Amazon's Seller Assistant Can Now Watch Your Account. It Still Can't See Your Business

TL;DR

What to expect from this article: A practical read on what Amazon announced for Seller Assistant at Accelerate 2026. It covers what the assistant can and can't see, which automations are worth turning on now, and where brands should keep a human approval step in place.

Key takeaways:

•       The real upgrade in Seller Assistant is memory and continuous monitoring, not autonomy. An assistant that tracks 100+ metrics and flags problems early is useful regardless of how much you let it act.

•       Seller Assistant is not omniscient, it only knows what lives in Seller Central. It doesn't know your cash position, supplier lead times, MAP policy, wholesale commitments, or that you plan to discontinue a SKU, unless you take the time to feed it that information.

•       Account health and returns monitoring are low-risk automations worth turning on immediately.

•       Pricing automation optimizes for visibility inside Amazon. Make sure to set hard price floors before enabling it, and keep anything changing price or ad spend on drafts-for-review until you're confident of its reliability with your account.

•       The fastest way to get value from custom automations is to translate checks your team already runs every week into plain-language rules, with approval required.

What did Amazon announce for Seller Assistant at Accelerate 2026?

Amazon gave Seller Assistant three things it didn't have before: persistent memory of your business, continuous monitoring across more than 100 business metrics, and automations that run in the background at a level of autonomy you choose. According to Amazon's September 23, 2026 announcement, Seller Assistant now keeps a persistent memory of each seller's pricing patterns, inventory cycles, and growth goals, and that memory carries across Seller Central, Amazon Quick, and Claude. On stage at the Accelerate 2026 Day 1 General Session, Amazon added that the assistant now tracks over 100 metrics (up from a handful) and can run workflows with hundreds of steps.

Three stages of capacities were announced, including a future stage currently in development:

On control, Amazon says every action is logged with a complete audit trail. For any workflow, sellers can choose whether it makes recommendations or takes actions on their behalf, and actions are reviewed and approved before they're carried out. The assistant is built on Amazon Bedrock with Anthropic's Claude models, and Amazon reports that sellers accept its recommendations more than 90% of the time.

Regarding this acceptance rate, a 90% acceptance rate can mean the recommendations are good, but it can also mean sellers are approving things they don't have the time or bandwidth to evaluate. For your account, this will depend on whether someone with full business context, and understanding of the big picture, appointed to be fully accountable, is doing the reviewing.

What can't Amazon's Seller Assistant see?

Seller Assistant can only see what's already in Seller Central, which means it actually lacks most of the information that is needed to decide whether a recommendation is actually good for your business. For instance the margins it calculates will be built from Amazon fees and your Amazon sales, not your full cost structure and logistics. Namely, it doesn't know:

•       Your cash position. A restock recommendation that makes sense in terms of sale velocity might not actually be good in terms of liquidity, especially going into Q4.

•       Your true margin by product. A product could look to have a good ROAS, but might be earning a thin margin once freight, returns, and trading terms with retail partners are included. Seller Central sees the ad return, but not the landed cost.

•       Supplier lead times and minimums. It sees inventory levels, not the fact that your manufacturer needs 90 days, requires a minimum order quantity, or just raised it. A restock plan that ignores MOQs either can't be placed or ties up more cash than it shows.

•       Your catalog plans. It will happily optimize a SKU you've already decided to discontinue.

•       Your pricing commitments off Amazon. MAP policy, wholesale agreements, and retail-partner pricing live outside Seller Central, and a price change that wins the Buy Box can create a problem with a retail buyer.

To bridge this with the missing data, Amazon's new Selling Partner plugin connects a seller's Amazon data to Amazon Quick and to Claude (in beta), so sellers can bring in data from the rest of their operations, such as financial, marketing, and manufacturing. The plugin helps to facilitate the process, but someone still needs to connect those systems and very importantly, keep them up to date. For most brands, the more immediate step is simpler: tell the assistant your margin floors, launch plans, and discontinuation decisions directly, since it now remembers them.

Which Seller Assistant automations should you turn on now?

Turn on account health monitoring and returns pattern detection now. Both are low-risk because they help identify problems rather than implement any changes, and both cover issues that can often go unnoticed until they've already incurred losses.

Account health is the clearest case. A Beauty & Personal Care brand we work with had a product doing roughly $30,000 a month get suppressed repeatedly. Across those suppressions, direct lost revenue came to roughly $19,000. At the core, it wasn’t an advertising issue nor a pricing issue, but a monitoring problem. And every hour a listing sits suppressed before someone notices is lost sales, and the fix was catching it faster, not spending more. That speed in identifying issues is exactly where a 24/7 account health watch showcases its value. (For how suppressions happen and how to prevent them, see Amazon Listing Suppressed? Causes, Fixes, and How to Prevent It.)

