Hello. I’m Nicholas Thompson. I’m CEO of The Atlantic.

It is my pleasure to welcome you here for this incredible conversation we’re about to have on where the internet is going. What is the new structure? We all know it’s going to change, but what’s it going to be?

How are people going to use it? How are we going to protect their rights? How are companies going to make money?

We’ve got a fabulous panel here this morning. Aman Bhutani, he is the CEO of GoDaddy. How are you?

I’m well. Thank you for having me. Malte Kosub, he is the CEO of Parloa.

How are you? Very good. Thanks for having me.

Helena Laurent, she is the Director General of Consumers International. Good morning. Good morning, sir.

Arjun Prakash, he is the CEO of Distyl AI. Welcome. Thank you.

All right. This is going to be great. Why don’t we just start with the big question?

Why don’t we go to you, Aman? How fast is the internet changing right now, and what do you see it becoming? Start us off, and then we’ll start to argue.

I also want to say we’re going to bring the audience about 30 minutes, so get your questions, try to identify the best points of tension, bring it up then, fire away.

I think the internet is changing every week, every two, three weeks. We’re seeing new protocols. We’re seeing new ways of interaction.

Frankly, it’s all moving so fast that I haven’t met anyone who can keep up with everything. Even with tremendous effort, it’s very, very hard to keep up. To understand where it’s going, I think it’s good to very quickly anchor us in where we’ve come from.

When the internet came along, and some of us remember that world, our access to the internet was CompuServe. You dialed into a walled garden into a closed ecosystem. But what really led to innovation on the internet was the World Wide Web, where you could type in a name and seamlessly go anywhere in the world.

Now, the question is, when agents roam the internet, and there is going to be more agent traffic very, very soon here than human traffic, or it feels like that, are they going to roam free on the internet?

What does that mean for brands? What does that mean for companies? What does it mean for the platforms?

How is that really going to evolve? That’s what I’m excited about.

I mean, isn’t there a subsidiary question like, will there even be an internet?

Yeah, I think you can ask that question. Will there be an internet the way we know it today? Will there be an internet?

Yes, I think there will be an internet. But will it be behind a gate? Will it be behind a toll booth?

I personally believe it shouldn’t be. I’m a big fan of what we call the agentic open web. The web should remain open.

And innovation is driven because the web is open. That’s why anyone can join the internet and be discovered, work with customers directly. But if there’s a toll booth, that’s going to be much, much harder.

But these toll booths also depend on the internet. We have great models in the world, but they have drunken the internet to sell it back to us.

I like that. They’ve drunk the internet to sell it back to us? I think it’s the best model in the world.

You take everything that already exists to drink it in and get people to pay you for what they had access to. It’s a better interface.

It’s kind of a gross image, I’ve got to say. But I’m going to work with it. So Malte, you rely on this internet that’s been drunken in and spit out.

Is it possible? So what we’re seeing right now, the paradigm was you would go to Google, and Google would send you off to the internet. And now you go to Google.

Maybe it sends you off. Maybe it summarizes. Maybe it just kind of sucks you into Gemini.

How does that affect your business, and where do you see it going?

So we personally help brands to have agents on the homepage in their customer support hotline. So we help our brands. For our customers, it’s relevant.

And if we look at how do we believe the internet will change, we need to look back what was the internet in the past. And you had three ways how to interact with a brand. First, someone went directly to a homepage of a brand.

Second, it was some kind of advertisement. So it was a Google Ads in search. It was an advertisement on YouTube or wherever.

Or third, you purchase a product on a platform like Amazon or booking.com. Those are the three ways. Some is changing, some of them are not changing.

So, if you directly want to go to a brand, that will not change. Maybe the interface will change, it’s not a homepage anymore, it’s maybe an AI agent of a brand, but that will stay the same, right? And to interact with a platform like Amazon Booking, it might also be the same, the interface might change.

But what I think is changing dramatically is how you discover, so search will drastically change, right? So, in the past, you search and then there were a couple of links and you clicked on one and then you own the experience. But what will change in the future if you search, you will probably not own the experience because a personal assistant like ChatGPTD Gemini, they will own the experience end to end.

So, they become the new platforms, but way more powerful platforms because those will be horizontal platforms, right? It’s not just for e-commerce, not just for travel or an insurance broker, they can do everything.

