Skip to content
Blog/

Meta's AI agent has 10 million users. Amazon blocked it.

Brands are spending to show up in ChatGPT. Very few are ready to let an agent check out for them.

Last Wednesday, Meta said more than 10 million people in the US and Canada have set up Muse, the personal AI agent it launched in August, and about a quarter of them use it every day (AP). TechCrunch reports shopping is one of the top things people use it for. Zuckerberg on the earnings call: “if you want to buy something, it’s not just navigating to that business’s website; it’s talking to their agent” (TechCrunch, via Winzheng).

Not everyone is letting it in. Amazon blocked Muse from its store less than two weeks after launch, saying it “had not authorized the agent to access customer accounts, scrape data or process transactions.” Shopify went the other way and is letting Muse check out through Shop Pay (Let’s Data Science).

So we went back through our interviews to see where brands actually are.

For every 9 brands paying to show up in AI, only 1 is ready to let it buy

121 buyers shopping for tools to show up in AI answers (AEO / GEO)

14 buyers who put agentic commerce (letting an AI agent do the buying) in their top priorities

In the last 12 months, 121 buyers expressed that they’re actively investing in AI discovery. But only 14 are ready to let AI do the actual shopping. That’s a huge disparity!

Here’s how a VP of ecommerce at a $10B-$100B consumer goods company is evaluating tools to score how its brands show up:

“This is such a new space for us, and we are starting to track our search in terms of how our brands appear when people are conversing with LLM tools like ChatGPT and Gemini. To retain our competitive advantage, we need to ensure our brands are appearing in the right conversations and are visible enough for consumers.”

VP Ecommerce, $10B-$100B consumer goods company

But here’s the hesitation to take it any further. The same VP, on the biggest challenge:

“I think the biggest challenge right now is the amount of ambiguity around the tech related to AI and how we integrate those into our workstream. The AI technology, as you can imagine, is changing so fast, so how do we keep pace with that while embedding all of this new technology into our legacy systems…”

VP Ecommerce, $10B-$100B consumer goods company

Even the brands that want agents to sell for them worry about what AI says on their behalf. A VP of ecommerce at a $100M-$500M baby products company is ready to sell through Shopify’s agentic storefronts as soon as they open, but draws a line here:

“Given how intensely regulated the business is of infant car seats, especially around safety… we absolutely have to be extremely sensitive and pay attention to the types of answers that an AI engine is providing about child car seat safety and laws to avoid lawsuits. So when I think about utilizing AI as a front-end tool on our website, I have to be extremely careful, and we have to have some real tight guardrails…”

VP Ecommerce, $100M-$500M baby products company

The ones getting ready to sell through agents start with product data

PIM is having a big moment. A VP of ecommerce at a $1B-$10B specialty retailer already has a team on this, and brought up Muse by name this month:

“We have product management who are focused on not only the core website improvements and optimizations, but they also work on the AI visibility and agentic commerce initiatives that really are helping position us to have visibility within ChatGPT, Gemini, and the other emerging LLMs, including Meta’s Muse.”

VP Ecommerce, $1B-$10B specialty retailer

Their next purchase is a PIM. Asked why:

“Data hygiene, data sourcing, data supply chain and hygiene, and we’re also understanding the role of PIMs in terms of AI data enrichment, supporting overall AI visibility, and the longer-term trend of agentic commerce.”

VP Ecommerce, $1B-$10B specialty retailer

A marketing manager at a $10B-$100B beauty company has the same goal and the same bottleneck:

“…and ensuring agentic commerce is up and running and utilizing that our data is able to inform the buy box on LLMs such as OpenAI, ChatGPT, and Gemini.”

Manager Marketing, $10B-$100B beauty company

“We know what we need to do to improve our product data, but we don’t necessarily have the bandwidth within the company right now to do all of those optimizations.”

Manager Marketing, $10B-$100B beauty company

And an ecommerce manager at a $10B-$100B retailer described what’s wrong with the feeds an agent would read:

“For example, feeds might have duplicate images… something might say blue shoe and it might be the red option. What else? Like just missing product attributes.”

Manager Ecommerce, $10B-$100B retailer

That said, most brands are still watching

For a lot of brands, agentic commerce is on the radar but not in the budget yet. A $1B-$10B personal care company listed it as one of the biggest shifts it’s planning around:

“And another one is to fully anticipate how agentic commerce will really impact us, in a way that either purchase directly from AI or influenced by AI.”

Director Ecommerce, $1B-$10B personal care company

The CMO of a $10M-$100M food company said they’ve “leaned heavily into AI for agentic commerce.” Asked whether that changes what they’ll buy:

“Not currently, no. But I mean, we are looking at our CRM software that has more agentic properties to it.”

CMO, $10M-$100M food company

What to do about it

  1. If you sell AEO or GEO tools, the next question is checkout. Your buyers are paying to get mentioned. Once 10 million people have an agent that can buy, the follow-up is whether that agent can actually complete the order (Shop Pay, Stripe and the new commerce protocols). Have an answer.
  2. If you sell PIM, feeds or product data, lead with agents. The buyers furthest along are tying product data directly to agentic commerce, and the ones behind say bandwidth is the problem. Show how little work it takes on their side.
  3. If you sell anything else on the site, ask how it looks to an agent. Search, reviews, promotions, checkout. If buying starts with “talking to their agent,” that’s the new front door.

Ten million people already have an agent that can shop for them. Most brands are still working on getting mentioned by one. That gap is where the next round of budget goes.

Jon

P.S. Those 121 AEO and GEO conversations also cover who’s already working on agentic checkout and what they’re buying for it. If you want the breakdown, let us know. Best case, you spot your next opp. Worst case, you know what you’re up against.