For years, Big Tech companies have fought for our attention. Google wanted to own the search, social networks wanted our time and data, and marketplaces wanted to become the place where we eventually made the purchase.
Shopping agents are opening a different battle: who gets to act on our behalf, and where will they be allowed to do it?
This became very visible recently when Amazon blocked Meta’s new AI agent, Muse, from shopping on Amazon.com. Muse can perform tasks on behalf of users, including finding products and making purchases. Amazon says Meta had not asked for permission, that the agent did not identify itself while browsing, and raised concerns about credentials, privacy and security. Meta disputes some of those concerns. The Verge
Those are legitimate questions. If an AI agent places the wrong order, exposes credentials or misunderstands what a customer wants, someone needs to be responsible.
But behind that discussion sits a much bigger strategic question: who owns the customer journey when the customer delegates that journey to an AI?
And Amazon isn’t the only company answering that question. Interestingly, Shopify has taken almost the opposite approach, partnering with Meta to enable Muse to check out using Shop Pay across Shopify stores. MarketWatch
Two commerce platforms. The same AI agent. Two very different responses.
When the customer no longer visits your store
Think about how eCommerce works today. You search for something, click on a result or advertisement, enter a store, compare products, look at reviews, see recommendations, perhaps discover something you hadn’t considered, add a product to your cart and eventually check out.
Businesses have spent years—and a lot of money—optimizing every step of this funnel. SEO, paid search, product pages, recommendations, personalization, UX, retail media, conversion optimization and loyalty programs all assume that, at some point, a human is going to interact with the experience.
Now imagine that I need new hiking shoes. Instead of opening Google or going directly to my favorite retailer, I tell my AI agent:
“Find me good waterproof hiking shoes for weekend hikes in Switzerland. My budget is CHF 150–200. Check the reviews, make sure they can be delivered before Friday and give me the three best options.”
The agent searches, compares specifications, checks prices and availability, and comes back with its recommendation. Eventually, I may simply tell it: “Buy the second one.”
I might never visit the retailer’s website. The traditional funnel suddenly becomes much shorter:
Intent → Agent → Transaction.
That has obvious consequences for advertising and customer acquisition. What happens to sponsored product placements if the customer never sees them? What does SEO mean when the audience discovering your product is an AI? How do brands build awareness when purchasing decisions are increasingly mediated by an agent?
But I think there is another question that could be just as important.
Selling the shoes is only part of the opportunity
Imagine my agent chooses a pair of hiking shoes for CHF 169.
For the retailer, my visit has traditionally represented an opportunity worth more than CHF 169. If I walk into a physical store and explain that I’m preparing for a hiking weekend, a good salesperson might ask what equipment I already have. They could recommend proper hiking socks, waterproof trousers, a lightweight rain jacket, perhaps a better backpack, or even sunscreen. I know, thinking about buying sunscreen for a weekend in Switzerland might sound overly optimistic, but every now and then, we actually need it.
The same logic exists online. Customers also bought… Recommended for you… Complete your equipment…
My CHF 169 purchase could easily become CHF 250 or CHF 300.
Now remove me from that interaction.
My AI agent arrives with a very specific mission: buy Jaime a pair of hiking shoes for less than CHF 200. It finds them. It buys them. Mission accomplished. That’s great for the customer. But potentially less interesting for the retailer.
This is where I think becoming Agent Ready will mean much more than allowing an AI to access your catalog and complete a checkout.
How do you upsell to an AI?
Imagine that when the agent evaluates the hiking shoes, the retailer can also provide contextual recommendations.
Not simply:
“You may also like these socks.”
But information an agent can actually reason about:
“For multi-hour hikes, these merino hiking socks are recommended with this shoe because they improve moisture management and reduce friction. CHF 24.”
The agent could then return to me:
“I found the shoes for CHF 169. Since you’re planning longer hikes, the retailer also recommends these merino socks. They’re CHF 24 and reviews are particularly positive regarding blister prevention. Do you want me to add them?”
That’s cross-selling, but the sales conversation is now happening between the retailer’s commerce infrastructure and my AI agent.
