Quick answer: Meta’s Muse passed 10 million sign ups in its first six weeks, about a quarter of its users open it every day, and more than a quarter of weekly users buy something through it, at an average of over $200 a transaction. Walmart opened its doors to AI agents on October 21. Amazon locked its own. For any business that sells a product, the job between now and Black Friday is simple to say and takes a month to do: make your products the ones an agent can find, trust and buy. Here is where to start.
What changed in two weeks
When I wrote about my first three hours with Bruno, the name I gave my Muse agent, the story was a curiosity. Meta had built an agent that could audit my Instagram with data no other AI could see and then buy a bat bag with my money. Two weeks later it is not a curiosity. It is a channel.
Here is what Meta told investors on its third quarter earnings call. Muse is the top app in the App Store almost every day. More than 10 million people in the United States and Canada have signed up. Shopping is the number one thing they do with it, and the volume ran so far past Meta’s own forecast that Stripe had to raise the credit limits in week one.
Then, on October 21, Walmart did something I have been waiting for a major retailer to do. It launched an Agent Gateway, a front door built specifically for AI agents, with its own protocol. Walmart’s reasoning was one sentence long and every business owner should tape it to the office wall: “The way agents shop is different.” Your website was designed for a person with a mouse. An agent does not scroll, does not get persuaded by a hero image, and does not forgive a missing size chart. It retrieves, compares and acts.
Amazon made the opposite call and blocked Muse with an anti bot wall. Both companies are rational. Amazon wants to be the agent. Walmart wants to be found by every agent. You do not have Amazon’s $588 billion, so you only have one of those choices.
Amazon wants to be the agent. Walmart wants to be found by every agent. You only have one of those choices.
The numbers that should get your attention
Shopify surveyed holiday shoppers and found 65 percent will use AI for at least one holiday task this year, but only 29 percent trust it to complete a purchase. That 36 point gap is the whole game. Trust is what a clean checkout and plain policies buy you, and most stores have neither in a form an agent can read. PayPal found only about one in five merchants has most of its catalogue in machine readable form. Four out of five stores are, to an agent, half empty shelves.
And the shoppers who do arrive through AI are worth more. Adobe measured AI referred visitors converting about 40 percent better than site average, and forecasts AI referred traffic to U.S. retail sites up 130 percent this holiday season. Salesforce expects AI chat agents to drive 20 percent of global holiday e-commerce traffic. One in five holiday visits may start with an agent, not a search box. And the agent gives three named products, not ten blue links. The store the AI can describe accurately in one sentence wins the mention.
Three of the ten things I told NSGA members last week
As AI Expert in Residence for the National Sporting Goods Association, I spent Thursday afternoon with independent sporting goods retailers walking through a ten step, 30 day sprint to Black Friday. The full sprint belongs to NSGA members, and if you sell sporting goods that alone is worth the membership. But three of the steps apply to anyone who sells anything, and they are the three I would start with tomorrow morning.
1. See yourself through the AI’s eyes
Before you change a single thing, ask Muse and ChatGPT the exact questions your customers ask, not the questions you wish they asked. “Best running shoes for flat feet under $120.” “Youth catcher’s gear set near me.” “Warmest women’s ski gloves.” Write down who gets named, and whether you appear at all. Ten real customer questions, about an hour, and you will know your invisible spots. Most owners have never done this. The ones who did it live on Thursday were not discouraged when they were missing. They were relieved, because now they knew the job.
2. Fix your top 20 SKUs first
Not 2,000. The twenty products that pay the rent need complete attributes (size, fit, material, weight, intended use, age, gender), a live price and stock status, and a real GTIN or UPC so the assistant can match the product with confidence. AI reads raw product data, not marketing copy. A partial match gets skipped for a competitor with full detail, and stale stock is a silent disqualifier. Two to three hours, done when all twenty have zero missing core attributes. Then rewrite the titles so a machine can read them. “The Hartley” is beautiful brand poetry and invisible to an assistant that ranks on facts. “Merino crew neck, mid weight, navy” wins the mention.
