By, Ray Fernandez, Espacio Media Incubator
Agentic commerce is on the rise, as AI agents become increasingly embedded in all types of online transactions, from traditional banking to online shopping.
In a recent report, Juniper Research forecasted that agentic commerce transactions would reach $3.5 trillion USD in 2031, up from 2026’s $8 billion USD market value – a jump of 43,240%. Driving this shift in how money moves are AI and tech firms, and banks and credit-card companies.
At the heart of agentic commerce are agentic commerce protocols, the backbone core engine that makes agentic commerce transactions possible.
While agentic commerce has only been around for a year, several protocols compete in the market today, including Anthropic’s Model Context Protocol (MCP), the Agentic Commerce Protocol (ACP), co-developed by OpenAI and Stripe, Visa’s Trusted Agent Protocol (TAP), Mastercard’s Agent Pay, and Google’s Universal Commerce Protocol (UCP).
Google’s UCP was developed with Shopify, Etsy, Wayfair, Target and Walmart, and endorsed by more than 20 others across the ecosystem, such as Adyen, American Express, Best Buy, Flipkart, Macy’s Inc., Mastercard, Stripe, The Home Depot, Visa, and Zalando.
With Google now announcing new UCP features, rollouts across Canada, Australia, and the UK, and integration of UCP for advertisers on YouTube, we spoke to experts to answer: what is UCP, how big will it go, and what tech features stand out the most?
Understanding UCP: Meet the experts
UCP is an open-source technical standard designed to enable agentic commerce, according to Akshar Prabhu Desai, software engineer at Google, speaking in a personal and independent capacity and not representing his employer.
“It acts as a universal language for digital commerce where AI agents can discover products, place orders, negotiate terms, make secure payments, and manage shopping carts,” he told Fintech News Media. “It bridges the gap between conversational AI agents and existing merchant backends.”
As a fintech AI pioneer and IIT Bombay alumnus, Desai has a proven track record of shaping the future of digital commerce. Before building card tokenization and virtual card standards at Google, widely implemented by other financial institutions, he guided breakthrough AI initiatives at high-growth startups like Rover and Revcontent.
His research interests include Fintech and AI, and his published work has been cited by major bodies such as the International Monetary Fund.
An April 2026 PayPal report found that merchants view agentic commerce as a requirement rather than merely an upgrade, with 91% of small businesses, 97% of mid-market merchants, and 99% of large enterprises being at least somewhat familiar with the technology.
AI agents are now impacting all points for the retail process, with companies like Effie AI creating an agentic retail execution platform for merchandising and stocking products, all the way to agentic AI shoppers who buy the products.
“Right now, every AI agent that wants to buy something needs a custom integration to every store it talks to,” said Jay Rungta, an engineering manager specializing in large-scale payments infrastructure, also speaking in a personal capacity.
Rungta currently leads AI deals and payment systems engineering at YouTube, and previously built real-time trading infrastructure at Millennium Management.
“That gets messy fast once you have hundreds of agents and thousands of merchants on the other end,” he added, “a protocol like this is basically a shared rulebook, so any agent can talk to any store without custom code for each one.”
The expert also explained that this pattern is not new, with card networks solving a similar problem decades ago – to facilitate any card to work at any store without special integrations required.
“If something like this catches on, it won’t be because of clever marketing,” said Rungta. “It’ll be because it removes real engineering pain for both sides.”
Meanwhile, Desai stressed the future of UCP looks promising because it involves collaboration with retail giants including Walmart, Target, and Wayfair.
And although the protocol is led by Google, it is open-source and now has strong technical involvement from hyperscalers like Amazon and Microsoft. “If agentic commerce grows, UCP will provide its very foundation,” said Desai.
Headless commerce and the tech behind it
Traditional e-commerce involves merchants managing complex applications, payment processing partnerships, and shopping cart management through a website and/or mobile apps, Desai explained
“Each merchant website looks, feels, and behaves differently, though conceptually they all provide the same features such as product discovery, shopping cart management, checkout and payment experiences, and shipping details,” he said.
“UCP has the potential to enable what I call headless commerce, where merchants can focus on their core competence – making excellent products at great value – and rely on UCP and AI agents to handle discovery, shopping cart management, and other activities without even presenting their own UI,” Desai added.
That shift is also underway independent of UCP. The global headless commerce market was valued at $2.1 billion USD in 2026, and is projected to grow at a compound annual rate above 20% through the early 2030s, as retailers decouple their front-end shopping experiences from back-end order and payment systems.
Much of that growth has been driven by the same pressure Desai described: merchants trying to deliver a consistent experience across an expanding number of surfaces. Analysts expect the broader headless commerce platform market to grow at close to 19% annually through 2035, driven largely by demand for API-first storefronts and modular commerce architecture.
Established headless platforms – commercetools, Shopify’s Hydrogen framework, Salesforce Commerce Cloud, and others – already let merchants separate storefront design from backend commerce logic. Commercetools, for instance, released a microfront-end toolkit in late 2025 that adjusts storefront components in real time based on device type and user behavior.