So, okay, so I run a media company, this makes us nervous, right? Because if you go right now, people go in and they’re like, hey, tell me what happened in Davos. You click and there’s an article by David Frum, it’s awesome, right?

Soon, tell me what happened in Davos, Gemini will just answer. But it’s kind of worse for you, right? You make customer service agents for Booking.com, somebody goes in and they’re trying to book a trip, they go and they go to the customer service agents, they have a conversation, that’s great.

You make money, Booking.com makes money. In the future, they’re just going to ask Gemini, Gemini will produce its own agents, it’s not going to use your agents, right? How do you make money in that new web?

Yeah. So, in the end, it will still be a platform behind a ChatGPT that executes a booking for a trip, for example, right? So, you will still have that product.

So, then there will be a distribution of attention. So, some customers will go to ChatGPT to ask a customer support question on their booking, and then ChatGPT needs to interact with Booking.com. So Booking.com needs to guarantee to send the right information at the right time and we support those brands to do that.

Or a customer would directly go to the brand, to Booking.com, and try to get their questions answered and there we help as well. So it is all about the distribution, how much will then be funneled through the personal assistant like ChatGPT, Gemini, and how much will be funneled directly to the brand. And brands will fight to continue to still own the experience, right?

And it’s a bit like the platform game, right? Did someone buy a product via Amazon or did someone buy a product via the company itself? And companies try to optimize to own the experience end-to-end.

Let’s stay on this and actually, Arjun, I’m going to go to you and then Helen, I’m going to ask you a sort of a more fundamental question. You know, what Malte is talking about, it sounds like there’s a good route for his company, for my company, for lots of other companies. Assuming there’s still a web, but eventually if you just get so used to doing it within your own chatbot, you’re just used to accessing all information through Gemini or OpenAI or whatever gets invented, why would you even go to a browser?

And if you don’t go to a browser, how would you ever go directly to a website? And so then doesn’t this path that Malte is talking about, which works well for him, works well for me, doesn’t that go away?

I think we’re going to have to rethink how we engage with the internet. The models are, in my opinion, closer to utility. The value creation story in consumer is going to come from new interaction patterns that allow you to engage with intelligence in a different way.

Historically, there’s been a browser predominantly around a search-based experience. Now what’s possible is agents that can take action based on your intent and that requires an entire new set of interfaces that, honestly, I think we’re still figuring out as an industry. Voice, certain things that involve directly understanding what you’re thinking, robotics, etc.

So in the consumer space, I think the interaction pattern is going to look very different and whoever cracks that is going to fundamentally own distribution. And on the enterprise side, it’s very different because there the bottleneck has historically been the models don’t understand what is unique to your enterprise because it’s unique to your data, it’s unique to your processes.

So in the enterprise game, whoever cracks integration is going to probably end up owning enterprise in a first class way. So models are increasingly going to become utilities. Consumer is going to probably gravitate around whoever figures out interaction and distribution by corollary.

And enterprise is probably going to gravitate around whoever figures out the integration story.

Define what you mean by understand the integration story?

So the models don’t understand the data in your company. They’ve never seen it before. They also don’t understand your business processes.

So if we subscribe to the idea that we are moving to an action economy, not just an attention economy where AI is going to take actions, it needs to understand how to plug into your enterprise to take actions and do it correctly.

So now you need an entire new set of software and infrastructure to educate the models on how to exactly operate within the confines of your company, which still stays proprietary. That’s the integration problem.

And you don’t think the models will just be able to solve it, that cloud code will just look at it and figure it out?

There’s a version of the world where enterprises still want to own that because that is ultimately their secret sauce.

So they’ll want to own it, but will they be able to own it?

Of course, because it’s their data. And they’re going to still use the models as intelligence, but they’re going to still own the data, which makes them unique. And there’s going to be need for certain software that lives with the enterprise that bridges that gap.

So you can still use the intelligence, but not have the intelligence trained on your data.

Okay. Helena, let’s go to you. And I want to ask you a more sort of fundamental question, which is you’ve been here in Davos, you’ve been fighting 20 hours a day, you said.

I’m going to try to protect at least, she’s already had, I saw her, I first sat down with her 30 minutes ago, during which I think she’s had three, possibly four cups of coffee. So part of the reason you’re doing this is because consumer rights have not been entirely at the forefront as we designed the last web. Now we’re designing a whole new thing and it’s not clear they’re going to be any more at the forefront.

