Stop Building for Clicks. Build What AI Agents Will Pay For.

For 30 years, the internet has been built around one basic objective:

Get a human to click.

That simple action created enormous industries.

Google organized information and sent people to websites. Publishers gave search engines access to their content because traffic had value. Businesses built landing pages, bought ads, optimized for SEO, collected leads, and tried to convert visitors into customers.

The bargain worked.

I think that bargain is starting to break.

My bet is that we are moving from the Human Web to the Agent Economy—and the founders who understand that shift early are going to build a very different kind of company.

The Human Web was built around pages people visit.

The Agent Economy will increasingly be built around resources machines can use.

That sounds like a small distinction.

It isn’t.

It changes what gets monetized.

The Click Is Becoming the Request

Think about how buying something works today.

You search.

You click.

You visit several websites.

You compare.

You fill out a form.

You make a phone call.

You check availability.

You eventually buy something.

Now give most of that work to an AI agent.

The agent doesn’t need to admire your homepage.

It needs your price.

Your inventory.

Your availability.

Your credentials.

Your specifications.

Your reputation.

Your rules.

Your API.

Maybe it needs a specialized capability your company provides.

The agent gathers those resources, compares the options, makes a recommendation and eventually may execute the transaction.

The human might never visit your website.

That’s fantastic for the customer.

But it creates an uncomfortable question for the owner of the information:

Why should I give valuable data to a machine for free if that machine no longer sends me the customer?

That is where things get interesting.

The old economic unit was the click.

I think the new economic unit may increasingly become the request.

Agent → request → resource → decision → transaction.

For founders, that changes the question from:

How do I attract more traffic?

to:

What do I own that an agent needs to get a job done?

HTTP 402 Is the Signal. x402 Is the Payment Rail.

There is a piece of internet plumbing worth watching closely.

HTTP has had a status code called 402 Payment Required for decades. The official HTTP specification still lists it as reserved for future use.

Now that old placeholder is becoming relevant.

But there is an important distinction.

HTTP 402 is the signal. x402 is an emerging payment protocol built around that idea.

With x402, a machine can request a resource, receive payment requirements, authorize payment and continue the transaction programmatically.

Cloudflare is already moving in this direction. Its Pay Per Crawl system allows publishers to set prices for AI crawler access, with crawlers receiving an HTTP 402 response containing pricing when payment is required.

In July 2026, Cloudflare also announced its Monetization Gateway, designed to let customers charge for resources including webpages, datasets, APIs and MCP tools using x402. The product is currently being offered through an early-access path rather than as universal infrastructure across the web.

And this isn’t just one company’s experiment. The x402 Foundation announced its operational launch under Linux Foundation governance in July 2026, an attempt to develop x402 as open infrastructure for internet-native payments by agents, applications and APIs.

That doesn’t mean x402 has already won.

It means the infrastructure for machine-to-machine commerce is being built right now.

Founders should pay attention.

Stop Building Another Chatbot

Everyone wants to build the agent.

I am increasingly interested in what all those agents will need underneath them.

The durable businesses may be less glamorous:

clean data, verified information, proprietary workflows, identity, reputation, permissions, specialized APIs and executable expertise.

I would rather own a resource 10,000 agents need than build chatbot number 10,001.

Cloudflare’s AI Index points in this direction too: websites can expose machine-oriented resources such as structured search APIs, MCP access and bulk-data interfaces rather than relying only on human-readable webpages.

That is a preview of a different internet.

Not just websites.

Machine-readable businesses.

Build a Niche Data Refinery

One of the businesses I would seriously consider starting today is almost boring.

Clean up messy information.

AI models are powerful.

Real-world business data is a disaster.

It lives in spreadsheets, PDFs, emails, accounting systems, old websites, text messages, CRMs and people’s heads.

Humans have spent decades working around that mess.

Agents won’t.

An agent cannot reliably execute a transaction if it cannot determine what is true.

So don’t try to organize all healthcare.

Start with orthopedic practices in Atlanta.

Don’t organize construction.

Start with commercial roofing contractors in Florida.

Don’t organize hospitality.

Start with independent vacation rental operators along the Gulf Coast.

Take 100 businesses.

Determine what an agent needs to understand them:

services, prices, availability, credentials, locations, inventory, policies, scheduling rules and verified reputation.

Then clean it.

Structure it.

Verify it.

Keep it current.

Make it callable.

At first, businesses may pay you to clean their data.

Good.

Get paid to build the asset.

If you eventually become the most reliable machine-readable source for that niche, you are no longer just running a consulting company.

You own infrastructure.

And infrastructure can be queried.

Sell Agent Readiness Before Everyone Knows They Need It

Here is another business I would start manually.

Ask the leading AI systems basic questions about a company:

What does it do?

Who should hire it?

Where does it operate?

What makes it different?

Is it reputable?

Then compare the answers with reality.

That gap is the product.

Call it an Agent Readiness Audit.

For 20 years companies asked:

Where do we rank on Google?

They are increasingly going to ask:

What does the AI believe about us?

If agents influence recommendations, purchasing and vendor selection, being misunderstood by AI isn’t just a branding problem.

It becomes a revenue problem.

The company that agents cannot correctly understand may never make the shortlist.

Turn Expert Knowledge Into Capability

There is another enormous asset sitting around doing almost nothing:

expert knowledge.

Lawyers have decades of briefs.

Engineers have processes.

Doctors have protocols.

Investors have models.

Operators have systems they developed over entire careers.

Most of it sits inside documents, videos, courses and archives.

Don’t build another chatbot that tells me what the expert said.

Build something that does the job the way the expert does it.

Don’t give me a procurement expert’s library.

Give me the capability that evaluates the vendors.

Don’t give me the pricing expert’s framework.

Give me the tool that runs the analysis.

Don’t make me search 20 years of legal briefs.

Build the capability that applies the methodology to the next problem.

Information tells you what to do.

Capability does it.

And eventually another agent can hire that capability.

Start Manual. That’s the Advantage.

Founders are often embarrassed by manual work.

Right now, I think manual work is an advantage.

Do the first ten Agent Readiness Audits yourself.

Clean the first thousand records manually.

Run the expert workflow beside the expert.

Find the edge cases.

Find where human judgment matters.

Most importantly, find out what somebody will actually pay for.

You cannot discover that on a whiteboard.

Maybe a verified record is worth $3.

Maybe an analysis is worth $50.

Maybe a completed transaction is worth 2%.

Do the work and find out.

Then follow a simple progression:

Service → Software → Infrastructure.

Service teaches you the problem.

Software automates what repeats.

Infrastructure makes the capability available to everyone else—including agents.

Then you put a price on the request.

Build the Door

I don’t believe websites disappear.

People will still browse. Brands will still matter. Search will still matter.

But another economic layer is being built on top of the web.

And increasingly, machines acting on behalf of humans will participate in it.

That creates a founder opportunity that is much bigger than building another AI interface.

Ask yourself:

What do I own that an agent will need to accomplish something valuable?

Data?

Trust?

Verification?

Expertise?

Access?

A workflow?

An API?

A transaction?

Find it.

Structure it.

Verify it.

Make it callable.

Make it indispensable.

Then put a door in front of it.

My bet is that the internet is moving from pages humans visit to resources agents use.

The Human Web monetized attention.

The Agent Economy will increasingly monetize utility.

And a lot of the most valuable doors have not been built yet.