End of the Click, Rise of the Agent

For decades, we have built brands for a simple moment: the moment someone chooses. A supermarket shelf. A scroll through a feed. A search result. Each of these environments shared the same assumption – that a human would do the choosing. 

That assumption is starting to break. 

We are entering a new phase of decision-making. Not searching, not scrolling, not even being guided towards what we might like next. Instead, delegating. 

From the Oracle of Delphi 3,000 years ago to the rise of Google search, humans have always looked for ways to outsource choice. We didn’t go for information. We went for answers. But generative AI has created something new. For the first time, we are not just asking for help. We are conversing with agents, delegating decisions, and increasingly asking them to act on our behalf. 

This shift introduces a new actor into the system: the machine. Not as a tool, but as an intermediary. An AI agent that doesn’t just retrieve options, but filters, ranks, summarises and increasingly decides. In doing so, it reshapes how brands are discovered and chosen.

From the Traffic Tax to the Choice Blackout 

Today, the machine is reshaping search, no longer surfacing ten blue links for us to select from. They are deciding what is worth surfacing – and the number of options is shrinking. The latest generation of ChatGPT surfaces 37 per cent fewer brands than before, a trend that is set to continue. If you are not structured, understood and relevant to those systems, you are punished with: less visibility, fewer appearances and declining traffic back to the site. You will pay the Traffic Tax. 

But this is only the beginning. 

This is not just about losing position in search. Increasingly, there will be no search moment at all. As agents take on more of the journey – researching, comparing, recommending and even purchasing – the need to visit a website begins to disappear. There is no results page, no list of links and no opportunity to compete for attention. There is simply a decision. 

And increasingly, that decision is not made by us. It is made between systems. Your agent negotiating directly with a brand’s agent – comparing options, f iltering choices and selecting outcomes – all without you ever seeing the alternatives. This is the Choice Blackout. A world where brands do not gradually lose traffic; they are either selected or invisible. Will the agent choose your brand, or will you never even be considered?

The two-audience challenge 

This shift introduces a new reality for brands. 

For the first time, brands are no longer built for one audience. They are built for two: the human – messy, emotional and social – and the machine – logical, structured and statistical. 

Humans learn through culture, repetition and shared experience. Machines learn through data, signals and structure. Increasingly, it is the interaction between the two that determines who is chosen. 

That creates a fundamental tension. What makes a brand strong for one audience may be invisible – or even detrimental – to the other. The expressive, emotional signals that resonate with people can be difficult for machines to interpret, while the structured clarity that machines require can feel reductive to humans. 

How can we balance the two? 

Scraps, straws… and silicon 

Jeremy Bullmore once said that people build brands as birds build nests, from scraps and straws they chance upon. For decades, those scraps were designed around humans: ads, conversations, culture and memory. A thousand small signals, loosely connected, gradually building into something meaningful over time. 

But with two audiences, the requirement has changed. 

Those scraps are no longer collected only by people. They are collected, interpreted and weighted by machines. Reviews become data points, conversations become structured signals, and content becomes training material. The nest is no longer built solely in human minds. It is built in machines too.

The new gravity: brand embeddings 

This shift introduces a new layer to how brands exist – not just in culture or memory, but inside the model itself. 

Large language models build meaning through embeddings: weighted relationships and gravity between words, concepts and ideas. Every piece of content, every review and every mention contributes to this structure, forming something powerful over time – a version of your brand inside the model. 

This version is not defined by what you say, but by the totality of what is known. It becomes a kind of machine-held understanding of your brand: a statistical picture of who you are, what you stand for and when you should be chosen.

 Ask a model to describe your brand as a person – how they look, how they behave, who they talk to – and it will generate a surprisingly coherent answer. Not because it “knows” your brand in the human sense, but because it has mapped the relationships around it. Example: Heinz Ketchup mirror test; to see brand embeddings in action, ask AI to describe and visualise your own brand. 

This alone can be revealing. If there is a gap between how you believe your brand is positioned and how the model represents it, it is rarely the model that is wrong. 

This is the machine equivalent of brand perception – a semantic web of meaning built from data, rather than memory. 

And it leads to a fundamental shift: from perception to gravity – the ability to be consistently surfaced, selected and recommended by the systems that now shape choice. 

Building for humans and machines 

As machines increasingly make decisions on our behalf, brands have never mattered more – and have never been harder to get right. Trying to be fluent in both human and machine languages creates new tensions in how brands are built, expressed and managed. But the fundamentals endure. Brands must still be talked about, available and distinct, but we must rewrite what each means for this new era.

Caroline Reynolds is the global chief performance officer for T&P