So why don’t you lay out what you think the fundamental principle should be as we move to whatever the heck is coming.

Yeah. And I think this is because it’s so fascinating. I started in this role in 2019 and you say things like privacy and people are like, but perhaps because of the pandemic and because of scams, 80% of people at the very least have been scammed in some way, 1 in 4 never get any sort of redress.

And I think people are starting to look at this question that we’re asking here in the halls of Davos at the dinner table because it’s about that power. You say they’re going to own the data. Well, how do we take back some of that power and make that about consumer power and consumer voice and consumer representation, because that’s us and our role in the marketplace.

So very pleased to be in this panel, even though I probably won’t understand half of what is said in terms of the technical side of things, but our network around the world has thought about this in Korea, even flagging AI washing, because let’s face it, everything is being sort of touted as AI now when it’s not.

In Brazil, really fighting for rights and how data is used around the world, thinking about what are the principles. And I think we’re sort of emerging with, we need to be basing whatever that future model is. And it would be wonderful to think of an agentic AI future where my agent bands up with others and goes out there and changes the food system.

This is the sort of the brilliant. the bright vision that you know we’re painted with all of this. If we can chart a way through that by thinking about you know transparency.

How are we supposed to judge these things if we don’t know what’s going on? If we can’t see anything? There is no way in which an individual or an NGO could go and sort of explore and understand.

There is no way in which you can get redress in that system. You know you need to be thinking about what happens when it goes wrong. We need safety and security and privacy as basics and we need a sort of not just the auditability but fairness and what fairness means can only be designed if consumers are at the table as we think about what the future of the internet looks like.

So this is great why don’t I go to Malte or Arjun and ask about an interesting trade-off here which is you we’re all gonna have agents to go out and do stuff for us right and there’ll be a little bit of a trade-off between utility and privacy right.

If I give it all of my data and I feed it all my emails and all my interest and I give it access to my brain scans it’ll be able to argue a lot better on my behalf. On the other hand it’ll create a massive new attack vector and some of the unscrupulous agents not the ones that you guys run will be you know sold off third parties. So how do you guys think about this trade-off?

So I think the best way to reconcile what you just talked about is accountability. Fundamentally if you just subscribe to the fact that these agents are scaling intelligence it’s doing the action on behalf of some human who has given it an intent. If you just subscribe to that as the principle then there should be a way to number one hold the human who provided the intent accountable and number two have governance to ensure that the AI agents are taking actions that are in line and aligned with the intent of the human.

And either of these could go wrong. You could have technology that is not aligned to the intent of the human, and then that’s a bad outcome, and that is a very major risk. And you could also have a world where the humans are not accountable, and so now people are putting out agents with mass reach, doing very unscrupulous things, and like some of the things we saw with social media, there’s no way to hold them accountable.

That’s a thing we have to figure out, and it’ll come sooner than we think it is.

Do you wanna add on to that, Malte?

Yeah, I would add that I believe it is all about transparency. So the user needs to know what is happening, what data is stored, what data is used, what actions are being taken, and there needs to be guardrails, so clear rules, what those agents can do on behalf of customers without clear oversight of a human in real time.

Right now, those rules are not there, but they need to get in place.

If I could add, I think when we talk about this, there’s a lot of philosophical alignment on an idea called the agentic open web, which does what you’re talking about, that you have your agent and it goes and works with other agents and produces this outcome, which is a little different from big agents that work within organizational boundaries, right?

Interestingly, all the things we talked about, the transparency, to be able to have audit log of what happens, there’s literally a standard called agent name service that somebody wrote and put it in IETF that said when an agent comes onto the internet, it should have these things, and as GoDaddy, we actually took that spec and we enhanced it and we implemented it, and what happens in that is that when you bring an agent to the internet, it actually gets attached to your domain name and your domain name settings, so there is no rogue agent out in the world.

With that agent, we embed security certificates, so there is no spamming. If people follow that standard, there is literally no way to do something anonymously, which is what happened with the internet, right? We came up with HTTP, a protocol that didn’t have security, that didn’t have authentication.

We connected the world and everyone used it, and it was fantastic until we realized that, oh my God, that means anyone can be anonymous and do anything. So now we need a security layer, right? 25-30 years ago, 30 years ago you wouldn’t have gone to your banking website and put in your credentials because you’re afraid of the Internet And then what did we do?

