DAS kommt nach KI-Agenten: Die nächste Stufe der KI beginnt JETZT

The attention economy is the reason why companies have been doing everything they can for 20 years to take your time every single minute. Whether it's through Infinite Scroll, Autoplay or the free push notifications on your screen. Today, almost every person spends almost 7 hours a day on the screen and is expected to spend more than 6 years of their entire life

alone in social networks. But what comes next, after the attention economy and which AI developments lead to this paradigm changing now, but differently than everyone thinks. This is not going to be one of these idealistic videos of people through AI and AI-slop, we would all

...

I speak weekly with the world-famous AI experts from Germany, Switzerland, Silicon Valley and here in Hong Kong and also in China. And that's why I don't just know what our next years will be like, but above all how you will benefit from all these developments.

So, it's about the so-called "intention economy". Which simply means that companies or AI no longer wait for you to do something yourself, such as investing attention or submitting a prompt. In 2022, Meta published an inconspicuous paper called "Cicero in Science", named after the Roman speaker. Cicero is an LLM with 2.7 billion parameters, which is ridiculously small,

Diplomacy is a strategy game from 1959

in which seven players take over the great powers of Europe on the eve of the First World War. And the decisive thing is: There are no dice, there is no coincidence and no luck in this game either. Whoever moves an army does not conquer a field alone by their own strength,

Diplomacy is therefore less of a normal board game than a pure test of how well someone can

...

And that's exactly what Cicero does. Over 72 hours, the AI plays between August and October 2022 40 real online games against senseless human opponents. She read in real time

whose intentions from the chat, negotiates with them, convinces them, promises things and changes alliances without anyone noticing that he is playing against a machine. In the end, Cicero reaches more than twice the average human score and is among the best 10% of all players who have played more than one game

and that she was an AI person in front of her, people only learn after the tournament. How human Cicero acts in this way shows a small moment from the published game protocols. When the technology fails for a few meters and the AI can't answer for minutes, she apologizes to the opponent with the words "she was just on the phone with her girlfriend", although of course there was neither a girlfriend nor a phone.

But what happens when you get Cicero out of the game and connect it to a real market? Exactly this question is asked by two AI researchers at the University of Cambridge at the end of 2024, Yapuk Chowdhury and Jonny Penn, who give the whole thing in their paper "Beware the Attention Economy" their name. They describe a market

on which your attention is no longer handled, but your pure intention. As soon as AI assistants like ChatGVT are deeply integrated into your everyday life, they will record in the conversation right next to what you plan next, bundle this plan in real time

and increase it via so-called provider networks to the most providers before you have even traded. And because the same AI that you intend to read can also direct it in the same direction in a paid direction, a salesman would unnoticed from the helpful assistant

The two gentlemen at the same time serves as close to reality as you can see on Amazon, for example. Even today, Amazon's recommendation system reacts to every signal you give and the company was already allowed to patent a process called anticipatory shipping in 2013.

delivers the goods in your vicinity even before you have ordered them on the basis of the prediction that you have to buy them. The difference is only the last click, i.e. the actual decision, you still make until today. And exactly this step is just about to fall away, because Amazon's AI assistant, which is now integrated into Alexa and has been used by over 300 million people in 2025,

Buy by-for-me products already for you, also outside of Amazon, monitor prices around the clock and automatically close as soon as your desired brand is in the target area. In the intention economy, Amazon no longer asks you what you want to buy, but sells your already known purchase intention to the brand that offers the most for your agent to attack it even before you have decided.

So on the shopping page we are, to be honest, already long in the intention economy. Only nobody has officially informed us about it yet. But what about chatbots like ChatGBT or Claude?

So at the moment we still have to describe very strongly what we want and what we think.

agentic AI as a keyword, integrated into our everyday lives, then ideally she will know at some point, ideally from the AI, which also has other disadvantages, she knows what I want before I have to say it at all. She can more or less, by accompanying me in everyday life, get exactly that and then only relatively little input is necessary to control it.

That's the core idea of the intention economy, because this constant support is not a side effect for the manufacturers, but the actual business. Then they get the most private thing you can get, namely the AI that accompanies me in everyday life and gets pretty much everything I do, at least digitally, maybe even offline.

With this currency you would pay in this case and we have to think about whether we want to do that and to what degree we want to do that and especially with which manufacturers, with which companies we want to do that, especially when they are abroad. Your intention is to become a currency and the more interesting it is how voluntarily we give it. Because Oma says in the same way that exactly this AI will make our everyday life so comfortable that most people will not be able to do this business.

without long delay. And in my opinion, it has to come in the same way, not only for the comfortable everyday life, but also in real business life.

To understand what I mean, it's worth taking a look at the journey that most of us have already done with ChatGPT ourselves. At the beginning, ChatGPT was a simple web chat in which you typed prompts and got an answer back. Then came web search and tool calls. That was around the end of 2024.

