bunkers.

Speaker B: Right. Advertising, I think in this, in this world splits into where you're trying to provoke emotional decision making from people and rational decision making by machines. So you want to create things like desire, values, alignment, recall brand preference with people so that they're asking their agents to prioritize your brand. But you also need to be Providing things like price availability, performance, service details, commercial terms to agents who are essentially becoming household procurement departments. People still decide what matters, but agents decide whether or not your product qualifies.

Speaker A: I think that's mostly true. I think there's still things we don't know entirely. I've mentioned this before, but Dave Clark talking about early agents opening pictures of national parks or something in the middle of whatever they're doing. They might like a catchy jingle maybe.

Speaker B: Yeah, it's true.

Speaker A: We don't know.

Speaker B: We don't know how these systems are going to develop. They may actually have strong emotions. Um, and, uh, who knows, maybe they'll be sentient.

Speaker A: So I don't know about that, but

Speaker B: maybe we'll be in a fight with our agents about what they're actually, what they think we should prioritize.

Speaker A: Yeah, you're like, I prefer the taste of Coke, but Kate Pepsi's on sale. No, it really has to be co. Yeah, all right.

Speaker B: This future's becoming less and less appealing to me, so.

Speaker A: But brands are still speaking to humans, obviously, because they have to create those brand preferences. They have to create the urge for someone, like I just said, to ask for a brand by choice because it matters so much to them. Right. So there, there is absolutely still a track of messaging and we just talked about the importance of fandom and things like live sports for engaging humans. Those are going to be more importantly the places where brands can continue to reach people. And especially if you're thinking about like, new brand introduction, where they don't have a deep history for LLMs to be searching. Or if you're trying to conquest and someone already has a product on auto order and you need to subvert that, interject yourself into the agent preference system. Those are going to be the only places I think increasingly you can do that reliably.

Speaker B: Yeah, you need strong brands in order to give people a reason to override agent freedom. So, uh, if people are just asking agents to find me a pair of running shoes, everything becomes commoditized. But if the instruction is always by Nike, unless there's a compelling reason not to, you, you're retaining the power of the brand. Ultimately, if you're a brand, you want to become a standing instruction or preference inside, uh, each customer's setup.

Speaker A: And if an agent can express everything valuable about your product in just a spreadsheet, that may not answer the human response question. There are still intangible things that matter. Humor and nostalgia and all the rest of it.

Speaker B: Though, who knows, maybe in this agent, uh, Centric future. The spreadsheet qualities of a brand will be more important than the human brand preference. We'll see. We'll see.

Speaker A: Who knows? All right. I thought we started out with such a rosy picture of this feature. Okay, so advertising takes on new forms. We're saying there have to be reasons why consumers still, you know, instruct their agents to deal with brands. There's, I don't think we're envisioning a future where people can just say, don't accept any offers or communications from brands because that is the market ultimately supposedly they want to be able to purchase or want to learn about new offerings that might suit their needs better. So it doesn't necessarily just block all commercial interests, but filters and prioritizes.

Speaker B: Yeah, people buy things, that's what people do. And we've learned that people like to buy things and uh, you know, they have a whole set of preferences for what they choose to buy. But I think that in this world, advertising to agents is less like delivering a message and more like having a standing commercial offer where you're publishing authenticated machine readable claims and prices that agents evaluate against each of their users rules.

Speaker A: All right, so does that mean promotion becomes negotiated rather than broadcast? I mean it gets a little bit into these, but we're talking about hyper personalization, people asking for specific things. There's current backlash right now against dynamic pricing. Things could come like I can pay for faster delivery and then that dynamic price is something that my agent understands and has negotiated versus being tied to a personal characteristic. Which I think is what a lot, a lot of people are uncomfortable with for very good reason.

Speaker B: Flexity is just an inherent feature of an open, um, broadly distributed future. And you can, it's complex enough to imagine a world in which brands have to deal with billions of individual customers, right, on a one to one basis. But then imagine each one of those transactions being a negotiation where those people are represented by an agent that says we'll consider your offer. But can you match this price? And also can you extend the warranty and uh, guarantee some local service? That you're just going into a world of infinite complexity.

Speaker A: This is, this is it. Because I think that the other thing with dynamic pricing as it stands right now, is that all the power is on one side. One company has algorithms that can generate that dynamic pricing because there's asymmetric information and what it knows about shoppers that shoppers don't have. But if we're saying in this world shoppers all have their own AI, that is as High powered as any distributor retailer, brand, then that asymmetric power imbalance isn't, doesn't exist. In the same way, the consumer can say, okay, I've looked at all the offers, I think I want yours, but make it worth my while. Even if pricing is dynamic for various reasons, in that sense the information asymmetry is not the cause for it. It feels less predatory in that sense.

