Amazon sellers were already using AI before Amazon Seller Central caught up

About 90% of Amazon's independent sellers already use some form of outside AI to run parts of their business. Amazon shared that number on Wednesday at its seller conference in Seattle, alongside the announcement of a new plugin that lets sellers manage their stores from Anthropic's Claude and from Amazon's own Quick assistant, without ever logging into Amazon Seller Central.

Put those two facts side by side and the story changes shape: the plugin is new, but the behavior it serves is old news to the people who were already doing the work.

Mary Beth Westmoreland, Amazon's vice president of worldwide selling partner experience, described the goal plainly.

"Our vision was that they would never have to log into Seller Central. We would just bring it to them where they work."

Where sellers work, it turns out, had already moved. For years the Amazon Seller Central dashboard was the job itself for a huge number of small businesses: updating prices, checking stock, pulling sales reports, and fixing listings one screen at a time. People became remarkably skilled at it, building careers on knowing exactly where every setting lived. That skill started losing value well before Amazon gave its replacement a name, and it didn't wait for anyone's permission to start.

The skill started expiring before the product launched

This is what I mean by the half-life of skills. The time between learning something useful and watching it go obsolete keeps shrinking, and the clock rarely starts with a company announcement. It starts when the people closest to the work find a faster way. By the time a platform ships the official version, the old skill has usually been fading for a while.

The scale matters here. Independent sellers account for more than 60% of the units sold on Amazon worldwide, and in the second quarter their seller-services fees brought Amazon $46.8 billion, more than AWS earned over the same period, according to GeekWire. Behind those figures sit family businesses, garage operations, and founders doing the books after dinner. They are the ones who started handing their dashboard work to AI agents, one store at a time, without waiting for a launch event.

Inside companies, the same behavior goes by the name shadow AI, which is employee innovation happening in real time, under a less flattering label. Amazon's sellers aren't employees, but the pattern holds regardless: the people nearest the work adopted the tool first, and the platform eventually followed where they had already gone.

Amazon went toward its sellers

I spent part of my early career at Napster, and I watched the music industry meet a behavior it hadn't sanctioned the way industries usually do. It ignored the new habit, shamed the people using it, and pushed for strict rules. Listeners kept going anyway, and the industry spent years catching up to what its customers had already decided.

Amazon made the opposite choice, and it deserves credit for that. It measured what its sellers were doing, then built an official door where they were already standing. Setup takes about 60 seconds with no coding, and the plugin is in beta for the U.S. marketplace now, with more AI partners planned.

It echoes a story I wrote about recently, when Amazon shut down Mechanical Turk and it became clear that many of the workers there had been using AI on their tasks long before the platform closed. The same pattern shows up inside the AI companies themselves: Anthropic has disclosed that Claude now leads 26% of its own internal AI research and development. I follow more of these cases in my Agentic AI coverage, and they keep teaching the same lesson.

Platforms usually name the change only after the people doing the work have already started it.

Every company has its own Seller Central

Most enterprises run on some internal system that people spent years learning: the procurement portal, the CRM, the reporting tool that closes the quarter. If Amazon's number comes anywhere close to your own company's, a large share of your people are already doing parts of that work with AI tools you never handed them. That can feel alarming. I'd treat it as good news, because it tells you where the real workflow lives right now.

I asked versions of this question at Cisco more times than I'd like to admit, and the answer usually surprised us. So here are two questions I'm taking into my own meetings; borrow them. Which of our systems are our people already routing around with AI, and what are they doing with the hours that come back? And when we build the official version, will it meet them where they already work, the way Amazon's plugin does, or will it ask them to come back to a screen they've outgrown?

If you're the person who knows the dashboard better than anyone, that knowledge still counts. The judgment behind the clicks, like knowing when to cut a price or which listing is worth fixing, is exactly what an agent needs a person to check. I call that working Above the Algorithm. Sort your own job into the clicking and the knowing, and put your learning hours into the knowing.

Relearning happens one storefront at a time

Multiply that choice across millions of small businesses and you get a picture of an economy relearning in real time. The families who run those stores feel the half-life of skills long before any survey records it, and they have been adjusting without a manual. What they need from the platforms they depend on is what Amazon just offered: tools that follow them instead of fighting them.

Amazon found out where its sellers were by counting. Most leadership teams haven't counted yet. If someone asked your people tomorrow how much of their work already runs through outside AI, would their answer surprise you? Take that question into your next leadership meeting and see what comes back. And if you want to compare notes, I'm easy to find.

Here is what makes Alex a credible voice on this topic: he shaped Cisco's $1.1B innovation portfolio, where the hardest part of the work was getting official processes to catch up with ideas that people at ground level had already put to use.

Bring Alex to your next leadership meeting to talk about what your people are already doing with AI →