We put lock boxes everywhere, which are search, it’s SSL search Now you don’t see those lock boxes because we all feel comfortable with it because it just works You go in and type your favorite, you know Institution’s name or company’s name or whatever and you don’t guess that something bad’s gonna happen You actually trust that the Internet’s gonna deliver because there’s a global infrastructure that supports humans on the Internet It’s called Domain Name Service And it sits with SSL, Secure Socket Layer Those two technologies make every person comfortable that when you’re in the open web, you’re fine That same infrastructure can support Agents on the Internet.

It’s called Agent Name Service And that’s what actually we should move towards and we want to talk to a lot of people about that Because if we do those things a lot of things you talked about can happen and the things you guys are building that You know, you want access to the information and you want to know that it’s transparent in the Agent Name Service There is what is called a Merkle tree log, an immutable log that records every version change So if you want to go back and see that the agent was this, well who did it?

It’s the domain name It’s who owns the domain that’s registered. What did it do? What version it was?

What time and date?

But hold on Aman, so is this something that you have built in GoDaddy for GoDaddy agents?

And you have… No, we’re championing an open standard on it.

So an open standard that you’re implementing at GoDaddy. And who else is using it? Because it only matters if everybody uses it

Well, it matters. Yes, if everybody uses it, it works. But even if a few large players use it It will…

It’s a domino effect then Because if you get used to getting your agents information when you pull the domain name from DNS It’s just, you know, it’s just so simple. It’s so elegant. It’s the simplest way to do it.

The infrastructure already exists It replicates globally. It doesn’t matter where you are in the world You put up a website or you add a domain name, five minutes later, usually a few minutes later, it’s available anywhere in the world. In 24 hours, it’s definitely available.

Do you guys think this is a good idea?

I can answer. I mean, the two questions I would have are, you know, what you really want is that the… Who is controlling that?

How concentrated is that system? And who has the right to change?

No, we’re not. I’m just kidding.

Well, that’s number one.

It’s completely open. Anyone can implement the standard. Add that information to the domain record, and now you have a perfect accountability because actually what the standard does is it embeds certificates both to communicate with the agent securely and the identity of the agent.

And the identity of the agent has levels. So if somebody, it doesn’t have to be GoDaddy, authenticates the company and says, this is XYZ company, they have this certificate, we have checked with humans that this is the company that has the certificate.

Now that cannot be misused on the internet. And that’s how the internet works today. If somebody else wants to do fraud and they find some misspelling or they try to, you know, sort of usurp it some way, there is existing practice of how you nuke that globally, how you take that down.

So it really, the minute you attach it to the existing entities, now you know, like, you know, pick your favorite brand. We’re the largest domain registrar in the world. About over 20% of domain entries, when you type something in, it originated with us.

Like where it has to go, originated with GoDaddy, right? So we are very interested in domain fraud. We’re very interested in fraud for our customers.

I have, you’d be surprised how large a trust and safety team I have because people can do a huge amount of fraud on the internet because of the way it’s structured. And with agents, it’s going to get way, way worse.

That only works, and actually, let me bring in the rest of the panel. That only works if we have a, I think, if we have a web that’s kind of like the web we have now, where if you want to… I don’t know.

You want to access Sephora. It’s a brand a lot of people like. It’s through sephora.com, right, where they have a website.

If the way it works is that the way you get Sephora is through an app within the ChatGPT store or within the Gemini store or the Anthropic store.

And where does that store get to Sephora? How do they know about Sephora?

Presumably through Gemini OpenAI Anthropic.

What does Gemini do? Gemini scrapes Sephora.com to get content about Sephora.

Or Sephora feeds it in directly into Gemini. I mean Sephora.com could just be wiped out in this new paradigm. It doesn’t matter for your example what gets wiped out.

Every platform needs information about every company. A large company may input it directly and currently there’s no, like you can’t go to ChatGPT and change the model behind it. That’s not a thing, right?

The model is trained and the way it’s trained is that it scrapes the internet, drinks it in, like I said earlier, right, because it goes scrapes everything. And Google already scrapes everything. That’s how you get links back, right?

It scrapes the internet. It synthesizes it. And when you’re looking for Sephora, it then finds it.

But where did it get Sephora’s content? It got it by typing or understanding Sephora.com. The minute it does that to get that information, it should have Sephora’s agents right away.