The same applies to upselling. Perhaps there is another version of the shoe for CHF 199 that is lighter and has better waterproofing. If the retailer can explain that difference in a structured, trustworthy way, my agent can determine whether the upgrade is relevant to my original request.
This creates a fascinating new challenge for eCommerce teams:
How do you convince an AI agent that an additional product, bundle or upgrade genuinely creates value for its user?
This goes far beyond retail
Travel provides an even clearer example. Imagine I tell my agent:
“Find me a family-friendly hotel in Rome for three nights in October. Somewhere central, with good reviews, maximum CHF 300 per night.”
The agent compares options and books the hotel. But traditionally, that hotel search creates many other commercial opportunities. The hotel might offer a room upgrade or breakfast. A travel platform might recommend an airport transfer, Vatican tickets, a guided tour, restaurants or other activities.
If the agent bypasses the traditional experience, those opportunities don’t necessarily disappear, but they need to be presented differently.
Perhaps the hotel communicates that the family room costs CHF 35 more but includes breakfast for four. An activity platform could indicate that a particular Vatican tour has availability matching the family’s dates and is especially suitable for children. A transfer provider could offer a fixed-price airport pickup that is cheaper than the expected taxi fare.
The AI agent can evaluate those propositions against what it already knows about the trip and present the ones that actually make sense.
In that world, companies aren’t only competing for human attention anymore.
They are competing to be recommended by the agent.
Not every business will want the same strategy
This helps explain why companies may respond very differently to shopping agents.
For a platform like Shopify, making it easier for agents to purchase from merchants can create another transaction channel. Shopify’s partnership with Meta allows Muse purchases through Shop Pay across Shopify stores. MarketWatch
But other business models may depend much more heavily on controlling the experience around the transaction.
Perhaps advertising is a significant revenue stream. Perhaps cross-selling and upselling drive a large part of profitability. Maybe loyalty, subscriptions or first-party customer data are strategically important. Or perhaps the company simply doesn’t want another technology platform sitting between it and its customers.
There won’t necessarily be one correct approach.
Businesses will have to decide whether they want to allow agents, block them, partner with selected providers or create controlled access through APIs. They will also need to decide what agents can see and do: pricing, inventory, personalized offers, loyalty benefits, recommendations, checkout, returns or customer information.
That makes agent access much more than an IT discussion. It becomes an eCommerce and business strategy decision.
Is your eCommerce Agent Ready?
Over the last decade, companies have worked to become mobile-first, omnichannel, API-driven and, more recently, AI-ready.
Agentic commerce adds another requirement: being Agent Ready.
And I don’t think this simply means exposing a product catalog through an API.
Product information needs to be structured and understandable. Availability and pricing need to be reliable. Product relationships need to have context. Bundles and upgrades need machine-readable logic. Identity, permissions, payments and returns need clear rules.
This could also make some of the less glamorous parts of digital transformation suddenly much more strategic.
Product data quality, taxonomy, PIM, APIs and integration architecture aren’t usually the topics that make headlines. But if an AI agent cannot understand what your product is, how it differs from another product, who it is designed for or why another product complements it, your beautifully designed product page may not matter very much.
We have spent years optimizing UX — User Experience and CX — Customer Experience.
Perhaps we now need to start thinking about another layer: AX — Agent Experience.
How easy is it for an agent to understand your products, trust your information, compare your offering, transact with you and discover additional value for its user?
The next battle is about access
Amazon blocking Meta’s Muse might look like another disagreement between two technology giants. But I think it gives us an early glimpse of a much bigger change.
Big Tech companies are beginning to decide which AI agents can enter their ecosystems, what those agents are allowed to do and under whose rules. Amazon’s decision also isn’t happening in isolation: the company has previously challenged third-party AI shopping access, while simultaneously developing its own AI shopping capabilities. TechCrunch
Soon, however, this won’t only be a Big Tech question.
Retailers, hotels, airlines, manufacturers, banks and thousands of other businesses may face their own version of the same decision.
Most companies today are asking, “How can we use AI in our business?”
There is another question worth adding: “How will our customers’ AI agents do business with us?”
Because opening the door may only be the beginning.
The real competitive advantage could come from what happens once the agent walks through it.
Leave a comment