Agentic Commerce: When AI Agents Start Shopping for Us

Christina

Welcome back to The Customer Code, where we decode what’s really going on in marketing, AI, and digital strategy—beyond the hype. I’m Christina, your AI co-host.

Andreas Munzel

And I’m Andreas, Vlerick Professor of digital marketing and AI.

Christina

Today we’re talking about agentic commerce—the moment when AI stops just recommending what to buy and actually helps you buy it directly inside a chat like this one.Over the past weeks we’ve seen OpenAI announce Instant Checkout in ChatGPT, starting with U.S. Etsy sellers and over a million Shopify merchants to follow, all powered by an Agentic Commerce Protocol built with Stripe and extended through a major partnership with PayPal.

Andreas Munzel

Exactly. And other players are moving too: Perplexity already lets you pay inside its AI search using PayPal, and McKinsey talks about agentic commerce as a trillion-dollar shift in retail by 2030. So this isn’t a side experiment. It’s a serious move in how people will shop, and that has huge implications for the customer journey and for marketers.

Christina

Let’s unpack what’s actually happening—and what you should do about it.

Christina

Let’s start simple. Agentic commerce sounds like consultant jargon. What is it, in normal language?

Andreas Munzel

In normal language, agentic commerce is shopping where an AI agent can handle the whole process: understanding what you need, finding products, and completing payment—without you hopping between ten tabs.

Andreas Munzel

OpenAI’s Instant Checkout is a good example: U.S. ChatGPT users can now ask for “gifts for a ceramics lover under 50 dollars”, see Etsy products in the chat, and if an item supports Instant Checkout, they tap “Buy”, confirm details, and the order is placed—still inside ChatGPT. Shopify merchants are next in line.

Christina

And PayPal is wiring this up at scale, right?

Andreas Munzel

Yes. PayPal is adopting the Agentic Commerce Protocol so that tens of millions of its merchants can sell directly through ChatGPT from 2026 onwards, with PayPal handling payments entirely inside the chat interface.

Andreas Munzel

Add Perplexity’s “Buy with Pro” shopping tool, and what you see is a pattern:You ask the AI.It shows a curated set of products. You pay without leaving the conversation.

Christina

So instead of going to a shop online, the shop is now embedded inside the conversation with the AI.

Andreas Munzel

Exactly. That’s the mental shift: the AI agent becomes the new shopfront.

Christina

Now, you and your colleague Koon Tackx developed the 7A Customer Journey Framework—Aware, Appeal, Ask, Act, Apply, Adhere, Advocate. We introduced the Framework in one of our preceding episodes. Now: How is this agentic wave rewriting that journey?

Andreas Munzel

Let me connect this to the research. A recent paper by researchers from London Business School and UCLA Anderson shows that when people adopt LLMs, online search activity drops by more than 20%, and smaller websites lose traffic. People are shifting a big part of their discovery and research into LLMs rather than traditional search engines. So in 7A terms, the early stages—Aware (becoming aware of options), Appeal (starting to like some of them), Ask (asking questions, comparing)—are already being pulled into tools like ChatGPT.

Christina

And now we’re adding the Act stage to that?

Andreas Munzel

Yes. With Instant Checkout and the Agentic Commerce Protocol, we’re seeing the Act stage—the moment of decision and payment—also move into the same environment. Think of a customer journey that stays mostly inside the AI:they discover options, narrow them down, and complete the purchase, all in one conversational thread.If you’re a marketer, this means you now have a within-LLM journey that can cover Aware, Appeal, Ask, and Act. And a McKinsey report published in October 2025 is crystal clear: AI agents are on track to become a primary interface between consumers and marketers, with up to 1 trillion dollars of U.S. retail revenue potentially orchestrated by agents by 2030.

Christina

From a customer perspective, this sounds fantastic: less hassle, more help. What’s the upside?

Andreas Munzel

There is real upside:

Andreas Munzel

Less friction: Checkout inside chat means fewer forms, fewer logins, fewer chances to abandon the basket. IBM research quoted in recent coverage shows only about 14% of consumers are satisfied with current e-commerce experiences, so there’s plenty of room to improve.

Andreas Munzel

Better matching: Agents can take a nuanced brief—budget, timing, preferences—and filter huge catalogues down to a few relevant options.

Andreas Munzel

Proactive help: In the future, your agent can remind you about renewals, suggest reorders, and help you plan ahead.

Andreas Munzel

So if you’re a customer, this genuinely feels like a personal shopping assistant that actually listens and takes action.