What UCP adds, in Desai’s framing, is a further step: merchants may not need to present a storefront UI at all, and can instead let AI agents handle discovery and checkout directly against their backend.
This thus represents a meaningfully bigger architectural leap than headless commerce as it exists today, and is part of why Desai and Rungta both describe payments and trust infrastructure – not the storefront – as the harder problem still to be solved.
The most exciting tech features of UCP
An early 2026 IBM Institute for Business Value study surveyed 18,000 consumers and concluded they are ready to take the next step in AI-assisted shopping, finding that use of AI apps and chatbots like ChatGPT and Google Gemini has surged 62% over the last two years – and even more among Gen-X (82%) and Boomers (92%).
The study also found that more than half (52%) of consumers are comfortable sharing their data, but 83% still expressed concerns over privacy, data misuse, and unwanted marketing.
Part of that trust gap stems from perceptions that AI agents make purchasing decisions fully autonomously. In reality, most online retail agentic commerce today – with the exception of financial and trading agents – still runs on an agent-human relationship model, in which agents assist consumers by searching for and comparing products and enabling seamless checkout, even though many are technically capable of handling the entire process on their own.
Given this, legacy infrastructure, compliance and regulatory demands, and companies’ ability across small, medium and large businesses alike to understand and deploy the tech stack become key priorities.
“One of the most critical aspects of commerce is payments,” said Desai. Due to complexity, risk, and the regulatory environment, this space generally is more conservative and slow-evolving compared to other aspects of tech.
“Agentic payments are one of the most interesting engineering problems to solve,” he added.
While UCP manages shopping, trust frameworks like ACP secure transactions. “Open protocols such as the Agentic Payments Protocol (AP2) address this, but what I am interested in is watching how legacy systems like credit card tokenization systems evolve to seamlessly allow agentic payments without compromising on security, privacy, and safety,” said Desai.
But that trust layer is also the most challenging engineering problem within this paradigm, according to Rungta. “Every time an AI agent buys something on your behalf, someone needs proof you actually approved it,” he said.
“It will be really important for anyone in this space to build this auth layer from day one, so every purchase carries proof of real consent.”
Another UCP feature that Rungta is excited about is the core checkout, which he said stays simple and stable while more complex features get added on the side, without breaking everything else.
“Sounds like a small detail, but in payment systems, that kind of discipline is usually what separates something that scales calmly from something that breaks every time you add a feature,” he said.
Opportunities in the agentic commerce supply chain
Asked what opportunities are opening up in the supply chain, Desai was succinct. “Procurement and B2B supply chains are complex and slow to work out,” he said, “once supply chains are established, they are hard to change.”
With AI agents becoming more common through UCP and other technologies, a distributed network of AI agents could make this significantly more dynamic and prone to reacting in real time to changes in the economy, he elaborated.
“For example, when a certain producer ends up with higher inventory than they desire, we could envisage a pool of liquidation AI agents being used to quickly help sell this additional inventory by offering discounts or targeting niche buyers.”
Because UCP acts as a universal language, Desai stressed, it lowers the integration friction between merchants and third-party logistics (3PL) providers. “Warehousing and shipping agents can dynamically bid on fulfilling a shopping cart’s contents, and if a carrier has excess capacity on a specific route, their agent can lower prices programmatically, winning the fulfillment contract instantly from the merchant’s automated backend.”
In sum: agentic commerce is evolving fast, and while the tech is already deployed and in the market, its foundations are still being built.
From proof of concept to scaling the success
“Every innovative technology needs a proof of concept that becomes popular and provides clear value to the wider economy,” said Desai. “Email was instantly successful in that sense, but blockchain has largely been a solution in search of a problem.”
The contrast is deliberate: email succeeded because it involved an obvious problem for everyone who used it, immediately and without needing to be explained. Blockchain, by Desai’s account, never found the same clear point of contact with ordinary use.
“Agentic commerce and UCP present a huge opportunity, and I am excited to see merchants and AI hyperscalers bringing early prototypes to consumers,” he added.
That optimism is notable given how much remains unresolved: trust, consent, the payment layer underneath it all. Excitement here isn’t a blind claim that those problems are in the rearview, but rather a conviction that the harder engineering work is worth doing because the demand side of this equation is no longer in question.
“Once that initial success is proven, the rest is just a matter of scaling that success,” said Desai.
In other words, scaling is the easy part once trust is established. And Rungta’s card-network comparison points to why that assumption might hold. Card networks spent decades building the invisible trust that now lets any card work at any store; and once that trust layer was in place, scale followed quickly.
The open question for UCP, then, is whether an equivalent layer can be built faster this time, given that the counterparty on each transaction may now be an agent acting on someone’s behalf rather than a person swiping a card.
Whoever solves that first won’t just be building a checkout flow. They’ll be setting the terms for what “consent” means in a marketplace where the buyer might not be human. Such is the foundation UCP is racing to become.
About Author:
Ray Fernandez is a journalist with over a decade of experience reporting on technology, finance, science and natural resources. His work has been published in Bloomberg, TheNextWeb, TechRepublic, The Sunday Mail, and eSecurityPlanet, among others. He is a contributing writer for Espacio Media Incubator.
