When Nobody Visits

Adobe measured AI driven traffic to US retail sites growing 4,700 percent year over year.

Not 47 percent. Not 470. Four thousand seven hundred.

IBM's Institute for Business Value found that 45 percent of consumers already use AI for at least part of the buying journey. McKinsey put the agentic commerce opportunity somewhere between three and five trillion dollars by 2030, which is the kind of number you should treat as a direction rather than a forecast.

Here's the version of that which actually affects you, and it's a question rather than a statistic.

When a machine evaluates your business on behalf of a customer who never loads your website, what does it see?

Most owners have never asked. And the answer, for the overwhelming majority of small and mid sized businesses, is "almost nothing, and some of what it sees is wrong."

Quick orientation, because the acronyms are genuinely confusing and you need about four minutes of context to make good decisions.

OpenAI and Stripe built the Agentic Commerce Protocol, live since September 2025. Google announced the Universal Commerce Protocol at NRF in January 2026 with Shopify, Etsy, Wayfair and Target in the coalition, plus more than twenty payment and retail partners. Google also has the Agent Payments Protocol, which records a shopper's instruction as a signed mandate so a merchant can verify the agent actually had authority before taking money. Anthropic's Model Context Protocol, now under the Linux Foundation, handles the connection between models and live data like catalogs and inventory.

On the payments side: Mastercard went live with agentic transactions in Asia Pacific in early 2026. Visa launched its Trusted Agent Protocol commercially after piloting with more than a hundred partners. American Express shipped a developer kit and purchase protection for registered agent purchases. All three US card networks now support agent initiated payments.

One correction worth having, because a lot of content still gets this wrong: OpenAI deprecated Instant Checkout in March 2026. The model shifted from "buy inside the chat" to "discover and compare inside the chat, transact on the merchant's site." That's a meaningfully better outcome for you, because you keep the customer relationship and the checkout. Discovery is what moved, not the transaction.

You don't need to implement any of this personally. If you're on Shopify, Agentic Storefronts abstracts most of it into a toggle. What you need is to understand what the plumbing is asking for, because the plumbing is asking for things you probably don't have.

An agent evaluating you has no eyes. It can't be impressed. It doesn't perceive design, or a well shot photo, or the fact that your shop has been on that corner since 1987 and everybody in town trusts you.

It reads structured facts and it compares.

It needs complete attributes. Not marketing copy. Fields. Dimensions, materials, compatibility, capacity, lead time, warranty, what's in the box. A missing field isn't neutral, it's disqualifying. If a buyer asks for something that fits a specific space and your listing has no dimensions, you don't rank lower, you're simply not in the comparison. PwC's framing on this is that agent ready commerce depends on clean data and processes agents can reliably act on, and clean is doing heavy lifting in that sentence.

It needs live availability and price. An agent asked for something by Friday has to know whether you can actually do Friday. If your inventory updates every thirty minutes and your competitor's is real time, you will lose comparisons you should have won, and nobody will ever tell you it happened.

It needs an unambiguous answer to who this is for. Agents match constraints. "Premium quality and exceptional service" matches nothing because it's not a constraint. "Fits vehicles from 2018 onward, ships in two days, rated for commercial use" matches a query. The vaguer your positioning, the more invisible you are to a matching system, which is a funny reversal because vague positioning also loses to humans, it just loses more slowly.

It needs a transaction path it can complete. A booking system a machine can reach, a catalog it can query, a quote flow it can start. If the only way to buy from you is to call during business hours, you've opted out.

Run your own business through those four. Be honest. Most people score one out of four and the one is price.

FROM THE AI NEWSROOM

The exact systems behind everything in this issue. The audit sheets, the routing logic, the templates and the review cadences, built out step by step so you can copy them straight into your own stack. One time, forty seven dollars.

Get the Blueprint for $47

.....

The retail framing hides the point and I want to pull it back out, because the service business version is bigger and nobody's talking about it.

Agents book appointments. They compare providers. They qualify vendors. They screen candidates. Somebody asking an assistant to find a commercial plumber who handles restaurant grease traps, is licensed in their county, and can come out this week is running exactly the same matching process as somebody shopping for a dining table.

So the service version of those four requirements:

Your service catalog as structured facts. Not a paragraph on an About page. A list. What you do, what you don't do, what each thing costs or what band it falls in, how long it takes, what you need from the client to start.

Your coverage as data. Geography, hours, response time, capacity. "We serve the greater metro area" is not coverage data. A list of counties is.

Your qualifications as facts. Licenses with numbers. Certifications with issuers. Insurance with limits. Years in business with a founding date. These are exactly the fields a screening agent filters on, and they're exactly the fields most service businesses bury in an image on a page nobody reads.

A booking path that doesn't require a human. This is the one that hurts. If you don't have one, Go High Level will give you a bookable calendar, a service menu and a quote flow in an afternoon, and that is the cheapest version of this whole problem. Plenty of excellent service businesses run entirely on phone calls, and that's been a perfectly good business for forty years. It stops being one the moment the buyer delegates the search.

None of that requires new technology. All of it requires somebody sitting down for an afternoon and writing facts in a structured form, which is precisely why it doesn't get done.

Let's be clear eyed about both, because the honest accounting is more useful than the hype in either direction.

