Amazon’s New AI Agent for Sellers: What It Can (and Can’t) Do

Give it real credit. It can tell you how much stock you have left relative to your actual sales pace, not a rough guess. If a listing gets flagged for a suppression or a buyability problem, it walks you through the fix with a preview before anything submits. If sales dropped last week, it pulls your own traffic and sales data and tells you why, instead of you digging through five reports to find out.

If you’re newer and don’t yet know your way around Seller Central’s Issues tab, this saves real time. It’s Tier-1 work sellers currently pay a VA to do.

If you’re running one account with a handful of low-stakes SKUs, none of this matters much. You ask a question, you get an answer, you move on.

It starts to matter once your catalog’s big enough that no one person tracks every listing, or you’re running more than one account. Amazon’s agent is scoped to whichever account you connect — no view across five accounts at once, no one place a team can see what’s flagged across every marketplace. That’s five separate conversations, not one operation. We found the same shape of gap when Amazon’s Ads MCP launched earlier this year.

And inside a single account, the bigger the catalog, the more slips through. Buy Box competition and hijackers never show up — nothing in the toolkit watches for either. Neither do the quieter things: an FBA fee that quietly went up, a listing that lost its Amazon’s Choice badge, a category reclassification, a parent-product change that reshuffles which reviews and images a listing shares. None of these are “issues” to Amazon’s own system. They’re just things that happen as a catalog grows, and none of them get flagged, because nothing broke.

Depends less on how many accounts you run than how much is riding on the ones you have.

A one-person, one-account seller with a handful of low-stakes SKUs doesn’t lose much — asking the agent instead of digging through Seller Central is a straight upgrade. But plenty of single-account sellers are running six figures through a handful of products, and account count stops being the number that matters there. One suppressed listing or one lost Buy Box on your best seller can cost more in a week than a monitoring tool costs in a year, and Amazon’s agent only catches that once it’s already been flagged.

Scale that up to a brand running several accounts across marketplaces, and it’s the same problem multiplied — reviewing one drafted fix at a time in a chat isn’t a different job than reviewing one alert at a time on a dashboard. What both of them are actually paying for is coverage that doesn’t depend on knowing to ask: fee changes, badge and category drift, competitive activity, watched continuously instead of one conversation at a time.

This took some time to ship. Why ? Amazon updated its seller agreement in March 2026 with a formal policy for AI agents — PPC Land covered the details — that requires a real human sign-off before any agent takes high-impact action. Amazon’s keeping its own AI on a short leash here, not just everyone else’s.

Also, this is a 1.0 beta, not a finished product. I’ve built agent-driven features on similar infrastructure, and the pattern repeats every time — the cautious, reactive version ships first, and the gaps close once real usage data comes in. I wouldn’t bet against Buy Box or fee-change detection or other comprehensive signals showing up here within 6 months or so. Today, it isn’t there.