AI CAN ACT
TFTT Friday Review
On Monday, we looked at AI changing the way customers search. By Friday, the story has moved on. AI agents are starting to act.
That distinction matters. We have spent the last couple of years talking about generative AI helping customers discover products, answer questions, compare options and make decisions. This week, we saw several developments that move the technology another step forward: AI is increasingly being built to take action on the customer's behalf.
On 8 September, Meta launched Muse, its new personal AI agent. Meta says Muse can work across the apps people use, open a browser, fill in forms, negotiate on a user's behalf and continue working after the user has closed the app. Crucially for commerce, Muse can also make purchases when the user gives approval. Meta has integrated Stripe's Link for checkout, with Link's wallet for agents able to generate a one-time-use card so the customer's real card details remain hidden.
That is a very different proposition from a chatbot.
You're no longer simply asking:
“What should I buy?”
You are potentially saying:
“Go and do it.”
And that changes the role AI can play in the customer journey. The agent can increasingly move from answering the question to executing the task. It can search, compare, navigate websites, complete forms and, with the appropriate permissions, move towards the transaction itself. For retail, that is a significant change because the customer journey has historically been designed around a person doing the work. The shopper searches. The shopper compares. The shopper decides. The shopper adds to basket. The shopper pays. Agentic commerce starts to put another decision-maker into that journey. The agent.
TFTT INSIDER'S VIEW
For years, retailers have fought for visibility through stores, marketplaces, Google, social platforms, paid search, SEO and increasingly social commerce. Now another layer is emerging. AI commerce channels.
The question is no longer simply whether a customer can find your product. It is whether an AI agent can find it, understand it, evaluate it and ultimately choose it.
That is why Shopify's announcement this week matters. On 8 September, Shopify added Meta as an AI channel inside its Agentic Storefronts. Shopify says products are shared with Meta by default through Shopify Catalog, while merchants can manage catalogue access, direct checkout and performance through their agentic settings. That may sound like a technical platform update. It isn't. It is another indication that AI is beginning to become part of the distribution infrastructure of commerce.
For fashion brands, that could be important. The digital shelf has historically meant your own website, retailer websites, marketplaces, search engines and social platforms. The emerging AI shelf is different because the consumer may not necessarily browse the shelf themselves. The agent may do the browsing. And if the agent is doing the browsing, the way a brand presents its products to a machine becomes increasingly important.
THE AI SHELF
This is where the conversation becomes much more practical for fashion. An AI agent needs more than a product image. It needs to understand what the product is, who it is for, what it costs, what sizes and colours are available, when it can be delivered, what the returns policy is and whether it actually satisfies the customer's requirements.
In other words, product information becomes part of the sales experience for the machine as well as the human.
That is reinforced by research published this week by Akeneo. Its Agentic Commerce Reality Check surveyed 1,000 senior IT decision-makers across the US and Europe. 95% said they were confident their organisations were ready for AI-driven commerce, yet the research highlights the very different reality underneath that confidence: data preparation, integration and governance remain major challenges. Akeneo says 89% of respondents are spending more than a quarter of every AI project preparing data. That is a fascinating tension. Businesses may believe they are AI-ready. But being AI-ready is not necessarily about having a chatbot or buying another AI tool. It may start with something much less glamorous:
Is your product data actually good enough for a machine to understand and act on?
For fashion, that means the quality and consistency of product information suddenly has a much bigger commercial implication. If an agent is comparing ten products on behalf of a customer, inaccurate descriptions, incomplete attributes, unclear availability or inconsistent pricing could potentially take a product out of consideration before the human customer ever sees it. That is a very different kind of digital competition.
THE INFRASTRUCTURE IS BEING BUILT
And this isn't just happening at the consumer-facing end. The payments and commerce infrastructure is moving at the same time.
On 8 September, Mastercard published a new report predicting that more than one in ten online shoppers could routinely use AI agents to purchase products on their behalf by 2030. The report describes a future in which agents can search, compare, negotiate and transact within boundaries set by the consumer.
Then, on 9 September, Mastercard announced Agent Connect and expanded its Agent Suite for Merchants. The proposition is designed to help merchants make products discoverable across AI ecosystems, support product discovery and consumer-authorised purchases, while maintaining control over product information, customer relationships and business rules.
That last part is particularly important. Because agentic commerce isn't simply a technology question. It is a commercial control question.
Which agents can access your catalogue?
What information are they seeing?
Which price is being presented?
What stock is available?
What happens to promotions?
Who owns the customer relationship?
How does the brand maintain control over the experience?
And what happens when the agent makes a decision that the retailer would not have made itself?
These questions are no longer entirely theoretical. The infrastructure being announced this week is designed specifically to start answering them.
BUT THERE IS A BIGGER BARRIER
There is, however, one major constraint. The consumer still has to trust the agent.
Visa published its new Trust Index for agentic commerce on 9 September. The research found that only 23% of US consumers currently trust generative AI to handle payment transactions on their behalf. That is a striking gap. People may be increasingly comfortable using AI to discover products. They may be happy for it to recommend a restaurant, find a product or compare options. But allowing it to actually press buy on their behalf is a much bigger psychological and commercial step. And Visa's research reveals something even more interesting.
When respondents were asked about trusted payment brands, 61% said they would trust Visa to handle an agentic transaction. The survey was conducted by Harris Poll for Visa among 2,065 US consumers, with the payment-brand question tested among samples of approximately 1,028–1,034 consumers per brand.
So the challenge isn't simply:
Can AI do it?
It is:
Do I trust AI to do it for me?
And perhaps even more importantly:
Who do I trust to stand behind the transaction?
That could become one of the defining questions of agentic commerce.
TFTT INSIDER'S VIEW
For fashion, this creates a fascinating new dynamic. The industry has spent decades building trust around brands, retailers, marketplaces and payment providers. Now an AI agent potentially sits between the customer and all of them. That means brands will need to think about more than simply whether their AI tools work.
They will need to think about whether their products, data, pricing, policies and brand proposition can be understood and trusted by an agent. The traditional customer journey was designed to persuade a person. The emerging agentic journey may need to persuade both the person and the machine acting for them. And that is where this becomes much bigger than another technology trend.
If AI becomes a meaningful route to discovery and purchase, then product information becomes distribution. Availability becomes discoverability. Trust becomes infrastructure. And the ability to transact becomes part of the customer experience.
THE SIGNAL
This week gives us a pretty clear picture of where the market is heading. Meta launched an agent designed to take action, including making purchases with consumer approval. Shopify added Meta as an AI channel, connecting merchant product catalogues to an emerging agentic commerce environment.
Mastercard is building infrastructure to connect merchants, AI agents and payments, while predicting that more than one in ten online shoppers could routinely use AI agents to shop and pay by 2030.
Akeneo's research highlights the less visible challenge underneath all of this: AI ambitions are running into product-data, integration and governance realities.
And Visa's research highlights the other side of the equation: consumer trust has not yet caught up with technological capability.
None of this means traditional ecommerce is disappearing tomorrow. It isn't. Agentic commerce is still developing, consumer adoption is still evolving and there are significant questions around trust, security, permissions, attribution and control. But the direction of travel is becoming much harder to ignore.
AI is moving from answering questions to taking actions.
And when AI can increasingly search, compare, recommend, transact and complete tasks on behalf of the customer, the question for fashion brands changes.
It's no longer simply:
“How do we get the customer to find us?”
It becomes:
“How do we make sure the AI can find us, understand us, trust us and choose us?”
That is not just an AI question.
It is a new retail question.
THE FASHION THINK TANK
INSIDE FASHION. NOT OBSERVING IT.
