The Agent Is Now the Customer

Marketing is being rebuilt around a new decision-maker.

For the last twenty years, marketing has been built around a human who searches.

They type a query.

They click a result.

They visit a website.

They compare options.

They make a decision.

The entire marketing stack evolved around that behaviour.

SEO optimised the page.

Paid search bought the click.

Content marketing captured attention.

Brand built preference.

Analytics measured the journey.

The objective was fundamentally the same:

Get the human to you.

That architecture is now changing.

Because increasingly, the human isn’t doing the searching.

The agent is.

The agent is now the customer

ProFound’s recent announcement captures the shift remarkably well.

Its CEO describes a world where people stop clicking blue links and start asking AI for the answer.

But there is an even bigger step beyond that.

The agent doesn’t just retrieve information.

It forms an opinion.

It gathers sources.

It interprets them.

It builds context.

It evaluates alternatives.

It selects.

And increasingly, it can act.

That means the fundamental unit of marketing changes.

It is no longer:

Query → Click → Website → Conversion

It becomes:

Intent → Context → Understanding → Trust → Selection → Execution

The website is no longer necessarily the destination.

It can become just another piece of evidence in a much larger decision system.

This changes what “visibility” means

The obvious response to AI Search has been:

“How do I get mentioned?”

That’s useful.

But it is already becoming too narrow.

Because a mention isn’t the outcome.

The mention is the output.

The real asset is the system that caused the mention.

Why did the model understand the company?

Why did it consider it relevant?

Why did it trust the information?

Why did it select it?

Why did it reuse that selection in another context?

And eventually:

Why did it act?

This is a fundamentally different problem from traditional SEO.

SEO asks:

How do I rank?

AI-mediated discovery asks:

How do I become the answer?

And agentic execution asks:

How do I become the action?

From ranking systems to decision systems

Search engines largely organise information.

AI systems increasingly organise decisions.

That distinction matters.

A search engine can return ten results and leave the user to work out what to do.

An agent is being asked to resolve uncertainty.

“Find me somewhere for our leadership offsite.”

“Which software should we use?”

“Book the best option.”

“Compare these suppliers.”

“Draft the proposal.”

“Buy it.”

The further agents move toward execution, the less valuable raw visibility becomes.

Because the agent doesn’t need ten thousand possibilities.

It needs a confident resolution.

This is why I think the emerging AI economy will be less about attention and increasingly about certainty.

The new marketing stack

We can already see the beginnings of the stack forming.

1. Context

The system needs to understand the brand.

Not just its homepage.

Its products.

Customers.

Positioning.

Proof.

Capabilities.

Constraints.

History.

And the relationships between them.

2. Coherence

That information needs to agree.

Across the website.

Content.

Reviews.

Third-party sources.

Structured data.

Social channels.

Customer experiences.

Partner networks.

The agent is effectively constructing a model of the company.

Contradictory signals create uncertainty.

Coherent signals reduce it.

3. Trust

The system then needs reasons to believe what it has understood.

This is where trust networks become important.

No single piece of content necessarily determines the answer.

The answer emerges from converging evidence.

4. Selection

Eventually the system has to choose.

This is the layer that traditional marketing has largely ignored because humans historically performed it.

AI changes that.

Selection becomes computational.

And when selection becomes computational, the architecture surrounding selection becomes incredibly valuable.

5. Execution

And finally, the agent acts.

The recommendation becomes a booking.

The shortlist becomes a purchase.

The research becomes a procurement decision.

The answer becomes an action.

At that point, marketing and operations begin to converge.

The moat moves

This is why I think the current obsession with model intelligence is only part of the story.

Models are becoming increasingly capable.

Intelligence is becoming cheaper.

Smaller models are becoming remarkably capable.

Inference costs are falling.

Open models are accelerating the optimisation loop.

So the scarce resource gradually moves elsewhere.

From intelligence to trusted resolution.

The model doesn’t necessarily need more intelligence to answer the question.

It needs better information about what to believe and what to do.

That creates an entirely different competitive landscape.

The moat isn’t necessarily the intelligence.

It can be the relationship between intelligence, context, trust and action.

The end of the funnel

There is another consequence.

The traditional marketing funnel assumes that humans move through stages:

Awareness.

Consideration.

Intent.

Conversion.

But agents don’t necessarily behave like that.

They can collapse the funnel.

A user can say:

“I need a three-day leadership offsite near London for 25 people in October.”

The agent can immediately:

• interpret the requirements

• discover options

• assess suitability

• compare them

• check availability

• evaluate evidence

• recommend one

• and potentially book it.

The funnel hasn’t simply become shorter.

The funnel is being replaced by a pathway.

Funnels optimise for attention.

Pathways optimise for certainty.

And this is where the real opportunity begins

The first phase of AI marketing has been about being discoverable.

The second phase is about being resolvable.

Can the system understand exactly what you are?

Can it distinguish you from alternatives?

Can it establish that you’re credible?

Can it confidently map a user’s intent to your proposition?

Can that resolution be reused?

And eventually:

Can an agent execute the decision without needing the human to intervene at every step?

That is a very different discipline.

It isn’t SEO.

It isn’t traditional content marketing.

It isn’t simply GEO.

It is something closer to decision-system engineering.

The agent doesn’t want your content

This may be the most important distinction.

Humans consume content.

Agents consume context.

They don’t necessarily care that you published another 1,500-word article.

They care whether the information they need to make a reliable decision exists, is current, is coherent, and can be trusted.

That changes the question marketers should be asking.

Not:

“How much content are we producing?”

But:

“What does the decision system need to know about us to confidently select us?”

That is a much harder question.

And potentially a much more valuable one.

The next marketing platform

ProFound is betting on the application layer.

That makes sense.

But I think the bigger architectural shift is still ahead.

Because once agents become the interface between intent and action, every company is going to need to understand its position inside these decision systems.

The winners won’t simply have the best campaigns.

They’ll have the clearest machine-readable identity.

The strongest trust networks.

The most coherent context.

And the highest probability of being selected when an agent encounters relevant intent.

In other words:

Brand building doesn’t disappear.

It becomes infrastructure.

And the ultimate metric may no longer be:

“How many people saw us?”

It may become:

“How often did the decision system resolve the right intent to us?”

That is the beginning of the Selection Layer.

And I suspect it will become one of the most valuable layers in the AI economy.