3. Unblock the AI crawlers
This is the one that made the room go quiet. Check yoursite.com/robots.txt. The search crawlers you must allow are OAI-SearchBot, Claude-SearchBot, Googlebot, Bingbot, Applebot and Meta-WebIndexer. Blocking these erases you from AI answers entirely. The training crawlers (GPTBot, ClaudeBot, CCBot, Bytespider) are your choice, and blocking them does not remove you from answers. The user triggered fetchers (ChatGPT-User, Claude-User) fire when a real customer pastes your link, so leave them alone. The 2023 “block all AI bots” wave is the trap. One line in that file can erase a store, and a lot of web developers added that line two years ago and never looked again. Fifteen minutes with your web person, and changes take about a day to show.
The other seven cover reviews, policy pages, checkout posture, measuring AI traffic on its own, product feeds and briefing your floor team, and they are sequenced week by week so the whole thing is done before Black Friday without adding hours to anyone’s day. If you are an NSGA member, the recording and the slides are in your member resources. If you are not, the Business Audit below is where I do this with you, one business at a time.
The calendar
Black Friday is November 27. Agents re crawl on their own schedule, not yours, so changes shipped in the second half of November may not be seen in time. Fix product data this week. Ship page changes by November 14. Test your own checkout the week after, as a guest, on your phone. Set up an AI referral segment in your analytics now so you can see what the agents sent you, and track the prompts your customers actually use, like “best gift for a runner under $100,” because the sale is decided in the answer before anyone reaches your checkout.
What Muse still cannot do, and why that matters
Muse completes checkout in the chat for Shopify stores today. For a retailer like Dick’s that is not on Shopify, it hands off to an approval step instead of buying silently. It cannot like, comment, follow or reply on Instagram. And when I asked Bruno to be specific about editing my own website’s crawler settings, his answer vanished behind a safety notice twice. You are building on someone else’s platform, and the rules can change under you without warning. Diversify. Be findable by every agent, not just the one that is winning this month.
From the retailer’s chair
I spent high school on a sporting goods floor selling skis and running shoes, and I co led one of the largest computer vision deployments in U.S. retail. I have watched a lot of technology arrive in stores with a lot of promises. This one is different in one specific way. The customer is not coming to you. Her agent is. The only question left is whether your products are the ones it can find, trust and buy.
The customer is not coming to you. Her agent is.
Two ways I can help
If you want this on your stage
Agentic commerce is the conversation every retail and association audience needs before the 2027 planning cycle, and I deliver it from the store floor, not the slide. I grew up above my family’s restaurant, sold skis and running shoes through high school, and co led one of the largest computer vision deployments in U.S. retail before I ever stood on a keynote stage. If your conference, leadership meeting or association is working out what AI agents mean for your members, let’s talk about a keynote or a workshop.
Looking for a keynote speaker on AI, agents and the future of retail?
If you want it done for your business
The Get FOUND Business Audit is 90 minutes with me, live, looking at exactly what ChatGPT, Gemini and Muse say about your business today, why, and what to fix first. You get your written plan within two business days and a year inside the community that gets this information before anyone else. It is how I took Harmoni to #1 on ChatGPT in a weekend, and it is where I do this with you, one business at a time.
Find out what AI says about your business, and fix it.
Frequently asked questions
How many people use Meta Muse?
More than 10 million sign ups in the United States and Canada in its first six weeks, with about a quarter using it on any given day, according to Meta’s Q3 2026 earnings call.
Does Muse buy from small retailers?
Yes. Any eligible Shopify store is in its catalogue with in chat checkout. Other platforms depend on whether the agent can read your products and complete your checkout.
Does Muse work with Amazon?
No. Amazon blocked it on September 20, 2026 and is building its own shopping agents instead.
What is Walmart’s Agent Gateway?
A front door Walmart opened on October 21, 2026 so AI agents can shop its catalogue reliably, with spending limits set by the customer and no access to card details.
What should a business owner do first?
Ask ChatGPT and Muse to recommend a store for what you sell. If you are not in the answer, fix your product data and your robots.txt before anything else.
Is it too late for this holiday season?
Not if you ship your data and page fixes by mid November. Agents re crawl on their own schedule, so the earlier the better.
About Susan Sly
Susan Sly is an AI keynote speaker and AI ethicist, recognized as one of the Top 7 Female AI Thought Leaders in the World. A graduate of MIT Sloan and the MIT School of Engineering, she has led large scale AI deployments in retail and presents to enterprise teams at companies including NVIDIA, HPE, Intel, and Lenovo. She serves as AI Expert in Residence for the National Sporting Goods Association.