All right. This is an amazing idea. If you have more questions about it, let’s talk about it more.

This is incredible. Before we go to audience question, about eight minutes. So I’m going to put one great idea on the table.

We’ll probably spend an hour debating whether it’s good, bad, how to make better. What are some of the other most important ideas? Things that could be done in the next two years that would lead to an internet that protects privacy and that also like lets small businesses survive, right?

Because we don’t want a world where they’re just four companies. What needs to happen?

I can go first. All right. So I love ideas where there are some really interesting ideas which start from ownership of data by people who create that data.

and then, you know, then you start to see, you know, okay well to scale that what are the licensing agreements that then can be made that continue to bring the value back to the people who create and own that data and I think those are really exciting, still small scale, you know, you’ve got examples from New Zealand around, you know, Tehiku is a brilliant media company and that would be really exciting to explore and scale up I think the other piece is about making sure that people understand what can be done in lots of consumer information and talking about this so that we’re involved in the design and then I think we’re going to have to explore, you know, there are many, many possible futures of the internet that you could explore some of them really dark where, you know, it’s actually fragmented by countries deciding that they want that control, you know, we haven’t talked about that here you want one where you’ve got the opportunity of, you know, look at 80% now have a digital, 80% of people on earth now have a digital bank account, that is net good so how do we lean into that and start, I think the answer will come from really going back to those models where there is safe regulation that helps people take more ownership of their, what they are producing and the data out there because it’s that and the data brokerage and all of the system behind that is, and the fact that we’re talking about the internet in terms of brands the internet is about online connection of people, you know, so let’s start there perhaps before we get into the sort of the advertising and the marketing side of that which came after the invention of the internet

I’ll just say that there are two things that I think we need to be able to solve and I think will change, number one is a move in general across the industry to more what I would call outcome based monetization models and so you’re going to see, for example, in consumer a version of that where now Pricing models move increasingly towards actions that things take on your behalf I actually happen to think that that is well aligned with the interest of users and in the enterprise are also going to similarly see the death of an entire set of industries that today work on The base of like pricing inputs not outputs and we’ve generally seen whenever things have been priced based on output efficiencies coming So I think you’re going to see this on the business model side and it’s going to reflect also in the technologies that get built So that’s number one

Before you go number two, I’m not gonna I want you to finish number two explain exactly a transaction That would be outcome based like what you mean by that, right?

So I’ll give you an example from enterprise, right? So today for example the way that large enterprises Do business as they work with the systems integrated they work with a few software providers And then there’s a time and materials based model, which is called labor plus and you build a project It’s very input space in this new world You’re gonna move all pricing models both the licenses for the software as well as for the labor for the systems integrators The more of an outcome based model where for example, if I want a customer experience solution like like you might be providing there I’m paving being ping based on containment or based on a margin or a Percentage for every every sale that’s made got it.

That’s finally possible at scale. That’s what now number two Number two is how do you actually now set this up in a way that is governable and creates the right incentives? I think in the consumer space It’s gonna require figuring out an ownership model that makes everybody feel like they’re part of the ecosystem Otherwise, you’re gonna end up with a small handful of people who can scale very well and create an abundant amount of wealth But has disenfranchised a very large percent of the population Finding an ownership model that brings more people into the fold to create incentives.

I think is going to be very important It’s intense incentivizes them to take part in it and also Benefit from it. So that’s on the consumer side and then on the enterprise side there’s a there’s a version of it which is making sure that the AI in enterprises is Fundamentally accountable to the people at those enterprises that are creating those agents without that You’re nobody’s gonna trust the AI within their enterprise now I’m gonna see the productivity improvements that I think we have hopefully come to expect.

Malte.

And we are in front of one of the biggest shifts in human labor that has ever happened. And in those large shifts, opportunities emerge. Because entrepreneurs can start new businesses and help to use technology to change.

So I think there will be a lot of opportunities coming. That’s the first part. The second part, to talk about pricing, I fully agree that outcome-based pricing for very clear defined vertical use cases is the future.

If we look at the enterprise, I believe we will continue to move into a consumption-based model, because I believe large enterprises, they will want to own the experience of an agent. And who then defines the outcome? Who defines what to optimize for?

An enterprise might want to have a conversation for 10 minutes in order to have a longer conversation to get to know the customer better. That creates more compute costs. But maybe that’s the outcome that the company wants to optimize for.

