Everyone Has the Button

A client I consult for sells on a large online marketplace. I was describing a dedicated AI system for their product listings: one that knows their brand, their rules, their customers. They listened, then asked the most reasonable question in the world.

"Why would I use anything else? The software already has a button for that."

A far more experienced person in the same room dismissed the button in one line: "It's a lazy way to do marketing."

That's the instinct. I want to lay out the reasoning, because I don't think the question was a bad one. I think it was the right question in the wrong frame. They were asking which tool to use. The question that matters is what makes them different. The button can't answer that one. Nobody's button can.

The same good listing

Here is the problem with the button. Everyone has it. Every seller on that marketplace sees the same button, backed by the same model, fed the same kind of input: a product name and a few attributes. Press it and you get a good listing. Your competitor presses it and gets a good listing too. The same good listing.

When every competitor has the same tool, the tool stops being an advantage. It becomes the baseline.

I think of it like a jig in a woodshop. A jig is the guide you clamp to the work so every cut comes out the way you want it. You can buy one off the shelf, and it will make a clean cut. So will the same jig in every other shop. The shops people remember build their own, shaped around the work only they do.

Map it across. The AI model is the saw: a commodity now, and getting cheaper. The jig is everything you build around it: what the system knows about your brand, the rules it must never break, the checks that catch it when it makes things up.

And the button? The button is a jig built for the platform. Its job is to make millions of listings acceptable. Your job is to make one listing win. Those are different jobs.

Later, during an AI audit I ran for the same client, the question came back in different clothes. "Why don't we just hire someone internally?" Fair question, and the same reflex. Anyone can get access to AI now. That's exactly the point. Access is the commodity. The jig is not.

A small test

I wanted to know whether the convergence was real or just a tidy argument. So I ran a small test.

I made up five brands selling the same commodity product: a magnesium sleep supplement, same dose, bottles of 60 to 180 capsules. None of them said how many capsules to take a day. I simulated the button with one generic prompt, "write an optimized listing," identical for every brand and given only what a seller would type in. It's a stand-in, not any platform's actual feature. Then I built a small dedicated system: the same model, plus each brand's own knowledge (voice, customer, the words its customers use, what sets it apart), claim rules, and a separate compliance review. I ran both on two different model families.

The convergence was real. I counted three-word phrases that showed up nearly word for word in at least four of the five competing listings. With the button: 20 on one model family, 25 on the other. With the dedicated system: 6 and 11. "Highly bioavailable glycinate form." "Third party tested for purity." "You can trust." Five brands, starting to sound like one.

It invented proof

That part I expected. What surprised me was the rest.

Neither version wrote a disease claim. What the button did instead was invent proof. "The gold-standard choice recommended by health professionals worldwide." "Screened for 280+ banned substances." "Clinically relevant dose." Nobody gave it any of that. It also picked a dosage out of thin air: 8 of the 10 button listings promised a supply duration ("a generous 4-month supply," "a full 90-day supply") with no serving size anywhere in the input. The dedicated system did it 3 times out of 10. The standard supplement disclaimer appeared in none of the button listings and in 7 of the 10 dedicated ones.

The catch

Now the catch, and it matters.

The dedicated system invented things too. Less often, but it did. One of its listings promised "a four-month supply at one capsule twice daily" for a 120-capsule bottle. That's 60 days. The rules helped. They didn't solve it. Rules reduce invention. Only a real source of facts removes it. I learned that by building the thing, not by pressing a button.

The test also wasn't a fair fight, on purpose. The dedicated system got more input than the button. That's the whole argument, since the button can't know your business, but it should be said out loud. Twenty listings, one run each. It illustrates the point. It doesn't prove it.

When the button is enough

None of this makes the button bad. An off-the-shelf store builder is a great product, and the exact opposite of customization. For plenty of businesses, that's the right call. If your words aren't where you compete, press the button and spend your energy somewhere else. A dedicated system is worth building only where the text is the battleground and a wrong claim costs you something real.

But look at the moment we're in. Shifts in technology are exactly when the gap opens. Everyone gets the same new tool on the same day. Most people use it the way it came out of the box. A few build the jig. A year from now, you'll be able to tell which was which.

The button isn't the problem. Thinking the button is the strategy is.