“Agentic Commerce” is in fashion. The media talks about it. Investors are mesmerized by it.

Like many new concepts, “Agentic Commerce” has a definition that is a bit fluid. It can be roughly defined as a way of online shopping where AI agents autonomously shop on behalf of a human consumer. The AI agents can interpret human intent, discover product options, optimize for constraints (prices, features, delivery speed) and complete transactions.

For example, an often quoted use case is as follows: text or speak to an AI agent to “buy me a pair of running shoes under $100 that can be delivered tomorrow.” “Agentic Commerce,” thus, can reduce or largely eliminate human participation.

In the investment industry, analysts are anxious. “Agentic Commerce” is seen as the force that will disintermediate traditional e-commerce platforms. Because AI can complete entire transactions for consumers, going forward, consumers might skip e-commerce websites and apps entirely and only interact with AI.

However, I see it somewhat differently. My sense is that “Agentic Commerce” might not see massive consumer adoption anytime soon. And, if it is to be widely adopted, the process would transpire over a very long period of time, not anywhere near the same speed that consumers adopt AI chatbots.

These are the reasons why I think so.

(For the sake of clarity and focus, this article concentrates its discussion on consumer behaviors, not on back-end fulfillment logistics.)

Interfaces Are Not Created Equal

In the office, during work hours, it is not unusual that we see colleagues spend their spare time browsing e-commerce websites. Sometimes they need to buy something. Many times, I bet, they are just relaxing themselves and having a good time.

Think of another example: When was the last time you saw your colleagues, in their spare time, wanting to relax and have some fun, so they sought out a particular piece of enterprise SaaS software, launching it and clicking around? Really?

Imagine this: If you have unlimited resources and unbounded technologies, would you delegate someone (or an AI agent) to do your bookkeeping in the SaaS software you have? I guess you will say yes. Bookkeeping (and thus the usage of the SaaS software for bookkeeping) is work. For most people, it is not fun.

Now, imagine the following: If you have unlimited resources and unbounded technologies, would you delegate someone (or an AI agent) to eat a delicious meal on your behalf, to do a weekend half-day shopping trip on your behalf? Hmm…

Similarly, would you trade e-commerce sites that you are familiar with, you enjoy browsing, and you trust, for a text- or voice-based AI agent that handles the entire shopping journey for you, so you do not need to be involved in the shopping experience at all? Hmm…

A classmate of mine reminded me about the defunct “Amazon Dash Button.” Rolled out in 2015, the physical button helped consumers to instantly re-order specific household products by just pressing a button. In a way, it is the physical manifestation of “one-shot” AI shopping. Unfortunately, consumer adoption never took off in a meaningful way. To its credit, the “one-shot” button experience successfully reduced ordering friction. However, at the same time, it also reduced consumers’ confidence in the shopping journey: Did the order go through? How many units did I just order? What was the price that I paid? When would the order arrive? Are there newer or different versions of the same products that I should also take a look at? Are there any promotions going on of similar products that are selling at much lower prices? By 2019, Amazon shut down the program. The front-end interface cannot be easily abstracted away.

It is safe to argue that we humans want to have fun. And for most people, shopping is fun. For e-commerce companies, their front-end interface is part of the final product. Often, these front-end interfaces are fun to use. No one wants their fun activities to be taken away.

So, that is why I think interfaces are not created equal. For interfaces that are fun for humans, these interfaces have more staying power.

Humans Like Having Choices

Humans like having choices. This innate behavior is often reflected in the external world we live in. The grocery industry is a good example. Grocers, by and large, sell substantially the same things. You can get bananas, milk, meat, fresh produce, from almost all grocery stores, at roughly the same price points. A casual observer might conclude that one or two grocery stores would be enough, because we can already get what we need. In reality, however, no grocers seem to be able to own an entire market outright.

• The largest grocer in the U.S. is Walmart, and in its home state Arkansas where it has been around for over 60 years, Walmart’s market share is ~50%.
• Kroger, the second largest grocer in the U.S., has a market share of ~50% in its home city Cincinnati despite it having been operating in Cincinnati since 1883.
• Publix is a beloved grocer brand by both its employees and customers; it has most of its stores in Florida, where it is headquartered at; with so much consumer “love,” in Florida, Publix only controls a market share of 40-50%.
• It is the same picture in the online world as well. Instacart has disclosed, and I quote, “The average consumer on Instacart shops, on average, from five retailers. Instacart+ members shop at almost twice as many.”

We as consumers love having choices. “Agentic Commerce,” in its current form, risks taking choices away from us. In a “one-shot” prompting (thus “one-shot” shopping), AI strips away all our choices. Even in a conversational style process, an AI interface probably can only present 5 to 10 products in one go. The choices are too limited versus us humans wandering around by ourselves (online or offline). The former feels constrained and mechanical. The latter is liberating and fun—it is an adventure!

Early Adoption Signs

Early signals seem to support the view that, unlike chatbot and its rapid consumer adoption, “Agentic Commerce” is not seeing signs of rapid adoption.

Look at OpenAI. In September 2025, OpenAI launched “Instant Checkout,” allowing users to complete transactions without ever leaving the chat interface. In March 2026, it was reported that OpenAI scaled back its shopping plans for ChatGPT. Quoting that article, “OpenAI staff had realized that while ChatGPT users were researching products to buy in the chatbot, they weren’t using the chatbot to actually help them make purchases” (link). Then, on March 24, 2026, OpenAI published an article, which read, “we focus our efforts on product discovery” (link), signaling that OpenAI realizes its role in commerce is probably more in discovery, and less in automating away the entire commerce process.

Look across the Pacific. Consumers in China have a track record of rapidly adopting new technologies: superapp, e-commerce, ridesharing, food delivery, short-form video, etc. Early in 2026, ahead of the rest of the world, Chinese consumers were embracing agentic AI, rushing to install OpenClaw.

Nevertheless, in China, there are hardly any signs of widespread adoption of “Agentic Commerce.” Consumers continue to use their e-commerce apps. Consumers are not switching at scale to use AI chatbots nor AI agents to do shopping for them.

What is even more telling is this observation: After the initial wave of OpenClaw installation, now we are seeing a second wave of OpenClaw uninstallation. Consumers are rushing to uninstall their AI agents. This development does not bode well for the near-term development of “Agentic Commerce.”

At least, so far, there are little signs of wide adoption of “Agentic Commerce.”

Summary

For over 95% of the time that we humans have existed on this planet, we were hunter-gatherers. Perhaps, we are hardwired to like activities such as wandering around and gathering many choices.

Text or speak to an AI agent to “buy me a pair of running shoes under $100 that can be delivered tomorrow.” To do it in one-shot, without us participating, is that something that we really want? I would guess most people would love to see these shoes (on screen or in person), comparing various colors, seeing different designs, and checking out by themselves.

“Agent Commerce,” in its current form, takes away our front-end experiences and takes away our choices. The exploration is gone. The adventure is gone. It makes life boring. Instead of empowering consumers, it risks stripping away human agency. From this lens, we should not be surprised by the fact that there are little signs of consumers embracing “Agentic Commerce.”

Given how slow consumer behaviors change, even if we have a form of “Agentic Commerce” that consumers like, it would take a long time for most consumers to change their behaviors.

(END)