So I think for very clear defined use case, I’m fully with you, this will be heavily outcome-based. For large enterprises, I believe it will still continue to be consumption-based, because they want to own the experience.

All right. Sorry, just to touch on small businesses. We work with small businesses all over the world.

That’s our customer globally. To touch on your question, I fundamentally believe a better world is where there are more small businesses. None of us want a world where there are only a few choices.

Something we want a lot of choice. And if AI is going to create a ton of new jobs, it’s going to be because the technology allows micro-businesses and small businesses to provide us more of those options. So what needs to happen, was your question, what needs to happen is the new version of the internet, even with agents, needs to bring that choice to the consumer so that the consumer can have that variety of choice versus narrow down the choices somehow.

Which then I think leads to a much more negative outcome.

One last point, this point about outcomes is so important and then how that defines the measurement of the next internet. Because 80% of people have a digital bank account now but only 54% of people in the world can access emergency funds within a month if they have some sort of crisis with their kid. So now in that world we’ve measured how many people have access, probably how, you know, feedback about satisfaction, how long somebody would speak to you on the phone about your bank account but we haven’t measured does that actually help at the end.

And so can we explore what outcome means in this next world and there will be multiple pieces of that.

We’re going to move to audience questions in about one minute but very quickly Aman just said something very important and I’m curious whether you all agree which is the internet will be better and the world will be better if there are lots of small businesses and I imagine we all agree with that.

Are we heading in that direction or are we heading in the direction of a world where the internet is dominated by one player, two, three, four? Are we going in the direction of diversification or centralization?

I think you have two parallel dynamics, right? You have those personal assistants, Gemini, which become horizontal platforms that not just focus on e-commerce, travel, insurance, they can consult in everything. That could be an argument that there will be less.

On the other hand, it’s easier than ever before to start a business, right? And you see a lot of small businesses emerge. There are discussions about the first billion dollar company created by one person, right?

So we have those two dynamics and we need to enable those.

You almost did that. You’re a three billion dollar company with like very few employees, right?

Yes. So I think it’s easier than ever to create business. So there are two parallel dynamics going on.

I think it’s a matter of incentives. Right now, the incentives are set up in a way that I worry about this a lot. We are heading towards a few companies creating a lot of value.

I think as a country, country’s total output’s gonna grow, I think, but now they’re gonna be able to do more with less. And the current incentive system does not create an incentive for more people to come in and start small businesses and feel like they are able to benefit from it. And we have to rethink our economic models because there was value to being a knowledge worker and an intelligence worker in an enterprise, and now that is scalable with intelligence that’s just powered by electricity.

And we have to rethink our incentive systems for this to be an inclusive boom. It is gonna be a boom, I just hope it is an inclusive boom.

Great, so we’re gonna go to audience questions. Cutting all y’all panelists off. Please raise your hand, ask questions of our wonderful people.

If they’ve said something wildly wrong, please call them out. If you’d like to challenge them on an assumption, please do that. Okay, Helena, you can answer what you were gonna answer before I rudely cut you off because nobody raised their hands, but please raise your hands and as soon as you do, I’ll cut them all off.

I’ll keep it quick, and yeah, come on, come on, come on, because I’m sure I’m definitely wrong on so many things. I was gonna add to this that where are you seeing innovation? In food, in health, in energy, you need innovative companies coming forward and the investment in those companies that are driving a change, food environments, preventive health, all of the things that we need innovation around, that could come from a different model of the internet and in a different engagement with consumers.

That’s what we want to see, right? So it’s not just about the competition structure, which is a disaster, it’s also about where are we seeing the things that matter to us.

If I could just add, let’s just define what the micro-small business is. It’s your roofer, it’s your plumber, it’s your executive coach, it’s your coffee shop, it’s your dental office. Those are the people we’re talking about here.

Yes, I agree with your guys’ comments, but I would say being in that business serving over 20 million of those customers globally, the Internet has absolutely helped those businesses. And the cost of starting a business and letting people know what your business is has been amazingly helped by the Internet. What we need to do is make sure that as the Internet moves to agents, that we don’t break that good thing that was done.

That’s the really very important thing, that as we move to agents, that we don’t consolidate the power with a few people, with a few companies. And it’s very doable, and I’m an optimist, I’m a huge long-term optimist, and my view is that the whole reason technology exists is to help the small player compete more effectively with the big player.

