- Agentic commerce is growing fast, but fully autonomous checkout is still rare.
- Not all purchases will be delegated to agents. Emotionally driven, experience-first brands have a natural buffer against full automation.
- The brands that win in an agentic world will be the ones that give algorithms the right context, and that context lives at checkout.
There's a ton of hype out there about what AI agents will mean for ecommerce.
But if you zoom out, agentic commerce will eventually be just another type of marketplace: a place where consumers can browse and buy from different vendors easily. We all know this model from Amazon, from eBay, from Temu.
The difference is in the way you search for and discover products, and how you transact. What used to be a window shopping experience — where you click through product pages and then drill down into the details — will eventually be fully agentic. A user will go to their AI agent and say, "Help me find a backpack that’s this color and this price," and the agent will go do it for you.
That said, we don’t see agentic ecommerce volume coming in yet. The discovery traction is real. But the agent being the one purchasing, doing the whole funnel for you? Not there yet.
The infrastructure is coming. And when it does, the brands that have optimized their checkout experience will have a massive head start.
Not Every Purchase Will Be Delegated to an Agent
Agentic commerce will work for certain types of goods but not others.
If you look at the Shopify ecosystem, most brands are not selling basics. You don't buy toilet paper or toothpaste, and you don't buy show tickets, flights, or hotel stays from these stores. They're not necessarily the cheapest. They’re not the fastest to deliver, either.
Shopify brands are focused on giving shoppers a unique experience. There's an emotional element to the way you make a purchase. When someone buys a candle from an indie brand, or a limited-edition sneaker, or a piece of jewelry, they're not necessarily thinking about efficiency. They're feeling something. An agent can't replicate that impulse, and we wouldn't want one to take that experience away from us.
Agents will compete with the Booking.coms of the world, with the Airbnbs of the world, the Amazons of the world — all of these marketplaces. Agents will be able to look at multiple vendors and make a purchase based on your guidelines.
They'll also likely replace subscription purchases. Instead of a rigid 30-day cycle, you give your agent a task: "I love this cream. Buy it every 30 days, but ask me first." The agent pings you, you say "Remind me in five days," and it adjusts. That's going to replace the subscription model as we know it. And it'll be better for consumers.
For many purchases, humans will still be the ultimate decision-makers. Which means brands need to optimize for humans.
The Agent Is the Shopper, But You’re Still the Buyer
As a platform focused on optimizing every checkout session, we’re interested in the human intent behind an agentic purchase. Even if you send an AI agent to buy something on your behalf, you need to give it guardrails.
If you're buying a birthday gift, the delivery time frame is probably going to be the most important consideration for you. You don’t want to miss the birthday.
Buying something else that's not so time-sensitive? You're probably going to be looking at price, not delivery windows. An agent won't know that without the proper context. It needs more information to go out and represent your needs.
As a checkout optimization platform, I don't care, theoretically, who's clicking the button, an agent or a person. All I care about is the intent behind whatever action is being taken.
When we started PrettyDamnQuick, we knew that it was at checkout — not at "add to car"” — where consumers were deciding whether to commit to the purchase. That hasn't changed.
There's a narrative right now that agentic commerce will kill the funnel. I have to say, I doubt it. When you make the purchase too easy, when there's no effort and no perceived risk from the consumer, conversion actually drops. There has to be that risk-reward play.
Context is Everything at Checkout
So what does autonomous checkout actually look like? It's about finding enough indicators during the checkout process to understand consumer intent, and then throttling the experience based on what you find.
Here's a simple example: most brands give 10% off to all first-time customers. But not all first-timers are the same. Someone who reached checkout in 17 seconds, never looked for a coupon, and went straight to payment — they're going to convert anyway. Don't give them the discount.
Take that 10% and give a low-intent customer 20% off if they buy in the next 30 minutes. Add urgency. The algorithm reclaims budget from high-intent shoppers and deploys it where it's actually needed to drive a conversion.
This algorithm we're building to drive autonomous checkout will play with different levers based on customer intent, within the guardrails that the brand will set.
Overall, your metrics will rise. But you need historical data to train the algorithm and a framework that can adjust checkout in real time — with holdout groups, with A/B testing. Because everything I'm saying is a great hypothesis, but it's worth nothing if I can't prove that the algorithm is actually converting better.
The brands that will win in an agentic world aren't the ones who hand everything off to the algorithm. They're the ones who give the algorithm the right context — about their brand, their customer, and what actually drives a conversion. That starts at checkout.
Liran Erez is the Co-Founder and Chief Product Officer at PrettyDamnQuick.
