sync 2026 | How Shopify is helping merchants become transaction ready in the AI era

AI is changing how people discover products, but what happens when discovery becomes purchasing?

As AI-powered shopping moves from recommendations towards building carts and completing transactions, merchants need to think about more than simply being visible. They need to make sure their businesses are ready to transact.

At sync, we welcomed Ben Homer, Solutions Engineering Lead at Shopify, to explore what the changing purchase moment means for merchants.

During his session, Ben looked at how the relationship between shoppers, merchants and AI agents is evolving, why product data and trusted information matter, and how Shopify is helping merchants prepare for a new era of commerce.

From discovery to transaction

Ben opened with a story from 2015… A customer walks into a fashion store with a shirt they bought online. It doesn't fit, but the right size is sitting on the shelf. The assistant finds the customer's account and sees the order, but tells them they need to return it through the website because it was bought online. The customer is in the store. The product is in their hands. The right size is ten feet away. Yet the systems can't connect the journey.

In 2015, that was normal, today, it sounds ridiculous. That's because unified commerce became the baseline. What was once an edge case became something customers simply expect.

The question is no longer whether customers will use AI to discover products. It's what happens when they expect an agent to handle part of the purchase journey — or eventually all of it.

The customer hasn't changed. The reader has.

For a shopper, choosing a product can be instinctive. They can look at a jacket, feel the material and decide whether it feels right. An agent can't do that. It has to reconstruct that judgement from the information a business makes available. That makes product data increasingly important.

Ben highlighted two types of questions an agent might receive. A brand query, such as: “Find me a wax jacket from this brand.” Here, the shopper has already chosen the brand. Accurate data helps the agent represent that choice correctly. Then there's a constraint query: “Find me three child seats that fit across the back of a family car.” Here, no brand has been chosen. Complete product information is what gets a merchant into consideration.

In both cases, the principle is the same: make the product, promise and purchase clear enough that the agent doesn't have to guess.

We're still early

There's a lot of talk about fully autonomous shopping, but Ben was careful to distinguish the headlines from where the market is today.

He described four stages:

• Ask: shopping becomes a conversation.

• Build: the agent compares products and builds a cart.

• Confirm: the shopper approves the important choices and the agent transacts.

• Delegate: the shopper sets the outcome and lets the agent take over.

Most of the market is currently somewhere between Ask and Build.

AI discovery is already creating value, but completing a purchase inside a conversation still depends on the channel, buyer, geography and merchant eligibility. Shopify's Q2 data shows AI-attributed orders represented 0.22% of measured online orders, up from 0.14% in Q1. AI-referred traffic and orders from AI-powered searches have also grown significantly year on year.

The share is still small, but the direction is clear.

For merchants, the message is simple: don't try to predict exactly when autonomous commerce will arrive. Build the foundations that allow you to participate as the channel develops.

Four tests for transaction readiness

Ben's framework for getting ready came down to four questions:

• Can the agent understand what you sell?

• Can it trust the promise?

• Can it complete the purchase without breaking your business rules?

• Can it support the customer afterwards?

1. Understand

The first step is making product information complete and accurate. Shopify Catalog gives participating AI channels a merchant-supplied source of product information, including details such as price, availability and variants. That's important because a missing category, vague title or absent specification can mean the right product never appears in an answer.

The goal is simple: give the agent enough information to answer the shopper's actual question.

2. Trust

An agent also needs to know whether it can trust the promise behind a recommendation.

That means more than a keyword match. It needs confidence that the product is suitable, the merchant can fulfil the order and the transaction is unlikely to create a bad outcome.

Shopify can use first-party commerce signals around areas including fulfilment, risk and safety.

Trust, in this context, comes from evidence in the commerce system.

3. Complete

Getting a customer to a cart is only part of the challenge. The checkout still needs to preserve the rules that make the business work. Discounts, tax, duties, shipping, gift cards, validation and payment options are all part of the real transaction. A generic payment flow might complete the payment, but if it loses those rules along the way, the result could be a failed sale, broken promise or margin leak.

Completing payment is easy. Preserving the business logic behind it is the hard part. Shopify is working with partners including Google on UCP, the Universal Commerce Protocol, creating a shared way for agents and merchants to connect across products, carts, checkout and orders.

4. Support

The purchase doesn't end at payment, customers still want to know where their order is, when it will arrive and what the returns policy allows. Those answers need to come from the real order and the merchant's actual policies, rather than an AI model filling in the gaps.

Shopify Knowledge Base gives merchants a controlled source for information such as delivery, sizing and returns helping agents answer from the merchant's own information - the merchant remains at the centre of the relationship after payment.

What can merchants do now?

The good news is that becoming transaction ready doesn't need to start with a huge AI project.

Ben's advice was to start with four practical checks:

• Understand: Is your product data complete enough to answer real shopper questions?

• Trust: Can an agent find reliable information about delivery, sizing and returns?

• Complete: Does your checkout preserve the rules that make your business work?

• Support: Can customers get accurate answers about their order after payment?

None of this is particularly glamorous, but that's exactly why it matters.Readiness takes time, and the work is easy to postpone. The purchase moment is changing, but the fundamentals haven't. Customers still need the right product, a trusted promise, a seamless transaction and support afterwards. The difference is that increasingly, an agent may be helping them get there.

So the question for merchants is simple: if an agent asked for one of your products today, would it get the right answer and be able to complete the sale? If not, that's where the work starts.

Looking to understand your eCommerce operation?