What you lose. The browse. Somebody wandering your site, seeing three things they didn't come for, buying two. Impulse. Merchandising. The brand impression from your photography and your writing and the care you put into the page. All of that gets compressed into a row in a comparison table, and the compression is brutal.

You also lose some of the ability to sell against a spec. If you're better but harder to describe, the machine's not going to catch that, at least not yet.

What you keep. Everything after the first transaction. The agent finds you once. What happens next is entirely yours: the quality, the follow through, the relationship, whether they ever need to search again. Repeat business has never been mediated by discovery and it still isn't.

You keep the things that can't be listed. Judgment. Taste. Knowing which of two options the customer actually wants when they asked for the wrong one. Being the person somebody calls when it's gone wrong at eleven at night.

And you keep proof. Specific, verifiable, attributable proof outperforms adjectives with machines even harder than it does with people, because a machine can check some of it. Three case studies with real numbers beat six paragraphs about excellence, and that's true in both channels at once, which makes it a rare piece of work that pays twice.

Here's the strategic read, and it's not what most people expect.

Agentic discovery is brutally efficient at commodity matching. If what you sell is genuinely comparable to what four other people sell, this is bad news, because the machine will do exactly what it's designed to do and pick on price and speed. There's no relationship to lean on, no salesmanship, no charm. You will be optimized against.

But the same efficiency cuts the other way for anybody with real specificity. The reason obscure specialists have always struggled isn't that nobody wanted them. It's that nobody could find them. Discovery was expensive, so buyers settled for whoever was nearby and adequate. An agent that can evaluate two hundred providers against fourteen constraints doesn't settle. It finds the one that actually fits.

If you're the only shop within three hundred miles that does a specific thing for a specific kind of customer, agentic discovery is the best thing that's ever happened to you, provided the facts that make you specific exist somewhere a machine can read.

Which means the strategic response isn't defensive. It's to get narrower and more explicit at the same time. Say exactly who you're for and exactly what you do, in structured form. The broad, hedged, "we serve a wide range of clients" positioning that felt safe for twenty years is now the actively dangerous position, because it matches nothing and competes on price by default.

You've got a quarter left in the year. Here's how I'd spend it.

October: the audit. Two afternoons. Pull up your own listings, your own site, your own booking flow, and fill in every structured field that's empty. Dimensions, specs, coverage, hours, credentials, lead times, prices or bands. Every blank is a query you can't win. Don't redesign anything. Just fill fields.

While you're there, run the visibility check. Ask three AI tools the questions your buyers would ask. Write down what comes back. That's your baseline and you'll want it in January.

November: the answer file. One document that answers every real question a buyer asks, in your actual voice, with real specifics. The objections. The edge cases. The "will this work for me if" scenarios. The honest comparison against the obvious alternative, including where the alternative wins.

This document does four jobs at once. It feeds a Sponsored Agent if you run one. It feeds your own site chat. It feeds your support. And it's the best sales training asset you'll ever make. If you do nothing else this quarter, do this one.

December: the path. Make sure a machine can start a transaction with you. Product catalog queryable, or booking reachable, or a quote flow that doesn't require a phone call at the first step. On Shopify that's a toggle. Off Shopify it might be a project, though for most service businesses it is a Go High Level calendar and a properly structured service list rather than anything that deserves the word project. Either way, know which one it is before January, because "we'll look at it next year" is how three years go by.

I'm going to be straight about the uncertainty here, because the confident takes on this topic are mostly selling something.

Nobody knows how fast this goes. Adobe's 4,700 percent is off a tiny base. Forty five percent of consumers using AI somewhere in the journey is not forty five percent of purchases being agent completed, and the gap between those two numbers is enormous. It's entirely possible this takes six years instead of two, and it's entirely possible the protocols consolidate in ways that make half of today's implementation work irrelevant.

Here's why I'd do the work anyway.

Every single item on that ninety day list makes your business better in a world where none of this happens. Complete product data helps human buyers. Structured service information helps human buyers. An answer file helps your team and your customers today. A booking path you can reach without a phone call has been a conversion win for a decade.

That's the tell for good strategic work under uncertainty. When the preparation pays off regardless of whether the prediction lands, you're not making a bet. You're just doing the thing you should have done anyway, and getting optionality for free.

The businesses that get hurt here won't be the ones that guessed wrong about timing. They'll be the ones that used the uncertainty as a reason to do nothing, and then discovered that the window for catching up was shorter than the window for getting ready.

FROM THE AI NEWSROOM

For operators who would rather build it with someone than build it alone. The full stack, installed alongside you, with the decisions made in the room instead of guessed at. Ninety seven dollars.

Join the Accelerator for $97

.....

Open a private browser window so your history doesn't color it. Go to ChatGPT, Gemini and Perplexity.

In each one, ask the exact question your best customer would have asked before they found you. Not a question about your company. The question somebody asks when they have the problem and don't yet know you exist.

Read all three answers carefully.

If you're not in them, that's your quarter. If you are in them but what's said is thin or wrong, that's your quarter too, and it's the easier version.

Either way you now know something about your business that you didn't know on Saturday, and it cost you four minutes.

Jordan

The AI Newsroom | Practical AI for people with a business to run.