What Really Is Agentic Commerce

In the commerce world, there is this buzz-phrase that is slowly becoming the epicenter of the industry: agentic commerce.

Since starting Kinect however, there have been over 50 definitions of agentic commerce that I've heard, so I decided to write down my own thoughts and definition of it. I'll also break down the nuances of the different “sides” of the industry and how it matters.

Agentic commerce is traditionally defined as autonomous AI agents acting as a user's digital proxy to research, compare, negotiate, and execute purchases.

If you look at my definition, you can see that agentic commerce doesn't outline a side, shopper or merchant. In fact, it doesn't even outline a specific change, problem, or solution. The reason is, I see agentic commerce as a shift of the entire commerce category because it impacts the entire ecosystem: shoppers, merchants, vendors, operators, and more.

How it changes shoppers

In the last year, we have seen a sudden shift in the way that consumers have started to find, research, and buy products. In fact, over 50 million queries a day on ChatGPT are shopping related. The age group of 15–38 averages one online purchase a week, of which over 50% are AI-assisted. All websites, especially e-commerce websites, have seen a dramatic increase in agentic traffic visiting their website: 11x in 2025 itself.

But does it really matter? Although there are queries and traffic, we don't really see transactions. It does, it matters a lot.

First, San Francisco. I think that SF catches trends ridiculously well. San Francisco fueled the .com boom, Uber and food delivery were launched in SF, and obviously AI grew fire in SF. Why? Because when the process is still ugly and duct-taped, SF doesn't care and engineers don't care. They hack around, and when they realize how awesome it is, they'll productize it in a way that makes it accessible for the rest of the world. That's where we are right now. People are running end-to-end purchases, companies are building infra for the future.

Second, and more important, AI purchases are also very hard to attribute, because most often the AI is involved in the top or middle of the funnel rather than the bottom. Someone will use ChatGPT to help find the best suitcase, but buy it during a sale 3 months later. Someone will find a product they like, ask ChatGPT about it, then come back to buy it the next week when the Friday paycheck hits. The last click will say “direct” or “email”, but arguably these were decisions made weeks prior using AI. These are agentic purchases that will never get the attribution line because it's near impossible. Roughly 40% of transactions are estimated to be AI driven already.

So it's not that there aren't transactions, it's that they are invisible.

Buyer agents

Here's how shoppers use AI.

• Discover: shoppers use ChatGPT, Claude, and Perplexity to find products in the first place.
• Research: they ask questions about products they're already considering. Fit, materials, care, shipping, returns, price.
• Decide: they ask the agent to compare finalists and make the call.
• Transact: some shoppers want the agent to just complete the purchase, end to end.

Most of the market is pointing its attention to discovery, while research is arguably where the most usage is.

How this impacts merchants

Commerce is defined as the buying and selling of goods and services. The above definition only covers buying. What if we were to think about seller agentic commerce? In the same way that shoppers will discover, research, and transact on products, why can't brands and companies have agents that go sell, advertise, and build awareness for them using autonomous AI agents?

I call these brand agents, seller agents, merchant agents. What brand agents look like in practice:

• Agents that act like sales reps on your own website and sell to every visitor, not just the ones who fill out a form
• Agents that answer buyer agents: when a shopper's AI asks about your sizing or your return policy, something on your side has to answer, correctly
• Agents that come up with new ads, test new angles, and optimize the funnel
• Agents that rewrite copy and generate new assets for the site
• Agents that enrich your catalog so machines can actually read it
• Agents that run customer research and sentiment analysis
• Agents that sell on other platforms: social, marketplaces, and the AI platforms themselves

In the last 3 months, I've been to 6 e-commerce related conferences, and agentic commerce has been one of the number one topics. But most of the time, the brands don't get past solving for discovery, AEO, or “How do I show up in ChatGPT”. But the issue is that showing up is such a small fraction of the shift. OpenAI's own research shared that most shopping conversations are related to research and information seeking, where intent is already expressed. We even saw the same in the conversations that Kinect agents have with shoppers on brand websites. Over 40% of them are research related, asking questions about a product they already like.

At the same time, a big piece of the conversation here is how brands can use AI to improve operational workflows and unlock capabilities that weren't possible before to drive revenue. Before AI, it wasn't possible to spin up thousands of ad angles and thousands of landing pages. It wasn't possible to generate images with the same fidelity as real photoshoots that take weeks to plan. It wasn't possible to create one-off experiences for every shopper and remember everything about them.

Buyer agents and seller agents will meet

I'm a big believer in the future of agent to agent commerce. I'll write a whole piece on the evolution to agent to agent commerce, but I do think that a non-marginal share of transactions will involve agents on both sides facilitating the conversation.

This also means that the meaning and actions behind “conversion optimization” change significantly. The next decade will be about creating long lasting systems that make your website both human and agent readable. It also means that instead of single actions, for the first time brands and agencies will be able to take hundreds of actions at a time, and let the data decide the path forward.

The current infrastructure

This section deserves its own article, but high level, there are a set of standards being created to make this future possible. Some of which are UCP, ACP, MCP, and agentic payment infra related to fiat, stablecoins, and more. The “rails” are being laid.

How this plays out

I've spent the last few months studying how the previous commerce shifts played out. There have been 3–4 major shifts we've seen:

• The shift from retail to online
• The rise of mobile
• The rise of social
• And now the rise of AI

The pattern across the past was repeated.

Buyers always change first. People started shopping online in 1998 and it was ugly, sketchy, but it caught popularity quickly. In 2010, websites didn't work on phones in the same way. In 2014, Facebook ads were a bit weird. But the pattern is there: when something is arguably a better paradigm and consumer behavior is shifting, commerce will change significantly too.

Some merchants died and some thrived. Borders, Toys R Us, and hundreds of other brands lost to Amazon. Small brands like Warby, Gymshark, and MVMT grew to new heights because they were first movers in ads. In every shift, there was a clear window of first mover advantage that compounded for the following years.

The vendors that won, and how. Honestly, it was the teams that could move the quickest that won. It's a bit of right place, right time, and it's a bit of luck. I wish the best for Kinect, but I know we are in the right place, and the right time.

More to come…

At Kinect we build brand agents: an AI sales rep that sells on your storefront to every visitor, the way your best rep would. If you're thinking about this shift, we'd love to talk.