And they are able to do that better and better today because of technology. And if AI is going to reach its true purpose, it’s going to do that even better.

Let me ask you a question that ties very directly into my business, but also millions of others. We at The Atlantic, we make money two ways. We sell subscriptions.

You want to read an article, at some point we ask you to pay, you have to pay, you subscribe, ideally you resubscribe, and we sell advertisements. You come to our website, you go to our app, and you see an ad for a company, and it makes you feel better about the company, and then you buy their products. How do ads work in an agentic world?

So when I have my own personal agent, and I say, hey, personal agent, go out and tell me all the news from the day, how are ads going to work and how are subscriptions going to work?

I think a lot of those things are going to be figured out.

I hope soon.

There will be ads. So OpenAI launched first concepts of ads just recently. Are there as much choice as before?

When you look at the Google search rank, maybe not, but 70% of clicks went through the first three. clicks either way, so we believe there will be ads. I think a big question is, there’s a lot of content and how do we pay the creators of content when content is surfaced in a conversational interface and we need rules for that, because otherwise platforms like yours, they have a challenge if you put content out there that is being consumed, but it doesn’t lead to customers going to your homepage, to your app.

And I think that is a topic that needs to be figured out and right now it is not.

And in fact, we may be going in the wrong direction. The one specific idea I’ve heard in Davos for rules for that is that the Trump administration should make sure there are no rules so that the big AI companies can keep snarfing the whole internet, drinking it all up and then building products that compete, which was not an idea I entirely agreed with, but there we are.

How will subscriptions work in this new model? Much of the web is based on subscriptions. If I send out an agent, I say, give me all the information I need, how will that model work?

I think subscriptions should absolutely follow an evolution of the current model. And what is the current model based on? There is an agreement between you and the aggregator of traffic.

The aggregator of traffic today, the largest one is Google search and you are a large content provider who makes amazing content. And you allow Google to have access to all your content because Google gives you links back. And you are able to drive both advertising revenue and subscription using the traffic that comes to you.

If the agentic or chat experience or AI experience breaks that fundamental relationship, then that relationship itself will have to evolve because you are not going to give your content for free to another entity that then doesn’t give you anything back in return.

All right, let’s get the audience questions going right here. Do we have multiple hands back or no? Okay.

Yeah, I’m just curious about when we look at the internet, it created different economies or jobs, right? You had the gig economy. You had a lot of different jobs come out of that.

I think when this AI boom came, there was a lot of talk on different roles. So I’d be curious to know, if we look at this new architecture of the Internet, what are some maybe macro or micro economies that could come out of it or different types of jobs?

Good question.

I think it’s a million-dollar question on what are the new jobs. We’re all sort of conscious of the jobs that we think are going to be impacted by AI, but we’re less clear about the jobs that are going to be created. Fundamentally, my thought process on that is that if you look at the technology, that’s been fantastic technology innovation the last few decades, what it has done is it has allowed more and more human beings to complete actions that otherwise they may not have fully been able to complete, that they might have needed way more expertise or way more time to do whatever that action was.

And AI is going to help them do those actions better and better. It is going to make them much stronger versions of what they want to achieve. My view of the world is that that should help more people drive their passion, their idea, bring it to the world faster, better, assuming, of course, the agentic open web exists, supported by standards, that you can actually get the traffic for it.

But as AI becomes better, people will be able to pursue more of their passion. And it doesn’t matter what that passion is. I think AI is really going to help with that.

And those should be the new opportunities.

I just want to add to that two things. Number one, I think the value of expertise is going to go up. And you’re already beginning to see that economy get created because if you’re in mathematics or in healthcare, you can actually now offer your expertise in return for the new form of data labeling.

That is already beginning to happen. And there are going to be more evolved versions of this, where expertise actually becomes incredibly valuable. So that’s number one.

Number two is I think it’s also going to create an entire new economy of people who are very good at describing what they want. Today, we’re very good at taking instructions and following instructions. The economic model has almost incentivized us all.

The educational system has also done that. But I think there’s a little bit of a re-education needed, where a good systems thinker can actually do a lot with very little and be extremely competitive. And that’s, I think, going to be very important.

And just to one point you had said earlier about subscription models, I do want to just say that subscription models have historically priced inputs, not outputs, because there’s no good way to measure value in the current world of attention.

But in this new world of an action economy where AI is taking actions, you can actually measure the output of the AI doing things. I think you’re going to move to value-based pricing across the board.

