Visa announced Intelligent Commerce Connect recently as a new system built for AI driven shopping. The company refers to it as an entry point for businesses that want to connect to agentic commerce, whether they are building AI agents themselves or just selling products through it.
The system connects through the Visa Acceptance Platform and gives businesses one way to plug into AI-driven transactions. Visa said more consumers are starting to use AI agents to search for products and complete purchases, which has created the need for a much simpler way to connect payments to these systems.
The platform is already in pilot with partners such as AWS, Diddo, Highnote, Mesh, Payabli and Sumvin, with more partners expected later this year. These early tests give Visa and its partners time to refine how agent payments work in real environments before a wider release.
Mandy Lamb, Head of Value-Added Services for Europe at Visa, said, “Commerce is changing fast, with AI agents starting to play a real role in how people shop and buy. That shift can only succeed if people feel confident letting technology act on their behalf.” She added, “Payments depend on trust, security and choice. Intelligent Commerce Connect provides that foundation for agent-led commerce, enabling businesses to innovate without compromising control or protection.”
Visa’s update comes exactly when more companies test AI agents that can help with the ecommerce journey, on behalf of users. The company is placing itself as a main influencer of how payments work in this new setup.
What Does Intelligent Commerce Connect Actually Do?
Basically, the platform gives businesses one connection point to accept payments from AI agents, instead of building separate systems for each protocol or agent type. Visa said this connection runs through its Acceptance Platform, so companies are able to plug into agent-led transactions using a single setup.
It covers the full transaction process needed for agent purchases, covering everything from secure payment initiation to spend controls. This allows an AI agent to complete a purchase using stored credentials, guided by limits set by the user or business.
Visa also confirmed that the system works with different card networks and not only Visa. This gives agents more payment options, which makes it easier for merchants and developers to take part without having to stick to just one network.
The platform works across shared agent standards that act as a common language for transactions. These standards define how payments are requested and approved, with examples such as Trusted Agent Protocol and Machine Payments Protocol.
It also makes merchant catalogues accessible to AI platforms. Product data such as pricing and descriptions can be read directly by agents, allowing users to search and complete purchases inside an AI interface rather than moving to a separate website.
Visa said it will manage the technical flow of transactions and PCI compliance for companies processing payments on behalf of merchants. This cuts down the work needed from businesses that want to accept agent-led payments but do not have advanced payments systems in place.
Who Else Is Building Agentic Commerce?
Lily Varon, Principal Analyst at Forrester, wrote in the piece “Agentic Payments in B2C Commerce: Where We Are Now”, that announcements in this space have continued to arrive “left and right” since mid November 2025, pointing to a fast build out of tools that let AI agents take part in shopping and payments. She also perfectly summed up the different companies who are venturing into agentic commerce:
Gap announced a partnership with Google to support shopping on the Gemini app, although the capability was still in testing at the time of writing. Gap’s chief technology officer told CNBC that Google’s Universal Commerce Protocol gives merchants more control over the shopping experience than OpenAI’s Agentic Commerce Protocol.
More from News
- Do Oil Price Shocks Make EV Investments More Attractive?
- Global Markets Show Early Signs Of Recovery After Trump’s Iran Ceasefire – But Can The World Breathe Easy, Or Is Volatility Just On Pause?
- Greece Is The Next Country To Ban Social Media For Under 15s
- World IoT Day 2026: Is The UK Ready For The Next Phase Of IoT?
- Why Are Digital Nomads Leaving The UK, And Where Are They Going?
- Are Hiring Tasks Turning Into Unpaid Labour For Companies?
- Artemis II Is NASA’s Mission, But Is Control Of The Moon Quietly Shifting To Private Companies?
- Is Your CV Alone Enough To Get Hired Today? Experts Weigh In
Google has also expanded its Universal Commerce Protocol with new features designed for agent-led shopping. These cover things like an AI-specific shopping cart and identity linking for loyalty systems, and the company said partners such as Stripe and Salesforce plan to bring this protocol into their own products.
Walmart has taken a different route by embedding its Sparky AI agent into platforms such as ChatGPT and later Google Gemini. In a Wired interview, the retailer said it recorded “three times lower conversion rates for the selection sold directly inside the [ChatGPT Instant Checkout] chatbot than those that require clicking out.” It also said data from its pilot showed users convert at roughly 70% of the rate seen on Walmart.com.
Amazon is working on its “Buy For Me” agent, which can complete purchases on behalf of users using stored account details. The company has also enabled product feeds so merchants can send inventory into its system, allowing the agent to act on that information during a purchase.
Stripe and Tempo introduced the Machine Payments Protocol and opened early access. Stripe said, “MPP provides a specification for agents and services to coordinate payments programmatically.” This points to a system where AI agents can complete transactions without direct human input.
Mastercard, working with Santander and PayOS, said it had completed “Europe’s first live end-to-end transaction executed by an AI agent” in a controlled environment, although it added that this does not represent a commercial rollout.
Varon’s article speaks to a space where businesses and platforms are testing different ways to connect AI agents to checkout systems, each taking a slightly different route on how much control the agent should have.
What Does The CMA Protect In Agentic AI?
The UK government said existing consumer protection law already applies to AI systems, whether decisions come from humans or algorithms. Under the Digital Markets, Competition and Consumers Act 2024, businesses must not mislead consumers or use design choices that push them into decisions against their interests.
The Competition and Markets Authority said enforcement has already covered digital interfaces that use misleading tactics such as false urgency claims. It explained that the same standards apply when AI systems present options or act on behalf of users.
The government said businesses are responsible for how their systems behave. This covers training systems to follow consumer law, checking performance in real settings and fixing issues where harm may occur. Human oversight is expected as systems operate.
It also said transparency is needed, so consumers understand how systems perform and what their limits are. Accountability applies as well, which means companies must take responsibility for outcomes created by their systems.
The guidance also speaks on supporting systems such as data access and digital identity. The government said agentic systems depend on accurate, user approved data and strong identity checks to act safely. Without this, risks such as fraud, errors and lack of consumer choice could grow.
The CMA also mentioned that companies that build these principles into their systems early are more likely to build trust and compete on the quality of outcomes delivered to consumers.
