The Catalog Is the Small Part
Ron Reynolds · 2026-11-20 · 6 min read
Read anything about agentic commerce this year and you'll see the same example. A shopper asks an AI assistant for a backpack: fits a 15-inch laptop, good for international travel, under $150. The agent searches structured product data, finds three that fit, and buys one.
The advice that follows is sound. Clean up your product data. Fill in the attributes. Keep price and stock accurate. Your catalog is becoming your storefront.
All true. But look at what got left out. Where the money actually goes
Americans spend about $20 trillion a year on themselves. Roughly a third of that buys goods: the things a retailer stocks and ships. The other two-thirds buys services: rent, health care, haircuts, plumbers, flights, hotel rooms, lessons, repairs, childcare, meals out, a truck for moving day.
Online retail, the slice nearly every agentic-commerce announcement is aimed at, comes to about $1.2 trillion. That's around six cents of every consumer dollar.
That count also leaves out business-to-business trade, which is larger than all consumer retail put together, and almost none of it fits in a product feed either.
So the industry is building the future of commerce around the part of commerce that was already the easiest to put online. Why everyone starts with the backpack
It isn't foolish. A backpack is easy for an agent to deal with. It has a SKU, a price, a photo, and a number in stock. It's the same backpack for everyone, today and tomorrow. You can describe it completely in one row of a table.
Now try the requests people actually make on a Tuesday: "A dog walker who can take a reactive dog, weekday afternoons, near me." "Someone to fix a leaking water heater before the weekend." "A van for Saturday, one way, under $200." "Ship a 40kg box to Lagos." "A beginner pottery class my daughter and I can do together in December."
None of these is a product. Each one is a question about whether a business can do something, for this person, at this place, at this time. The answer depends on a calendar, a service area, a weight limit, a licence, a policy about reactive dogs. It changes by the hour. It's different for the customer across town.
A catalog row can't hold that. A product page doesn't answer it either, which is why today's agents end up reading websites, guessing, and handing the person a list of phone numbers. What ecommerce is now
Today's ecommerce is a store you visit. A person goes to a website, scrolls, compares, and clicks buy. The first wave of agentic commerce automates the scrolling: the agent visits the store for you. It's a faster way through the same door.
That's real progress for retail, and retailers should do the work it asks for. But it keeps the old shape: a business publishes pages, a reader works out what's true from them. What it will be
The shape that replaces it starts from the question, not the store.
A person tells their agent what they need. The agent asks every business that might help, in one structured question, and each business answers with plain facts: what it does, where, when, how much, and what it won't do. The ones that can't help drop out on the facts, with a reason. The ones that can come back, side by side, and the agent books, orders, or buys from the one the person picks.
That works the same for a backpack and a dog walker. The backpack is just the case where the facts don't change much.
Three things have to be true for it to work:
Every kind of business needs a way to state its facts. Not only products. Hours, coverage area, capacity, limits, certifications, lead times. A shipper's weight limit is as much a fact as a backpack's size.
The agent needs one place to ask. A plumber shouldn't need a separate integration for every assistant, and the assistant shouldn't need to read a thousand websites to find three plumbers.
The answers have to be trustworthy. If paying for placement can change what the agent is told, the agent can't trust anything it hears, and neither can the person behind it. How we built ComOS for it
ComOS is built for the whole two-thirds, not only the catalog.
Every business on the network runs on platforms made for its kind of work. Retail is one of them. So are bookings, rentals, shipping, services, and more than a dozen others. A pottery studio gets a calendar with seats, not a product page pretending to be a class.
Every business publishes the same kind of fact sheet, whatever it sells, so an agent can compare a courier with a courier and a studio with a studio on the things that matter. Ask "who can ship 40kg to Lagos?" and anyone who can't is out, with the reason, before anyone talks about price.
One connection reaches all of it. An agent connects to ComOS once and can reach every business on the network.
And there's no way to pay your way into the answer. The order comes from the facts and how well they fit what was asked, nothing else. ComOS earns when a sale settles, so it only does well when the customer was matched with the right business. If you can't take a reactive dog, no amount of money says you can. The point for every business owner
If you sell products, the advice you've been reading this year is right. Do it.
If you sell your time, your skill, your space, your trucks, or your calendar, which is most businesses, don't conclude this wave isn't for you. It's coming for you next, and it will reward the same thing: the businesses that made it easy for an agent to know, for certain, what they can do.
The catalog was the first part of commerce to move online. Most of what's left isn't a catalog.
Put your business on ComOS → comos-federation.com/start Figures are rounded, for the United States, 2024–25. Consumer spending and its goods/services split: U.S. Bureau of Economic Analysis, personal consumption expenditures. Online retail sales: U.S. Census Bureau, quarterly retail e-commerce report.