The model was not only able to talk, but also to perform actions and to operate the first small apps. And since the end of 2025, we are in the middle of the age of agents, in which Codex has become the central work surface that opens AI itself as a super app.

and in which chativity and codex just merge into a single product. At the beginning you could have applications like websites programmed here, but you still had to take care of the rest like database and deployment. Today you can not only build simple applications with Codex Sites, but also let them deploy. With /goal you now only give the agent a target and let the agent be self-employed in the loop.

working for hours until it is reached and in the scene it is long said that such coding agents are no longer prompted by themselves, but only the loops, i.e. the loops, that the agents should then prompt themselves. Your agents no longer run on your computer, but continue around the clock in the cloud. So you may not work around the clock, but your agents in the cloud do.

you don't really need a lot of imagination to see how small the next step really is. So far you have the goal / goal in front of you, but what happens if the agent no longer needs this goal from you because he can use all your chats, your projects and your data

von selbst erkannt hat, was du eigentlich brauchst. Er sieht dann, dass du zum Beispiel seit Wochen an denselben Projektproblemen scheiterst und baut dir über Nacht ungefragt ein passendes Projektmanagement-Tool. Oder er bemerkt, dass die Sales KPIs nicht stimmen und schickt von sich aus ein fertiges Reporting an das Management, noch bevor jemand proaktiv danach gefragt hat.

For companies, this means that it is now more important than ever to build an infrastructure with which agents can act proactively at all. With simple PDF files or Markdown files in an Obsidian, of course, that won't work. You need a protected central database with all your corporate knowledge

all meetings, all decisions, all projects, all numbers and so on. So a single source of truth. One of the most important use cases that we have been implementing for our customers for 12 months is to build exactly such an ontology of the company, i.e. a company brain, which allows people and agents to work proactively instead of reactively and thus significantly

costs to save. You will experience all this directly in advertising itself. Because if a system already knows what you want next, it doesn't have to show you a pre-made ad, but can create one on the fly that exactly

foreign

ByteDance and Alibaba are the generative generation of ads in real time, a long-standing part of the system. At the end of this development, there is no more campaign and static creatives that are built for millions, but an ad that is only created for you at this moment, for your justly recognized intention, in exactly the second in which you are most susceptible to it. But all of this is still happening

behind the screen. The last hurdle between a company and your intention so far is your own head and that's exactly what works. Who would have thought?

The error rate was still around a third in the first version of 2025.

But the second version, presented in June 2026, reconstructs sentences in real time with a word accuracy of 61%, even 71% for the best testers. It still needs a closed laboratory scanner, but the direction is clear. Because at the end of this development there is an intention that is no longer read from your words or your behavior, but directly

For invasive BCI, i.e. brain-computer interface approaches, the claim is that the thoughts are free, who can guess? Yes, after all, already a relic of the past. In BCI you can read the thoughts, read them out of the cell, as well as add them if you want. With the new BCIs we have thousands of electrodes and with that we can of course

The manipulation of thoughts is practically unlimited. If you know where the thoughts or ideas are going,

Yes, of course, with all the chances, it is always important to emphasize the shadow sides. That is perhaps also the actually bitter part of this whole story. Because the intention economy was originally thought of as the exact opposite.

The term has already been coined by the American author Doc Searles in 2006, one of the co-writers of the legendary Cluetrain Manifesto, and he means something that is deeply customer-friendly: "A world in which no longer companies are to be fooled and advised about your attention,

but in which you as a buyer sit at the wheel, deliberately send your intentions and to your conditions, control your own data and the providers compete to make the best offer for you. In his own words, providers should no longer guess what the customer's attention is catching, but react to their actual intentions. 18 years later, in 2024, the same Doc Searles has to look at it as his term in the opposite direction.

The researchers did not mean the intention economy that developers have been working on for 18 years, but something completely different. A bad thing for which they only use the same name and brings betrayal into

In one sentence on point: His intention economy was an optimistic, possibly world-changing vision, while hers was a pessimistic, the world-worn possibility. From the tool that should put the buyer in control, it has become in less than two decades exactly the machine that your own inventor warns today. So the actual question of the next few years is not:

whether this intention economy is coming. It has been there for a long time in many places and already clearly visible in the other, but on which side of it you stand. You can either be the product whose intention

It's basically the exact difference that Doc Searles has made since 2006.

And of course, this whole view today will still sound very strange to many. You might be smiled at for thinking about having agents working proactively for you and thus of course also giving away control a bit.

But always remember that in the end it's the crazy ones and the pioneers who win the future and not the skeptics who in the end don't even understand in which world they find themselves.

What are your thoughts on all this and are you interested in these videos in which we talk about the great, extraordinary developments? Let me know in the comments and if you want to understand how these self-employed agents already control entire markets and ultimately entire companies today, then watch my video on the agent economy next time. We'll see you there again. Take care, your Leo.