Pricing

Documentation

Case Studies

Blog

About Us

Get Started

Table of Contents

The Signals: Big Tech Is Building Toward Agents

The Gap: Checkout Is Still the Unsolved Core

The Solution: Rye’s Universal Checkout API

The Future: Reliable Checkout from Any Merchant

Big Tech Is Leaning Into AI Shopping. Why Can’t They Make It Happen?

Arjun Bhargava

Co-founder and CEO @ Rye

Aug 5, 2025

Huge tech companies have failed. Rye has succeeded. Here's how we made agentic commerce work end-to-end on any e-commerce site.

None of Big Tech and generative AI’s marquee names have solved the hardest piece of agentic shopping: programmatic, universal checkout.

Google, OpenAI, and Perplexity, and others have billions of dollars and the world’s best engineers at their disposal, and yet the industry’s dirty secret is that “AI-powered shopping” isn’t actually powered by AI. Chances are a human behind the scenes, somewhere in the world, is manually entering your credit card info and confirming a purchase on your behalf.

This process is slow for consumers and expensive for businesses, so AI shopping experiences mostly don’t bother, leaving consumers to manually click through to buy. It’s the Breezewood of the internet, an infrastructure gap that needlessly disrupts an otherwise smooth journey.

Rye closes the gap by making agentic commerce truly, reliably agentic.

None of Big Tech and generative AI’s marquee names have solved the hardest piece of agentic shopping: programmatic, universal checkout.

Rye closes the gap by making agentic commerce truly, reliably agentic.

None of Big Tech and generative AI’s marquee names have solved the hardest piece of agentic shopping: programmatic, universal checkout.

Rye closes the gap by making agentic commerce truly, reliably agentic.

The Signals: Big Tech Is Building Toward Agents

A look at what some of these big players are doing reveals the limitations they’re encountering.

Perplexity has put AI-powered shopping into the spotlight, weaving recommendations directly into its core product and adding the option to click to purchase. But the company admits its “Buy With Pro” feature is limited, explaining that if the option isn’t available it’ll direct you to a merchant’s website. It’s proof that automated shopping remains unsolved: At this rate, we may get flying cars and jetpacks before we can get a robot that can buy you a pair of Birkenstocks off Zappos.

OpenAI has announced shopping within ChatGPT, starting with Shopify. This is a natural first step, but also a constraint. Either the recommendations must be artificially limited to chosen ecosystems or the buyer referred to a merchant to manually complete the sale.

Shopify itself is leaning hard into the agent opportunity. Its merchant network is vast, but not universal; it has a 29% share of U.S. e-commerce activity. The open web, where millions of products live, still sits outside its reach.

Google, meanwhile, has decades of dominance in product search, with an enormous product legacy in its Google Shopping price comparison product. Its recent AI Mode shopping experience has fancy tricks like virtual clothing try-ons and LLM-powered recommendations. But checkout still happens somewhere else — usually on a merchant’s page, outside the AI experience. The company may be hamstrung by recent government action, hesitant to build new features in-house that might make regulators think it’s trying to own too much of a transaction. (Just our guess.)

The biggest players are validating the promise of agentic commerce with their efforts. But as you can see, it’s not quite there. The world’s best engineers at Google, OpenAI, and beyond, have incredible feats of engineering. But they haven’t yet come up with a service that can programmatically broker a transaction on any given website quickly and reliably.

Why?

The Gap: Checkout Is Still the Unsolved Core

Checkout looks simple on the surface, but it’s where all the complexity lives:

The total price, including shipping and taxes, must be calculated in real time
Merchant sites’ layouts and order flows frequently change
Fraud detection systems block orders that don’t look human enough
PCI compliance and transaction responsibility scares off potential integrators
And latency matters, because people will not wait minutes to confirm their order when they can order faster by clicking through manually.

This is why even strong players struggle. Some competitors patch the gaps with humans behind the scenes—in some cases, for almost all orders. Others sidestep the calculation challenge by eating shipping costs because calculating them in real time is too difficult. Most limit their services to one or a few merchant platforms. None has solved the challenge of universal checkout.

Checkout looks simple on the surface, but it’s where all the complexity lives:

The total price, including shipping and taxes, must be calculated in real time
Merchant sites’ layouts and order flows frequently change
Fraud detection systems block orders that don’t look human enough
PCI compliance and transaction responsibility scares off potential integrators
And latency matters, because people will not wait minutes to confirm their order when they can order faster by clicking through manually.

Checkout looks simple on the surface, but it’s where all the complexity lives:

The total price, including shipping and taxes, must be calculated in real time
Merchant sites’ layouts and order flows frequently change
Fraud detection systems block orders that don’t look human enough
PCI compliance and transaction responsibility scares off potential integrators
And latency matters, because people will not wait minutes to confirm their order when they can order faster by clicking through manually.

The Solution: Rye’s Universal Checkout API

Rye’s early work in product discovery and aggregation gave us a deep understanding of how product data flows online. Recognizing the gap in the current landscape, we developed the Universal Checkout API, designed from the ground up to make execution programmatic, scalable, and safe.

It works on any site that allows guest checkout, not just with merchant partners, using just a PDP URL as input. It’s at least as fast as a human. It uses PCI-compliant tokenized payment credentials so neither the agent nor the integrator see this sensitive data. An LLM only has to figure out the checkout flow—including collecting tax and shipping info—once per merchant; on future runs, deterministic workflows are 8x faster and still able to call on AI to self-repair. Manual fallback is just that: the last resort to ensure our systems fulfill the promise when edge cases fail.

The Future: Reliable Checkout from Any Merchant

Industry analysts project agentic commerce will grow from $135B this year to $1.7T by 2030. But it won’t make that 10x jump by magic. Without infrastructure that can execute purchases reliably, universal agents will stall at the experiment stage.

That’s why we built Rye. To make programmatic checkout not just possible, but reliable. To provide the crucial connection that lets the next generation of commerce platforms scale. To turn stoplights into smooth ramps. We’d love to close the loop for your agentic shopping experience; here are the docs so you can get started.