In Part 1, I wrote about Amazon wanting to handle your warehouses, shipping and retail distribution. Now, let’s talk about who’s running the account.

What Seller Assistant promises

At Accelerate, Amazon showed a version of Seller Assistant that connects sales, inventory, advertising and compliance, recommends what to do next, and prepares actions for approval. Amazon’s recap

Even more importantly, a few of the accounts already have a “Connect to Claude” button. This could replace some of the custom-built MCP connectors your team or agency uses. I personally have not tested it yet, but I’m curious to hear feedback from those who have.

In the demonstrations, Seller Assistant compared promotions at different discount levels, including their effect on margin. It looked for pricing opportunities, spotted patterns in returns, prepared advertising campaigns and helped plan inventory shipments.

That’s useful when it works. So far, trying to get useful answers in real-life client accounts has been a challenge.

If you remember the free version of ChatGPT 3.5… Amazon’s Seller Assistant is not even there yet. I’d describe Seller Assistant as a very eager, easily confused but overly confident intern. Keep using it, but don’t actually rely on it in any way, shape or form.

The direction is clear, though, especially for those in the SaaS business developing or selling products for brands that sell on Amazon. Amazon wants sellers to use its tools rather than third-party tools.

What the recommendation leaves out

A promotion can look profitable on Amazon while using inventory you need for a retail launch or cash you’ve committed elsewhere. Unless the system has that context, it won’t be part of the recommendation.

I’ve worked with Amazon since 2011. I’m interested in anything that makes the work easier. I still want to know what assumptions are behind the recommendation before someone clicks approve.

  • A retail-first brand may already have shoppers looking for it on Amazon.
  • A DTC-first brand may be paying to introduce customers who then buy on Amazon.
  • An Amazon-first brand may need the account’s cash to fund expansion elsewhere.

The same recommendation to spend more on Amazon means something different in each business. Before approving it, I’d want to know where the demand came from and what else needs that money.

New to Amazon is not new to your business

Amazon’s Sponsored Brands new-to-brand reporting uses a 12-month purchase lookback on Amazon. That person could already be buying from your website or picking up your products at Target. Amazon’s measurement guidance

I’m particularly skeptical of using that number to judge customer acquisition. Even someone who bought from you on Amazon more than a year ago can count as new-to-brand again.

Amazon itself acknowledged the value of a longer history when it announced five-year purchase data in Amazon Marketing Cloud, pointing to better new-to-brand measurement for seasonal products and products people replace less often.

And a customer searching for your brand and clicking your ad may have been planning to buy anyway. Being counted as new-to-brand doesn’t establish that the ad brought them to you.

That’s one reason we built a new-customer dashboard in Enspescio. I want to look more closely at who we’re acquiring before deciding how much to spend acquiring more of them.

It doesn’t make the cross-channel question disappear. New to Amazon and new to your business still aren’t the same thing.

Everywhere. AI can download and read reports. It does not replace an expert. It makes them more productive.

Sales history needs context

One of the brands I’m working with is dealing with decreased FBA capacity for Q4. Its bestseller was taken down last November. The brand’s storage capacity is now severely limited, and the allocation appears to treat last Q4’s sales drop as weaker demand. But the listing was deactivated. Customers couldn’t buy it.

That’s the context I want someone checking before we let a system make the next inventory decision. A sales history can tell you what sold. It doesn’t necessarily tell you why it didn’t.

Another example comes to mind: one brand allowed several promotions to stack and, as a result, blew through its inventory in a day. Without that context, Amazon could view it as a successful promotion to replicate instead of a mistake that should not be repeated.

Where I’d keep an expert involved

Let the tools pull reports and flag changes. For prices, advertising budgets and inventory commitments, I’d keep someone involved who knows what’s happening in the business.

And yes, I would remember that my friends in Seattle are also selling the advertising and fulfillment services being recommended. That doesn’t make the recommendation wrong. It does mean I’d check it against our own numbers.

I want these tools to get better. I’d happily spend less time pulling reports and more time figuring out what to do with them. But someone still needs to know that last November’s sales dropped because the listing was down, or that a record sales day was actually a promotion mistake. That’s the expertise I’d want in the room as Amazon takes on more of the work.