So with subscriptions, if you have an agent and you go out and you say, hey, tell me what’s interesting in the world based on my interest, what I’ve read, what I think, it comes back and it gives you a summary.

You’ll reverse engineer from that summary, what contributed to it, and then you’ll ascribe value that way?

Let’s take a more basic example. If, for example, I need a particular item, today I go, I search for it, and I go buy it myself. So I’m going to tell an agent to do it, and it’s going to make a small commission on having bought it.

That’s value-based, and you’re going to see different flavors of this.

Right. So your agent will then have some weird incentives where, well, actually, I want to go buy a rake, it’ll want to buy the higher rake if it’s commission-based.

That’s a good point, and we’re going to have to figure out how to do that in a way that’s aligned with people’s best interest.

Helena?

Yeah, so first on that point, it’s the, currently what we’re seeing is personalized pricing that ends up working in certain instances, and it can work against the consumer, right? And so unless this is designed well, you will be the target for higher pricing for exactly the same product in this instance, unless you’ve got that sort of the agent actually working for you and your ability to switch at ease, right?

So that’s this part of pricing that can work in the other way, and I don’t think people realize that, nor can we measure it or see it very easily currently. So you need that accountability. And in answer to this question, I think the fascinating part will be, I think AI will get rid of tasks that can be eaten more easily replaced.

And the really exciting piece is, for example, in those spaces where so much of healthcare spending is now on preventing disease, so little of it is now on preventing disease, for example, versus on accidents, and the pressure on our health systems is so great.

So the opportunity to look at health, well-being, how will we, AI can help you with that and can give you some sort of ways in which I can then. measure my wellness but as we all know you don’t just want to be relying on you know a technical solution for your health and your well-being you need human contact there as well so I think sort of possibly some of the places we’ll see new jobs arise are in these spaces where we need to sort of interpret the results and make sense of them as humans that that will be really interesting.

We have we have a lot of ideas and a lot of agreement on what needs to happen what makes it better can I ask a strategy question so I sometimes think of the way I sometimes think of technology on the sort of these two axes there’s like this vertical race which is the gigantic trillion dollar companies building what they’re building and then there’s horizontal race which is you know everybody else it’s small businesses it’s civil society organizations it’s academics and they’re building tools that are sometimes absorbed in this and you know this is fascinating question whether the horizontal race is building as quickly as the vertical race when you’re thinking about strategy and you’re thinking about like how do we make the world better and how do we make it fair and how do we have a more competitive world is it more important to try to change the actions of the companies on the vertical race or is it more important on the horizontal race to like hope that things like Linux get developed from you know a finished programmer in his bedroom what do you guys think?

I think the mission of our company is to make opportunity more inclusive for all so we’re focused very much on the horizontal but this is not a choice I think to make it work for the horizontal we need the trillion dollar companies or the vertical to adopt certain things so for me I’m I don’t I think we have to be clear that a better world is where the horizontal succeeds and if that means that the vertical world has to meet certain certain criteria that are human first ideas that are society first ideas then I think that’s the right decision.

And I think I said I think it hasn’t been easier to build a business right now with all the technology coming so we are moving in the right direction there but I think we need to make sure that there are not two or three platforms that actually own the internet and and control where traffic is going and I think that is the part where we need to spend time we are still at the beginning but we need to spend time because if that happens the whole horizontal thing is in danger because just two or three players control the Internet.

Question in the back? One more one more comment that was the one minute warning actually.

It’s both right because the truth and the future will be somewhere in the middle of that right we yeah.

It’s both right because the truth and actually could have been three things the vertical race the one minute warning. So you think it’s both so when you think about your efforts and you think about your massively powerful organization all the people meet in Davos it’s horizontal vertical push.

It’s both because and actually you can do that you know on the ground our members around the world can challenge you know there are ways in which you can you can challenge you can bring class action you can fight you can campaign you can boycott there are all sorts of exciting things you can do and you can also work and shape beyond the idea is to get ahead into those conversations so that we’re not fighting so that we’re we’re building.

So one of the things about conversations about agents is the ratio of BS actual insights is usually about 10 to 1 and I got to say this panel zero BS 100% content you guys are awesome thank you so much thank you for hosting us here in Davos have a wonderful rest of the day.

That was great. So good. Thank you.

So good. Thank you.