Speaker B: One more reason to have established really strong brands that override the need to negotiate.

Speaker A: Right? So you might be willing, if the consumer's preference is it has to be Coke, it has to be Nike or whatever, then the agent accepts the higher price for those things. And that's always been true about strong brands, right? The, the ability to sell premium product is prefaced on building that brand preference over a long time, priming consumers, all the rest of it. Things aren't even that different. Sam, um, why are we doing this? Podcast 2035 and nothing's changed. All right, this is where I wanted to get to. Running agents costs money, right? The compute and the platforms and the systems to manage all these preferences and go have all these conversations with brands right now platforms can subsidize the price of AI and agents with advertising. But let's talk about the business models and the revenue models in this future because who's, who's paying for all that compute?

Speaker B: This is the, maybe the most important question in this future, right? You could have consumers paying subscriptions. Uh, perhaps agents get subsidized by employers devices. Uh, brands might also want to fund some of the service in exchange for being a ah, preference and consideration. Essentially today platforms monetize their knowledge of consumer intent. But in an agent commons, maybe consumers want to monetize that intent themselves. So instead of watch me and target me the exchanges, here's a need I have that I'm willing to disclose. Make me an offer.

Speaker A: Yeah, yeah. When I was shopping for a new car, I was very happy to see car ads because I was in active, active buying phase. Right. Once I'd bought the car, I was like, I'm done, I'm done.

Speaker B: There's nothing worse than seeing an ad after you've bought the car.

Speaker A: Yeah, yeah, exactly. All right, so you mentioned brands and devices potentially helping to fund this. I do think there's again, this wouldn't be that different from where we are today, but I think it's an interesting thing to talk through, like a potential for bifurcated access among those who can pay. You said subscriptions and maybe that's a like premium subscription model. Where we're seeing Apple move into that uh, this type of all in membership that grants people access to like a premium aging template, preloaded brand deals and subscriptions that might be entertainment, mobility services like Waymo, maybe device upgrades or home smart home stuff, connectivity and then agent instructions. Maybe some of those again are already like suggested again easy, make it easy for consumers. They might center around trust, exclusivity, customization. That might look very different to other systems in this world. Maybe those are slower older generation models that can be used without all the bells and whistles by people unable or unwilling to buy into those Premium memberships still have, you know, needs. They can still be explicit about what's important to them in terms of availability, reliability, price and brand messaging is still a part of both of those models. I just think it could look different and it's likely to look different. We're not likely to get one size fits all in this world because of you know, existing differences in populations today and people's just preferences and attitudes towards uh, personal systems.

Speaker B: There's at least a couple of good black mirror episodes in this world. The Agent Commons does not end or abolish advertising. It just makes the value exchange more explicit and gives you know, the buyer's representative uh, a seat at the table. So the danger is that the moment brands pay for consideration, you just get back to sponsored search inside the personal agent and the question is whether or not sponsorship expands choice or corrupts recommendation.

Speaker A: Yeah. All right, let's bring this, bring this back to the audience. What does this mean for advertisers, agencies? What are we talking about?

Speaker B: I think media separates into two different roles. You've got human media, uh, which is incredibly uh, entertainment and culture driven. Sport creators, communities, uh, all help form m human preferences. And then you have machine facing media which supplies structured facts, offers and evidence performance. Marketing becomes almost entirely machine to machine and brand marketing becomes unapologetically human.

Speaker A: Yeah, and there are questions about that being such a clean dichotomy but I think we do on the whole go more in that direction, sort of reverse the integration of those two elements that we've been seeing over the last couple of years. All right, so then in the closed loop the role of agency was basically to act as intermediate and protect brand from platform. In Agent Commons, what's the role of agency?

Speaker B: Well in the Agent Commons, agencies I think help protect brands from chaos. Right. Agencies help brands translate propositions into agent readable systems. They design offers and negotiation rules, they manage permissions, uh, they can test against different agent Preference models and preserve human meaning. Um, amid this radical ability to compare, um, agencies, become a market maker for markets of one agency's advantage is not, uh, owning the consumer's agent. It's understanding the population of agents better than any one advertiser can.

Speaker A: All right, so how are we going to get here? This is the M started talking about back casting, and we've taken 40 minutes just to paint the picture.

Speaker B: We need to backcast this episode.

Speaker A: All right, so what needs to happen in the near term? Right, this is where we're going to see open weight models, smaller models, edge computing, strengthen their capabilities, basically enough to support credible alternatives to platforms.