Returns monitoring earns its place for a similar reason. A rising return rate on one ASIN is often the earliest signal of a listing that's overpromising, a variation mismatch, or a quality issue in a new batch. Those problems usually show up in reviews weeks later.

Should you let Seller Assistant change your prices automatically?

Not yet, and not without hard price floors in place first. According to what Amazon showed on stage, the pre-built daily pricing check weighs category prices, fees, margins, and the last seven days of demand. That's a reasonable model for visibility inside Amazon. It's an incomplete model for your business.

Seven days is a short window to price against. Jungle Scout's 2026 Benchmark Report shows why: in Home & Kitchen, searches for Owala were up 29% on a 30-day view but down 49% on a 90-day view. A short-term spike was masking an underlying decline. A pricing rule reading only the last week of demand would have seen strength where there was none.

The practical guardrails:

•       Set a hard floor per SKU based on contribution margin after COGS, not Amazon fees alone. The assistant doesn't have your landed cost unless you give it.

•       Keep price and ad spend on drafts-for-review until you've watched the assistant's recommendations on your own account for several weeks and can judge how often you'd have overridden them.

•       Exclude SKUs covered by MAP or wholesale agreements from any pricing automation entirely.

This matters most for brands with retail distribution. A price move that wins on Amazon can create problems with retail partners that no Seller Central metric will show.

How do you turn your team's weekly checks into custom automations?

Start with the checks your team already runs every week. Rules your team already follows translate directly into custom automations, since you're describing a process you've already validated rather than inventing a new one. Amazon's own example is as simple as an alert when any top-10 product drops below a 4-star rating, paired with a drafted response plan.

A good way to start:

1.    List three checks your team does every week. Examples: weekly stock cover on top sellers, week-over-week conversion drops by ASIN, or new negative reviews mentioning a specific defect.

2.    Pick the one with the clearest trigger and write it as a plain-language rule.

3.    Run it with approval required for a few weeks and compare what it flags against what your team would have caught manually.

If it catches the same things faster, keep it and move to the next check. If it flags noise, tighten the rule before you expand.

Is Seller Assistant's event planner ready for Q4?

No. The event planner is arguably the most interesting feature in the announcement, but it isn't ready for use for Q4. It was described as planning full events, with promotions, ads, inventory, and listings adjusting together. But it was announced as coming later, with no date. Prime Big Deal Days is already scheduled for October 6–7, and holiday plans are largely locked by now.

Plan this Q4 the way you already do, but remember to revisit the event planner for Prime Day 2027, once it's live and Amazon showcases how well it handles promotion depth vs margin in the field.

Does Seller Assistant replace an agency or an in-house Amazon hire?

It can be tempting to think so but in reality, no. It takes over a growing share of monitoring and routine execution, but the most important thing that often gets overlooked is explicit accountability. The oversight and decisions still need someone accountable who can see the whole business, not just Seller Central. The practical question is which parts of your Amazon work belong to the assistant and which don't.

A useful way to decide is to sort your Amazon work into three groups:

1.    Work the assistant can automate. Monitoring, alerts, order summaries, routine reports, and first drafts of restock plans. Turn these on, spot-check them, and take back the hours. But don’t be complacent in periodic spot checks.

2.    Output that needs a person to interpret and challenge. Pricing recommendations, bid changes, restock quantities, and listing edits. The assistant can draft it, but someone with full cost, cash, and inventory context, and thus full accountability should approve or override it.

3.    Decisions that are about business strategy. Which SKUs to grow or cut, how much margin to trade for rank, how Amazon pricing interacts with retail partners, and when to expand. These depend on information the assistant doesn't have, and someone has to own the outcome.

If most of your Amazon workload sits in the first group, Seller Assistant may cover much of what you'd otherwise hire for. Chances are though, you have important tasks/decisions in the second and third category, which makes the review step in fact more important. An assistant that runs hundreds of steps around the clock can compound a bad assumption faster than a person could, and the cost of that shows up in margin, not in the Seller Central dashboard. (For how that judgment layer is usually staffed, see Amazon Agency vs. In-House Hire: The Real Math.)

Related reading

This piece connects to Kyzenn's Catalog Management and Advertising clusters. For why catalog health decides more than ad optimization, see Amazon Catalog Management: The Business Function Most Brands Treat as Housekeeping. For the in-house vs. agency cost comparison, see Amazon Agency vs. In-House Hire: The Real Math.