Want to know a new viral trend on Tik Tok? It’s not six seven. (and if you don’t know what 6-7 means, that’s what the kids are saying these days. If you don’t know what it means, just call over your preteen or teenage kid. They will happily explain it to you.)

Something strange is happening to the world of shopping. There is a trend where girls are taking a picture of their face, sending it to ChatGPT and asking for products to make their skin better. It’s called the “ChatGPT GlowUp”. I guess the thought is - why go to my expensive dermatologist if I can get the same suggestions on ChatGPT?

Go ahead. I’ll give you a few moments to try it on ChatGPT.

Impressed? Confused?

We’re slowly opening the door to a new way of online shopping. We had migrated from Ebay listings to Amazon to social commerce, to short video to now, something called Agentic Commerce.

For some websites, traffic from answer engines is already around 10% and that number is growing. In my previous article, I explained what brands can do to get discovered by answer engines. This article will explain how the different layers of agentic commerce will work.

A few days ago, PayPal signed a partnership with ChatGPT to be the recommended digital wallet while shopping on ChatGPT. While most of us are uncomfortable handing over credit card details, they have found a way around it which is using our PayPal credentials.

Soon, we will have AI agents shopping on behalf of us. That’s Agentic Commerce: a world where AI agents act, negotiate, and transact on behalf of humans and businesses. Imagine this: your personal AI agent knows your skin type, tracks your favorite brands, waits for a discount, then buys your moisturizer automatically.

‍

What Is Agentic Commerce?

Agentic Commerce is the next leap beyond e-commerce and automation. It’s a system where autonomous AI agents make commercial decisions: researching products, comparing options, and executing purchases or payments. You simply give it a goal, like – I want better skin. And the answer engine will crawl the internet and suggest to you what products you can buy.

Unlike traditional e-commerce that requires user input, agentic commerce thrives on machine-to-machine transactions. It combines language models, APIs, and payment protocols to create a seamless loop of intent, decision, and payment.

But there’s a big problem. Our entire global payments infrastructure was built for humans, not autonomous AI agents. Today’s payment systems assume a human is directly clicking buy. When autonomous agent initiates a payment, it breaks this assumption and raises some questions.

How can we prove that a user gave an agent specific authority to make a particular purchase? How can a merchant be sure an agent’s request is accurate and not just a hallucination? And if there’s fraud involved, who’s accountable? The user, the merchant, the issuer, or even the AI model?

‍

The Three Nodes of Agentic Commerce

Consumers: The Personal AI Layer

Your AI agent learns from your conversations, preferences, and purchase history. It knows your favorite colors, your sustainability goals, and your spending limits. Soon, you’ll delegate recurring purchases or complex comparisons entirely to these agents.

Brands: The Machine-Readable Layer

For brands, success depends on being found, understood, and trusted by AI agents. This requires Answer Engine Optimization (AEO) clear, structured data that answers questions the way an AI parses them. Product metadata, customer reviews, and schemas will matter more than website design.

Payments: The Protocol Layer

Banks and payment networks are creating the connective tissue for this ecosystem. Visa, Mastercard, and Google’s AP2 all serve as communication protocols for autonomous commerce, ensuring transactions between human or non-human agents can occur securely and instantly.

However, a protocol is only as good as the community that uses it. As companies like Visa, Mastercard, Google start writing their own protocols, it will be something notable to watch as to how many companies actually start using it.

‍

There are already problems between answer engines and marketplaces

This past week, we saw the tussle between Amazon and Perplexity, where Perplexity users are using Comet, Perplexity’s AI browser to look for and buy products on Amazon. Amazon in turn prevents AI agent crawlers from accessing its products. You would wonder why it would do that since ultimately it is still a sale. Why prevent a sale? Actually, if this starts happening regularly, it is a detriment to the intra app marketing machine that Amazon has built up. Currently, Amazon ads represent $56.2 billion in advertising services revenue. Amazon ads is Amazon’s internal retail media network.

Agentic commerce can be a threat to all retail media networks since the goal of the network is to monetize its existing traffic. If traffic from answer engine websites to marketplaces start to increase and marketplaces truly become a back end database, then there would less and less use of formal retail media network within the marketplace.

You see, retail media networks were built on a simple premise: shoppers browse, linger, and click. The longer they stay within a digital storefront, the more opportunities there are to advertise to them. Amazon, Walmart, and Instacart have turned this attention into a powerful profit engine, selling brands access to prime digital shelf space. Yet in a world where AI agents handle shopping decisions, the foundation of this system begins to erode.

When an intelligent agent makes a purchase, there are no eyes to attract, no clicks to win. The agent doesn’t pause to admire a banner ad or scroll through sponsored listings. Instead, it moves directly to optimization - comparing prices, analyzing delivery times, weighing reviews, even considering ethical sourcing. In this process, persuasion gives way to precision.

The retail media network’s greatest asset—its ability to target customers based on behavioral data - loses value because the agent already knows the consumer better than any algorithm designed for ads. It knows dietary restrictions, preferred delivery hours, and brand loyalties, making ad targeting redundant.

The result is a shift in the economic model. Retailers may no longer profit from advertising impressions or sponsored search but instead from data access, premium APIs, or strategic partnerships with the very agents that now control the flow of commerce. Agentic commerce doesn’t destroy retail media overnight, but it changes its purpose. What once sold attention will need to sell intelligence.

‍

Implications for Stakeholders

For Consumers:

In the future, we will be able to get our own highly intelligent shopping concierge. I’m sure there will be start ups working on exactly that to pull away from having all consumers shop directly within ChatGPT. This shopping concierge can “wait in line” for you for tickets that are hard to get (think Taylor Swift concerts or big music festivals). It can help negotiate pricing for you, handle returns, and handle getting items that are out of stock. Decision fatigue disappears, replaced by intelligent delegation. The consumer’s role shifts from choosing to training their AI.

For Brands:

Your digital shelf is no longer designed for humans. It’s designed for machines. Brands must structure data for conversational discovery, add schema markup, and train AI copilots on product knowledge. Large investments will be put towards AEO services and content production to increase the amount of traffic that is coming from answer engines.

For Retailers:

The shopping cart is becoming invisible. Retailers should focus on trust signals, authenticity verification, and seamless back-end integrations to support AI-initiated purchases. This means potentially entering formal relationships with answer engine companies to allow for smooth payment and logistics of transactions.

For Banks and Payment Networks:

The challenge lies in creating interoperability and compliance between human and AI agents. Banks must rethink fraud detection, KYC, and liability for non-human actors. If there is a dispute in a transaction, banks and issuers must provide a traceable audit trail.

In summary, this is the role that each section of agentic commerce plays and how it will change in the future.

‍

Risks and Regulation

‍As agents gain purchasing power, new risks emerge. Who is liable if an AI overspends or gets tricked by synthetic data? What happens when fake agents infiltrate marketplaces?

Regulators will need frameworks for AI identity, consent, and auditability. Financial systems must develop “Agent Authentication Protocols” to distinguish between human and non-human participants in a transaction.

‍

Conclusion

Of course, the speed with which this will all be adopted will be a slow and steady one. Consumers must feel the convenience for them to want to change their shopping habits. Consumers ultimately will buy things in the laziest way. If agentic commerce poses too many mistake purchases, improper audit trails, or is blocked by retailers (Amazon and Perplexity’s example) then this phenomenon will take a while to take off. So far, there’s too little data to point in either direction, but what’s true is product related searches are happening within answer engines. The way of shopping is changing under our feet.