Revolutionizing Commerce: The Emergence of Agentic AI with Visa and AWS

Transforming Commerce with Agentic AI: A New Frontier

Co-written with Sangeetha Bharath and Seemal Zaman from Visa.

In today’s digital landscape, the advent of agentic AI is revolutionizing how work is conducted across industries. By transitioning from traditional, manual interactions to more autonomous, outcome-driven workflows, agentic AI introduces intelligent agents that can reason, act, collaborate, and tackle complex tasks on users’ behalf. This evolution is reshaping sectors like travel, healthcare, banking, logistics, and customer service, enabling agents to research, plan, optimize, and execute processes with minimal human involvement.

The Shift in Payments

Similar transformations are emerging within the payments industry, as agentic commerce redefines the initiation, authorization, and completion of transactions. Gone are the tedious manual steps across multiple platforms; now, autonomous agents seamlessly handle discovery, decision-making, and secure payments in the background—a game-changing shift reminiscent of the eCommerce boom of the early 2000s. Just as digital checkout altered consumer expectations, agentic commerce is laying the groundwork for a more seamless, contextual, and intelligent payment ecosystem.

However, to facilitate this shift, developers need robust support to ensure secure and scalable implementation. Until recently, even advanced AI agents could assist users in planning and cart preparation, but actual transaction completion still required manual input. Without trusted, standardized integrations with existing payment networks, agents could not reliably initiate transactions from start to finish.

AWS and Visa: Pioneering Agentic Commerce

To tackle these challenges, Amazon Web Services (AWS) and Visa have joined forces to help enterprises develop agentic payment applications. In April 2025, Visa rolled out Visa Intelligent Commerce, enabling developers (and even non-developers) to connect directly to Visa’s payment network using straightforward natural language commands. This initiative supports network-agnostic flows and establishes secure communication channels between agents and merchants through Visa’s Trusted Agent Protocol.

Excitingly, AWS will facilitate this transformation by utilizing Amazon Bedrock AgentCore to host Model Context Protocol (MCP) tools for building extensive, network-agnostic agent workflows.

A Seamless Foundation for Intelligent Commerce

The primary goal? To provide a secure, scalable foundation for developing the next generation of intelligent commerce solutions.

Key Components of Visa Intelligent Commerce on AWS

In the AWS Marketplace, developers can explore Visa’s essential agentic commerce tools, including authentication, tokenization, and user intent capture. These resources enable autonomous, secure, and contextual payment flows. With the blueprints provided through the Amazon Bedrock AgentCore samples repository, developers can easily integrate security and transaction tools into their systems.

Amazon Bedrock AgentCore acts as the backbone for these intelligent commerce experiences. Key features include:

1. Secure Runtime: Each agent operates in isolated environments, protecting sensitive information while allowing for multi-step workflows.

2. Identity Management: Secure inbound and outbound authentication ensures user data remains safe.

3. Governed Access: The AgentCore Gateway allows agents to access payment solutions while meeting trusted access controls.

4. Contextual Memory: Allows agents to maintain long-term context over interactions, enhancing user experience and providing valuable insights.

5. Complete Observability: Full transparency into agent actions provides audit-ready records across workflows.

A Modular Supervisor Architecture

One of the standout features of the architecture is the Reusable Supervisor Pattern, which enhances development efficiency. This approach allows developers to use a shared supervisor agent across multiple commerce use cases, reducing the overhead associated with building distinct systems.

This pattern facilitates:

• Effective multi-domain deployments.
• The addition of specialized agents without altering the core orchestration.
• Consistent user experiences across various transactions.

Sample Implementations: Travel and Shopping Agents

Revolutionizing Travel with Intelligent Agents

Currently, travelers experience a fragmented process when booking trips, often hopping between multiple platforms. To simplify this, a travel booking agent powered by Amazon Bedrock AgentCore pairs up with Visa Intelligent Commerce.

Core components include:

• Supervisor: Central orchestrator managing user interactions.
• Travel Assistant: Handles planning, bookings, and recommendations.
• Cart Manager: Manages payments and purchase flows while ensuring security.

This multi-agent system transforms the travel experience into a streamlined process, allowing users to finalize their itineraries and bookings with just a few commands.

Redefining Shopping with Intelligent Commerce

Shoppers often navigate a convoluted landscape of websites, apps, and loyalty programs. With a multi-agent shopping assistant powered by Amazon Bedrock AgentCore and integrated with Visa Intelligent Commerce, the complexity simplifies drastically.

Users can issue commands like, "Find the best deal for a Sony PlayStation 5 Pro under $500," and the assistant autonomously executes the complete shopping flow—searching, comparing prices, validating promotions, and finalizing purchases—all without manual clicks.

Conclusion

The collaboration between AWS and Visa illustrates how agentic commerce can reshape the transaction experience. Transitioning from a fragmented process to a cohesive, intelligent journey allows users to explore, compare, and purchase seamlessly.

As the future of commerce unfolds, where intelligent agents operate tirelessly on users’ behalf, opportunities for innovation abound across various sectors. Join us at the forefront of this transformation and begin building secure, streamlined payment experiences utilizing Amazon Bedrock AgentCore and Visa Intelligent Commerce. Explore our GitHub repository for sample projects to get started.

About the Authors

• Sangeetha Bharath: AI strategy leader at Visa, specializing in neural networks and large language models.
• Seemal Zaman: Product leader at Visa, focused on applying agentic AI to enhance B2B and consumer experiences.
• Isaac Privitera: Principal Data Scientist at the AWS Generative AI Innovation Center, dedicated to building responsible AI systems.
• Hardik Thakkar: Sr. Security Prototyping SA at AWS, specializing in secure architecture design for financial services.
• Daniela Vargas: Prototyping Solutions Architect at AWS, creating innovative prototypes based on customer needs.
• Ritambhara Chaterjee: Senior Solutions Architect at AWS, helping financial institutions leverage AI in their operations.
• Ankit Pathak: Leader of ML and AI practices at AWS, driving generative AI adoption.

In this rapidly evolving landscape, the fusion of AI and commerce is just beginning. Stay tuned as we continue to explore new possibilities!