The Machine Customer

If you tried to pinpoint the day commerce moved online, one good answer would be August 11, 1994: the day NetMarket sold that Sting CD, ten dollars plus shipping, encrypted through a browser. Why?

Before then, buying was physical. Stores, catalogs, phone orders. The mall was the recommendation engine.

After then, the page became the store. The storefront was rebuilt as a visual artifact: photography, copy, color, persuasion. Amazon made it convenient. eBay made it a market. Shopify made it universal. Instagram made it beautiful. Every pixel argued with a human eye.

A science formed around that eye.

Heatmaps were drawn. Eye-tracking studies were commissioned. Entire careers were built on knowing where a thumb would pause.

Search rankings decided who was found. Conversion rates decided who was bought from. Thirty years of craft, all aimed at the same reader: a person with a credit card and a wandering attention span.

Every optimization ever made assumed the same pair of eyes.

In the beginning, nobody knew how to sell online. Maybe not literally nobody, but the number was small and they all worked within ten miles of a server room. Selling on the internet was low status, assumed to be for mail-order weirdos. There was no track to follow.

That soon changed. Shoppers arrived just as the technology hit its stride. Google made search the front door. Shopify put a store behind it. And like railroad tracks built across America in the 1800s, economic tracks were laid in the 2000s to support the demand for attention. Ad networks to buy it. Landing pages to convert it. Email flows to recover it. More each year.

All with good intentions.

In time, what started as an earnest effort to help merchants avoid mistakes grew into something different, and dramatically lowered the perceived risk of selling online. One playbook became hundreds, each promising a more predictable path to the cart. Playbooks for SEO. Playbooks for paid. Longer flows. Bigger popups. Wider funnels. Specialized playbooks for DTC. For marketplaces. For DTC-enabled marketplaces. Yes, we definitely have that funnel template.

Dependent on the tracks for their livelihood, agencies and gurus manifested more. Urgency timers. Exit-intent popups. Social-proof widgets whispering that someone in Ohio just bought this. Breakout brands, it was believed, would lay new tracks for others to follow. For everyone selling, following a proven track felt risk-free.

Conversion rate optimization told you to take your big idea and A/B test it small. Persuasion is a science. A formula. Math, if you will. If it is a disaster, you iterate. There is no way to fail, because you are always two percent from a better funnel. And nobody likes the possibility of failure, least of all anyone who has signed up for a track.

The message was clear: just get on The Conversion Track and pick a channel. You will eventually convert, or you will convert somewhere.

Like a river with a strong current, it was believed, the power of the track, not the individual merchant or product, could carry a store forward.

And for a long moment it looked that way.

By the early 2020s, one might argue that the track had grown into a sprawling apparatus that served itself more than the merchants at the heart of it. Selling online had become a prestigious career with set milestones and rewards, like being a banker, consultant, or tech executive.

It was deja vu all over again. Products that neatly fit the track were courted by aggregators at the finest demo days. Ad budgets climbed, but not as fast as the cottage industry of tools that promised to lower them. If accepted, you would join retreats with nine-figure founders who exuded a bravado previously reserved for celebrities. More than a few merchants were taken on a magic carpet ride to spend more on acquisition, with little concern for if they should.

Then the cracks showed. Acquisition costs rose and margins compressed. Attribution went from abundant to scarce. What once seemed a predictable path was revealed as more idiosyncratic. And underneath it all, the reader began to change. First slowly: comparison engines, review aggregators, buy buttons inside feeds. Then all at once: agents that shop.

An agent does not see your funnel. It is not retargeted. It is not rushed by a countdown timer. It is not moved by your brand film. It reads structure: feeds, specs, APIs, policies, price.

Stores optimized for persuasion are illegible to the buyer that matters next. The unease is palpable. Some have filled the void with more popups. Others swear the answer is more brand. Maybe it is one channel and done. Maybe you just need to be on TikTok. Or is it Amazon.

Or maybe the funnel did not break. Maybe the reader changed.

Today, something more powerful than the track has emerged, driven by a customer who was never on it. Procurement agents, shopping agents, assistants carrying budgets and instructions. Instead of entering your funnel and letting it take them somewhere, they arrive knowing exactly where they are going.

They do not browse. They resolve.

They compare every option, never tire, and never impulse-buy. They read the schema, not the hero image. They check the return policy before the price. They do not abandon carts; they abandon stores that cannot be read.

They choose on legibility and terms.

The Machine Customer arrived one protocol at a time.

This is a fundamentally different way to buy. It also stands in stark contrast to a sea of vague missions from commerce giants today. Take one whose mission is to revolutionize commerce globally. Is this a marketplace, a payments company, a website builder, or all of them? How will a merchant know if that mission has succeeded? Perhaps that is the point. Big platforms have become too bland to fail. This is the merchant’s opportunity.

The largest platforms in the world are already racing to wire agents into checkout, and they will wire in the stores that are easiest to read first. The rallying cry of the funnel era was: if we do not capture the click now, somebody else will. In the agentic era the truth is different: if your store cannot be read, nobody will ever know you were there.

Bright lines around legibility serve an important purpose: they drive focus. A merchant who knows exactly what an agent can and cannot read knows exactly what to fix next. The specificity leaves no room for the problem to be rebranded as a branding problem.

When the customer is a machine, readiness is the brand. A store is either agent-usable or it is invisible. There is no page two.

What these agents leave in their wake may look like a funnel. And surely many will try to sell you one. But they will all be disappointed. It is not a funnel. It is a specification.

Today, cynicism runs rampant. We are told daily that AI will flood the internet with slop, that agents will strip-mine small merchants, that commerce will become a toll road run by three platforms. On an industry level, we are exiting an exuberant period that has left many with a natural bend toward skepticism over optimism.

Against this backdrop, fear has no shot. This transition is not a threat to merchants; it is a leveling. Legibility is cheaper than persuasion, and a two-person store can out-read a giant. Open specifications keep the agentic web from becoming a toll road. The stores that move first will not just be found. They will be the default answer.

That is why we publish in public: the spec, the harness, the scores. Collectively, these artifacts take our gaze from the glidepath of the present and point it toward the catalysts of the near future. The agent-ready web should be measured, not marketed.

Indeed, building what matters can only come from the strength of a specific mission, not the strength of the track. For the most significant transitions, you measure the possible, and then you build it.