Agentic commerce is real

Perhaps fasting makes me a bit zestier than usual, but the topic of agentic commerce needs a more sober conversation than the one currently happening. If ecommerce is not your main focus, you may not be aware that we are in the middle of a full on agentic commerce hype cycle. And if you follow it closely, you already know the conversation is mostly happening inside a room full of ecommerce people.

Let me explain what I mean, and why it matters for your brand.

What’s agentic commerce

Agentic commerce is a model where AI agents act on behalf of the consumer to discover, evaluate, and purchase products. Here is an example: you ask ChatGPT for advice on what clothes to pack for a trip to Tuscany in July. ChatGPT gives you a list, and also offers where to buy them. You click through and complete the purchase. You skipped Google search, browsed no brand websites, saw no Meta ads, no SEO-optimized landing pages, no retargeting.

From a marketing lens, you skipped the entire persuasion layer that brands have spent years or even decades building. In a human-driven journey, a brand competes for attention and trust. In an agent-driven journey, a brand competes for inclusion in the agent's consideration set, and then for the agent's preference logic. That is a structurally different problem to solve for. Not a shift in a game of ecommerce, but a different game entirely.

So why am I skeptical of the hype? Because real and overhyped are not mutually exclusive.

The case for real (aka scale)

The case for agentic commerce being a completely new model of ecommerce that all brands have to learn to participate in can be built on the sheer argument of scale.

There are trillions of investment being poured into, and deployed here. (‘too big to be allowed to fail’).

• Amazon invested $50 billion in OpenAI, with integrated agentic shopping as a core part of the thesis

• Walmart embedded its own AI agent, Sparky, directly into ChatGPT and Google Gemini, keeping checkout inside Walmart while using the AI platforms for discovery

• Shopify is turning agentic stores on by default for every merchant by end of March, meaning products surface inside ChatGPT, Perplexity, and Copilot whether merchants opt in or not

• Amazon's Rufus generated nearly $12 billion in incremental sales in 2025

• Morgan Stanley and Bain predict agentic commerce could represent 10% to 25% of the U.S. ecommerce by 2030

• McKinsey goes further, projecting a trillion dollars in U.S. retail revenue will flow through agentic ecommerce by 2030

Years at Goldman Sachs taught me deep respect for the rigor of institutional research. I also know that projections about transformative technologies made at peak excitement have a consistent track record of being right about direction and wrong about timing. My scepticism isn’t about the direction, but the timeline and the magnitude.

Capability and utility are not the same thing

Capability is the power or ability to do something. As I mentioned above, trillions are going into building infrastructure and capabilities of agentic ecommerce, part of the larger AI-everything shift.

And capability surely exists. Rufus can autonomously purchase on your behalf. Sparky can build a full shopping cart from a conversational prompt. ChatGPT can help find products. It’s still very buggy, clunky, and not intuitive. But every technology starts there.

The utility question is different. I keep thinking how AI (and agentic commerce by extension) is a huge shift we didn’t ask for because utility, usefulness is not clear at times. Are consumers actually delegating purchasing decisions to AI agents at the scale the projections imply?

Bain found that while 72% of U.S. consumers have used AI in some form, only 24% said they would be comfortable using it to make a purchase.

But a different Omnisend survey, conducted two months later, found 80% of U.S. shoppers open to AI handling final transactions.

These two numbers cannot both be right in any meaningful sense. It seems that humans are curious, but not committed, in the context of agentic commerce.

The Walmart Instant Checkout story is the most honest data point so far: Walmart, the largest retailer in America, partnered with OpenAI, the most powerful AI company, built the capability, tested it with real customers, and discovered that conversion rates through ChatGPT checkout were three times worse than through Walmart's own channels. They pulled the feature, rebuilt with Sparky embedded inside ChatGPT instead, keeping the AI platform for discovery and Walmart in control of checkout. It was the right pivot, time will show the success of it.

The echo chamber problem

When I scroll through LinkedIn posts about agentic commerce from ecommerce experts the responses are almost entirely from other ecommerce experts. The conversation is happening within the industry, not with the customer.

This is not a criticism of the people involved. Being at the frontier of your field requires paying close attention to what is emerging. But there is a difference between what practitioners are excited about and what consumers are ready to adopt. The history of ecommerce is full of capabilities that practitioners declared transformative and consumers adopted slowly, selectively, and on their own timeline. Shoppable social media was going to end the traditional retail funnel in 2019. It did not. Ecommerce was going to kill malls. It did not.

A word on the Jobs inversion

Steve Jobs said it the best: “You have to start with the customer experience and work backwards to the technology. You cannot start with the technology and try to figure out where to sell it.”

A significant portion of what is being built in agentic commerce right now is doing exactly the opposite.

What it means for you as a brand

Here I want to be careful, because calibration kind of cuts both ways.

Dismissing agentic commerce as hype and staying purely in the traditional model would be a mistake. But so would be to overhaul your entire go-to-market strategy in anticipation of agentic commerce taking over. I think the most sober, and healthy approach is to is to recognize where this is directionally going, without overreacting to timelines or abandoning what already works.

The practical move here is not to build an agentic commerce strategy, but to read your own signals so to say: does your best and most traffic still coming from traditional channels. What data populates in the Prompts section of Amazon ads. How many customers tell you they discovered your brand through LLM. The fundamentals of building a strong product, clear positioning, credible reviews, and consistent availability do not become less important in an agent-driven world.

So the practical posture for now is relatively simple: stay informed, avoid over-indexing on hype, and continue building a brand and operating model that is structurally sound. If agentic commerce scales as projected, those foundations will translate. If adoption takes longer, you have not compromised your business chasing a narrative that has not fully materialized yet.

Saludos,

Irina