Commerce is on the edge of a profound transformation. Agentic commerce, where AI-powered agents autonomously conduct transactions, will redefine how consumers and businesses buy, sell, and interact.
These agents go beyond automating tasks. They understand intent, negotiate deals, and settle payments across interoperable systems while delivering personalized experiences at scale. Powered by emerging standards like the Model Context Protocol (MCP) and Google’s Agent-to-Agent (A2A) protocol, agentic commerce promises a future that’s not just faster or cheaper, but fundamentally smarter.
The Problem with Today’s Commerce
Despite decades of digital innovation, modern commerce is still full of friction.
Consumers bounce between apps, compare prices manually, struggle with clunky checkout flows, and run into surprise fees. The cognitive load is exhausting. It’s no surprise that cart abandonment rates hover around 70%, according to a 2024 Baymard Institute study. Research from the University of Minnesota shows that, on average, consumers visit 3 to 5 websites before making a purchase over $100.
Businesses, from freelancers to Fortune 500s, deal with slow supplier onboarding, fragmented payments, and siloed platforms that require expensive, custom integrations. According to Deloitte’s 2024 Global CPO Survey, a typical enterprise purchase involves over 15 approvals, multiple supplier negotiations, and complex payment reconciliation across disconnected systems. McKinsey research shows companies spend an average of 6.4 hours on administrative tasks for each purchase order.
This inefficiency continues because today’s systems are manual, disconnected, and reactive rather than proactive.
The Promise of Agentic Commerce
Agentic commerce flips the script. AI agents, which are autonomous and goal-driven software entities, act on behalf of users and businesses to streamline end-to-end transactions.
A well-designed agent can:
- Understand intent by analyzing context, preferences, and history.
- Negotiate dynamically using real-time market data and counterparty signals.
- Execute seamlessly by integrating with payment and logistics systems.
- Establish trust through verified identities and transparent reputation systems.
In essence, they operate like human assistants but only faster, more affordable, and available 24/7.
Standards Driving the Ecosystem: MCP and A2A
Open standards are essential to making agentic commerce work. Two in particular, Anthropic’s MCP and Google’s A2A protocol, are laying the foundation.
MCP (Model Context Protocol): Introduced in November 2024, MCP standardizes how agents connect to external tools and data. It turns APIs, files, and databases into accessible components through a local server model. By April 2025, it had gained adoption from OpenAI, Google, Microsoft, and others, becoming a unifying interface layer for AI applications.
A2A (Agent-to-Agent Protocol): Unveiled in April 2025, A2A enables agents to discover, communicate, and collaborate using open web standards such as HTTP, JSON-RPC, and Server-Sent Events. It allows agents to coordinate tasks dynamically without exposing sensitive internal logic.
Together, these protocols solve the M×N integration problem, where each app needs a custom connection to every other system. Instead, they introduce an M+N model, where every tool or app connects once to a shared ecosystem. The result is one-time integration, reusable components, and a much more scalable architecture.
Early Examples of Agentic Commerce in Action
This is not some distant future. It is already happening. Here are some early examples:
Amazon Buy for Me (April 2025): Launched as a feature in the Amazon Shopping app, “Buy for Me” allows users to purchase products from third-party websites without leaving Amazon. The AI handles product discovery, checkout, and payment, all within the app, removing the need to visit external sites.
OpenAI Operator (January 2025): Operator is an AI agent that completes complex tasks like booking travel or placing orders by interacting directly with websites. It fills out forms, clicks buttons, and self-corrects errors without needing APIs. It automates what a user would normally do in a browser.
Perplexity + PayPal Travel (May 2025): Perplexity AI integrated PayPal and Venmo to support travel bookings and purchases directly within chat. Users can now book hotels or flights during a conversation, powered by partnerships with Tripadvisor and Selfbook, with payments handled seamlessly through PayPal.
SAP Ariba AI: SAP Ariba uses AI to automate B2B procurement workflows, including supplier negotiations, purchase orders, and contract management. Enterprises report up to a 50 percent reduction in cycle times by eliminating manual coordination.
Coupa AI Spend Management: Coupa applies AI to streamline business spending, covering invoice processing, approvals, and vendor payments. Companies use it to reduce administrative overhead, accelerate payments, and improve cash flow visibility.
These are not isolated features. They are early signs of a larger shift. The foundation for agentic commerce is already taking shape.
How Businesses Can Prepare
Today’s commerce infrastructure was built for humans to navigate, not for AI systems to operate within. To thrive in an agentic ecosystem, businesses must become agent-readable and agent-cooperative. That means:
- Expose structured data: Product catalogs, inventory, and pricing must be made accessible via APIs or MCP-compliant servers.
- Adopt MCP and A2A: Tools like Speakeasy can help generate MCP servers from OpenAPI specs, fast-tracking implementation.
- Enable dynamic pricing: Let agents negotiate pricing based on volume, timing, or market conditions.
- Integrate logistics and payments: Ensure seamless execution by connecting to services like Stripe, PayPal, or ShipBob.
- Build trust signals: Use OAuth-based authentication, transparent return policies, and visible ratings to help agents prefer your platform.
Challenges Ahead
Of course, agentic commerce isn’t without hurdles:
- Interoperability: Competing standards must align. Early collaboration between Anthropic, Google, and others is encouraging.
- Regulatory compliance: AI-driven transactions will face increased scrutiny. Frameworks like MCP’s OAuth 2.1 model demonstrate how security and auditability can be built in from the start.
- Adoption: Most users aren’t developers, so interfaces must stay simple and intuitive. Klarna’s AI shopping assistant is a good example as it blends effortlessly into regular browsing without adding complexity.
- Security: Local agent servers introduce risks. Tighter OAuth scopes and sandboxing will be critical.
The Road Ahead
From Amazon’s AI shopping concierge to SAP’s autonomous procurement agents, we’re seeing the blueprint of an agentic future unfold.
In that future, AI agents don’t just help. They transact, negotiate, and deliver outcomes in real time. Businesses that expose their systems to these agents via open protocols like MCP and A2A will become preferred nodes in an intelligent, interconnected economy.
As the digital economy goes autonomous, the real question is whether your business is ready to lead it.