Speaker B: You've got to have. It's not just model quality, it's whether an, uh, agent can change models or service providers without losing useful memory.

Speaker A: Yeah. And then, uh, next maybe we see some of those conflicts within platform agents becoming visible. We talked about this. This was the push. Right. So the pull is really good models that you want to work with. The push is, do they really have your best interest? Who are they working for?

Speaker B: Right. By 2028, maybe we'll see a presidential campaign, for example, that features some of these issues. We'll have a very prominent election cycle in the United States. So maybe people will be asking questions about who does your agent really work for? Um, and make that much more of a mainstream consideration for people.

Speaker A: All right, so we've got a new government place in the US at that point, then 2029, 2031. In that time range, the protocols are going to have to mature. I mean, this is going to be a process, but we're going to have to get to a point where again, there are good standards for things like identity, uh, you know, digital wallets and permissions, um, dispute resolution. Those things are all going to have to be codified, you know, within the next, what, five years are we talking about?

Speaker B: Yeah, you would need to have regulation that enforces the. Well, the need, you know, for independent agents to be able to compare and negotiate and buy across major commercial ecosystems.

Speaker A: All right, and then a couple years on from that, we're at, uh, what, 2032 or so, Mass adoption. Right. So personal intelligence infrastructure becomes normal. Most new homes are built with personal data vaults and servers. Dashboards become sufficiently usable for ordinary households. And it has to be dead simple. Right.

Speaker B: It's as easy as switching your mobile provider.

Speaker A: Yeah. And, uh, by 2035, we're sort of at this consumer intent marketplace where consumers are opting into offers and messaging and brand conversations rather than being sort of the Recipient of those on the back of targeting.

Speaker B: If we've completed all these steps, then at this point material share of advertising expenditure might begin flowing through consumer controlled agents rather than platform owned media systems.

Speaker A: I mean that would be a very full scale change from where we are today.

Speaker B: Things are moving quickly. Kay.

Speaker A: All right, so uh, given that we have our three moves for 20, 26, one strategic bet, one risk to map and one signal to watch for what would point us towards this potential future. We'll start with the uh, strategic bet. What are you saying brands and advertisers should be thinking about today?

Speaker B: The strategic bet is to build the commercial interface for agents. Create clean, authenticated, machine readable information about your products, your inventory, your price, your claims and make it available through open interfaces rather than only inside individual platforms.

Speaker A: Okay, I like it. For mine, I'm going to lean into something I've been telling clients already, which is if we imagine people are going to create these personalized agents and feel comfortable opening up their, you know, wallet and purchase history, I think there's an imperative in the next year to two years to really focus on new to brand and ensure your brand shows up in as many people's past carts as possible so that there is a history of that uh, brand there. Yeah.

Speaker B: Okay. All right, uh, that's the strategic bet. What about risk? What, what risk should advertisers be mapping?

Speaker A: I mean there are some companies that are advertisers that are heavily reliant on performance media, platform ownership of audience data and profiles. And I think now today scenario planning, muscle building should be happening for like seating that top of funnel consideration building that more brand oriented storytelling. That can't be something you only decide to do sort of once your current mode of targeting is already vanishing.

Speaker B: That's exactly right. Can your brand survive radical comparability? Um, and what happens when consumer agents strip away your beautiful packaging, uh, all the smart thinking that uh, you've put behind it.

Speaker A: Yeah. All right, uh, last one. Signal to watch. What tells us that uh, this is the future we're headed for.

Speaker B: It's gotta be portability, not just capability. Uh, so you know, whether people can move, whether there's a way really for people to move agent memory and permissions and identity across services?

Speaker A: Uh, yes, I think that's right because without that you're still locked in and there's no benefit for me. I think I'm also watching how much of computing is moving to edge devices, uh, uptake of local data servers and companies that are providing that kind of, uh, tech and hardware, and then companies offering more turnkey solutions for personal data vaults. That edge world, uh, something I'm watching

Speaker B: really closely out there on the edge.

Speaker A: Oh, we need some outro. How much of music do you have to like before you pay royalties? Living on the edge. Can we play that as the outro for today?

Speaker B: We just soono it. No, not. Not quite.

Speaker A: No. All right. Thank you, Sam, as always, thank. Yeah, it was a good, good chat.

Speaker B: Uh, happy to be on this journey into the future with you.

Speaker A: As always. The WPP Media Intelligence Podcast is meant for informational and entertainment purposes only and should not be considered financial advice. We will see you next week.

Speaker B: See you next week.