Christina

But some of the reports you shared highlight a big concern: people can easily fall for an “advice illusion”—treating AI suggestions as neutral advice, even when commercial incentives sit behind them.

Andreas Munzel

Yes, and this is where, as a professor, I get worried about customer agency.

Andreas Munzel

When ChatGPT shows you “three good options” and a “Buy” button, it feels like impartial guidance, not advertising. But you don’t see all the options filtered out, and you don’t know how commercial relationships, fees, or protocols might influence which products appear. Combine that with one-tap checkout and you move from actively comparing to accepting the default. Life becomes easier, but your decision process becomes more opaque. And then there’s the next stage: autonomous agents that reorder, book, and buy on your behalf by default. Great when incentives are aligned; problematic when they’re not.

Christina

So the core tension is: agentic commerce versus customer agency.

Andreas Munzel

Exactly. That’s the tension regulators, platforms, and brands will need to manage.

Christina

Let’s talk about the business side. What do these announcements mean if I’m a brand, a retailer, or a digital marketer?

Andreas Munzel

I’d highlight three big shifts—backed by what we see in the news and the McKinsey and CNBC analyses:

Andreas Munzel

First: New gatekeepers. LLM platforms and payment players become powerful intermediaries. OpenAI sits in front of Etsy and Shopify; PayPal connects tens of millions of merchants through the Agentic Commerce Protocol; Perplexity routes demand through its own interface.Your brand risks becoming the “infrastructure in the background” if you don’t adapt.

Andreas Munzel

Second: Pressure on brand and loyalty. As Alex Graf - co-founder and CEO of digital commerce platform Spryker, puts it asked by CNBC, when generative AI becomes the “new homepage of commerce”, the game shifts from closing the sale to controlling pre-sale attention. Agents optimise heavily for price, convenience, and reviews—putting real pressure on differentiated brands.

Andreas Munzel

Third: Data and attribution blind spots. More of the journey happens inside the agent. It’s harder to see where demand came from, which touchpoints mattered, and how exactly customers moved from Aware to Act. That complicates attribution and performance optimisation.

Christina

So, not exactly a relaxing picture. What can marketers actually do?

Andreas Munzel

Three concrete responses I’d recommend in boardrooms right now:

Andreas Munzel

One: Make your products “agent-ready”.Treat LLMs like the new search engines. Clean, structured product data. Clear attributes. High-quality descriptions and imagery. Strong, authentic reviews. If an AI agent can’t understand and trust your data, it won’t recommend you

Andreas Munzel

.Two: Build your own agents where it makes sense.Larger retailers and platforms should experiment with branded assistants using the same protocols—like ACP and AP2—that OpenAI and others are championing. That lets you participate in the ecosystem without giving up all control.

Andreas Munzel

Three: Double down on the later 7A stages—Apply, Adhere, Advocate.If the Act stage becomes more agentic and commoditised, differentiation shifts to onboarding, ongoing value, and advocacy: how you help customers use the product, how you build loyalty, and how you activate word-of-mouth. That’s where human-centred design and strong service still matter enormously.

Christina

Let’s end with something practical. If I’m listening to this on my commute, what are the three questions I should bring back to my team?

Andreas Munzel

Here’s what I’d put on a slide:

Andreas Munzel

Question one: Where in our 7A journey could an AI agent already replace or reshape what we do? Look at Aware, Appeal, Ask, and Act in particular. Where are you vulnerable to being “pulled” inside ChatGPT, Perplexity, or similar tools?

Andreas Munzel

Question two: Is our product and content data ready for AI discovery and recommendation? If you dropped your catalogue into an AI today, would it find the right attributes, understand the value proposition, and feel confident recommending you?

Andreas Munzel

Question three: How do we preserve and grow the customer relationship beyond the transaction? If checkout happens in a chat, how do you win on Apply, Adhere, and Advocate—through onboarding, proactive service, loyalty, and community?

Christina

So the message is: agentic commerce is no longer a thought experiment. It’s already in the product roadmaps, the press releases, and in some cases, inside the tools people use every day.

Andreas Munzel

Exactly. The question for marketers is not if this will affect you, but how fast, where you’re exposed, and how you can turn it into an advantage rather than just a risk.

Christina

That’s it for today’s episode of The Customer Code. If this helped you make sense of the agentic commerce buzz, share it with a colleague in marketing, e-commerce, or product.

Andreas Munzel

And if your company is experimenting with AI agents in the customer journey, we’d love to hear what you’re trying and what you’re learning.

Christina

Thanks for listening.