What is agentic commerce — and why it's not just about shopping agents

When people hear "agentic commerce," they think shopping. AI that finds products for you, compares options, maybe even checks out on your behalf. And that's real — 58% of consumers have already replaced search engines with generative AI for shopping research, up from 25% the year before. Customers are using ChatGPT, Perplexity, and Google to discover, compare, and buy. Some of them are completing purchases without ever visiting a brand's website.

But if you stop there, you're seeing one layer of a much bigger shift.

The walls are coming down

Here's what's actually happening: the entire commerce experience is becoming conversational. The buying, the returning, the exchanging, the "I need help with my order," the "what should I get my sister" — all of it is collapsing into a single kind of conversation experience.

Customers already see it this way. They always have. When someone messages a brand about a return and then asks about a new product in the same breath, they're not switching from "support mode" to "shopping mode." To them, both are just talking to the brand. They don't care that one routes to the CX team and the other is supposed to live on the ecommerce site.

The systems have been fragmented — search tool, chat widget, returns portal, order tracker. But the customer's mental model never was. They've been waiting for the experience to catch up.

That's what agentic commerce actually is. Not a shopping assistant bolted onto your website. It's the convergence of commerce and service into one AI-powered experience that works the way customers already think — as a single, continuous relationship with the brand.

Think store associate, not chatbot

The best analogy isn't a chatbot that can also sell. It's the best store associate you've ever encountered.

A great store associate knows you. They remember that you run a half size large in that brand. They know you bought a jacket last month and you're probably going to need something to layer under it. When you ask about a return, they don't punt you to a different counter — they handle it, and while they're at it, they mention that the item you're exchanging for just came in and your size is in stock.

They help you, they know you, and they sell to you, all in one conversation. That isn't three separate functions. It's one experience delivered by someone who has enough context to connect the dots.

Online shopping has never had this. The internet scaled commerce infinitely — with one-click purchasing, next-day delivery, and endless selection — but it stripped out the relationship. The experience of being a customer online has been, for most people, a solo activity. You search, you scroll, you guess at sizing, you read conflicting reviews, and you hope for the best. When something goes wrong, you enter a completely different system to get help.

AI changes that. Not just because AI is smarter than a search bar, but because AI can hold context across the entire relationship and act on it. The store associate model, running everywhere the brand shows up.

The revenue sitting inside conversations brands are already having

Every service conversation is already a commerce moment that most brands leave on the table.

A customer reaching out about a return? That's someone you could convert to an exchange and a better-fitting product. A customer asking about their order status? They're engaged, they're paying attention, and they might be interested in something that complements what they bought. A customer whose subscription is about to renew? That's an upgrade conversation waiting to happen.

This isn't about turning support into a sales pitch. Nobody wants that, and the brands that try it will lose trust fast. It's about the AI having enough context to recognize a natural moment and act on it the way a good associate would. "You're returning the medium? The small actually runs true to size in that style — want me to swap it instead?"

This is efficiency and revenue. Resolution and relationship. Cost savings and customer value. For too long, the industry has treated these as tradeoffs — invest in AI to cut costs, or invest in humans to protect the experience. But you can refuse that framing. The same AI conversation can resolve a problem, save a customer, and generate revenue. And it's already happening in the service organizations that have figured out how to give their teams and their AI full customer context.

Salesforce research shows 91% of service organizations now track revenue generated by their service teams, up from just over 50% in 2018. The capability is there. What's been missing is the infrastructure to do it consistently, across every conversation, at scale.

What your AI actually needs to make this work

There's a baseline version of this that the market is converging on. AI that's available 24/7, across every channel, that can resolve common questions and take basic actions. That matters and it's a real improvement over decision-tree chatbots and phone queues. But it's just the starting line.

What separates the starting line from something that actually builds loyalty over time is whether the AI accumulates understanding.

Being available isn't the same as being useful in a way that earns the next visit. An AI that's always on but starts from scratch every time is a really helpful stranger. It can solve your problem in the moment, but it doesn't get smarter about you.

The difference is compounding knowledge. An AI that knows your sizing preferences from last month's exchange, that learned you prefer warm layers that aren't bulky from a product comparison in October — that's a different experience from a competent one that meets you where you are every time but doesn't actually know you. That context doesn't just make the current conversation better, it makes every future one better — and that's what brings customers back.

This is an architecture problem, not an intelligence problem. If your customer data lives in cases scattered across systems, the AI reconstructs context every time, no matter how capable the model is. If it's organized around the person — every conversation, every purchase, every channel, one record — the AI already knows who it's talking to. That's what makes the experience compound.

The loop

Here's the vision for where this goes.

When AI has full customer context and operates across both pre-sale and post-sale, something interesting starts to happen: post-sale feeds pre-sale. Not in a theoretical way. In a concrete, measurable way.

A return teaches the AI about sizing preferences. Next time the customer browses, the AI proactively flags fit notes. A product comparison teaches the AI about use cases — this customer shops for function, not fashion. Next time, the recommendations reflect that.

Every post-sale conversation becomes the starting point for the next pre-sale moment. The knowledge accumulates. The experience gets better each time. And the customer keeps coming back — not because you're marketing harder, but because the experience is so tailored that going somewhere else would mean starting over.

That's the loop. And when it's running, loyalty isn't something you buy with discounts or points programs. It's something you earn because every conversation makes the next one smarter.

What agentic commerce actually includes

If you're thinking about agentic commerce as a strategy, here's how to see the full picture. Three dimensions that go beyond the "AI shopping assistant" framing.

The full journey, not just pre-sale

Pre-sale and post-sale on one surface, powered by the same AI, with the same context. When a customer moves from browsing to buying to needing help with an order to shopping again, the AI doesn't reset. The experience is continuous because the underlying data is continuous. This is where the convergence of commerce and service gets real.

The whole site, not just the chat widget

Agentic commerce doesn't live in a chat bubble in the bottom-right corner. It's an intelligent layer across the entire brand surface. The AI sees browsing behavior, understands intent, and acts — surfacing the right product card, offering a size recommendation at the point of hesitation, nudging at the moment of cart abandonment. The whole website becomes conversational.

Beyond your own site

Customers are discovering products on AI surfaces that brands don't control — ChatGPT, Perplexity, Gemini. That's a real channel, and it's growing fast. Brands need to show up in those conversations through emerging commerce protocols. But showing up on third-party surfaces is discovery. The real goal is making your own experience so personalized, so contextually aware, that customers come to you directly. Get discovered there, build the relationship on your own terms.

What comes next

This is the first in a series on agentic commerce. We're going to break down the pieces that make this real:

• How to think about trust when AI is selling — the human element, graduated automation, and what happens when the stakes of a conversation are high.

• How emerging protocols are reshaping the way brands show up on AI surfaces — and what that means for owning the customer relationship.

• How to build your brand-owned experience so well that customers choose it over the third-party alternatives.

The market is still early in figuring out what agentic commerce means. Most of the conversation is about AI that helps customers find and buy products. That's part of it, but it's the smallest part. The bigger opportunity is in the full picture: commerce and service converging, every conversation compounding what the AI knows about the customer, and that knowledge driving the kind of loyalty that doesn't depend on discounts to sustain itself.

The brands that build for this will generate revenue from conversations they're already having. The brands that don't will keep treating service as a cost line and leave that revenue on the table.