A2A Commerce: How Agent-to-Agent Transactions Will Replace Traditional E-Commerce

Understanding Agentic Commerce: Why Agent-to-Agent Interactions Will Define the Next Decade

Last year, a developer tried to sell an old sofa. He didn't want to deal with classifieds, so he let his AI agent list it. A buyer's agent found it, negotiated a 8% price reduction, and completed the transaction - all while both owners slept. The sofa arrived the next morning. Neither human had typed a single message. Neither had opened a website.

Expert analysis of Agent-to-Agent (A2A) commerce: protocols, market forecasts ($3–5Tn by 2030), real-world deployments, liability gaps, and strategic imperatives for businesses.

This is Agent-to-Agent (A2A) Agentic Commerce. It's not science fiction. It's happening right now. And it's about to change how everything is bought and sold.

We will explore what A2A commerce is, how it works, the real-world examples already live in 2026, the staggering market projections, the technological and regulatory challenges, and how you can prepare.

A Trillion-Dollar Shift You Can't Afford to Ignore

Let's look at the numbers first. They tell a story of a transformation as big as the internet itself.

By 2030, McKinsey projects that the U.S. B2C retail market alone could see up to $1 trillion in revenue orchestrated by agentic commerce, with global projections reaching $3 trillion to $5 trillion.

This growth isn't just confined to the U.S. Edgar, Dunn & Co. forecasts that agentic commerce will reach $2.9 trillion in global retail transaction flows by 2030, representing 29% of total e-commerce and growing at a compound annual rate of 185% between 2026 and 2030. In the U.S. specifically, Morgan Stanley projects that agentic shoppers could represent $190 billion to $385 billion in e-commerce spending by 2030, capturing a 10% to 20% market share. According to a Bain & Company report, AI agents could account for between 15% and 25% of U.S. e-commerce sales by 2030, a market worth roughly $300 billion to $500 billion.

McKinsey has called this "a structural transformation of how products are discovered, evaluated and purchased". This transformation will achieve broad influence faster than previous revolutions because AI agents can ride the infrastructure built by the internet and mobile revolutions.

But to understand this market, you need to understand what A2A actually is.

What Is Agent-to-Agent (A2A) Commerce? A Clear Definition

Let's define the terms precisely.

What Are AI Agents?

Modern AI agents are not static "chatbots." They are sophisticated software systems capable of reasoning, planning, and acting across multiple platforms and AI tools. They operate with three characteristics:

Autonomy: They act without constant user input.
Reasoning: They adapt recommendations based on evolving conditions - price changes, stock depletion, and delivery delays.
Interoperability: They integrate across many AI platforms through open APIs.

Agentic Commerce vs. A2A Commerce

Agentic Commerce refers to the use of AI agents that research, compare, negotiate, or execute purchases - sometimes with human approval, sometimes without.

Agent-to-Agent (A2A) Commerce is the advanced subset where a buyer's agent communicates directly with a merchant's agent, creating an invisible negotiation layer that bypasses traditional websites and checkout flows.

The Interaction Models

Google has defined two primary interaction models for this new landscape:

Consumer-to-Merchant (C2M): A consumer's personal AI agent interacts with merchant agents on their behalf.
Merchant-to-Merchant (M2M): Merchant agents interact with other merchant agents, such as when a retailer's agent is out of stock and negotiates with another retailer's agent to source the product.

From Ask to Buy, Not Search to Click

The fundamental shift is from a "search and click" model to an "ask and buy" model. In traditional e-commerce, you manually search, compare, click, and type. In agentic commerce, you provide a goal - "Find me noise-cancelling headphones under $300 with great battery life" - and your agent does the rest. The entire journey, from discovery to post-purchase support, can be handled without you ever visiting a website.

How Agent-to-Agent Commerce Actually Works

Let's walk through a complete A2A transaction, step by step, to demystify the process.

Agent-to-Agent (A2A) Commerce Flow by careertechinsight.in

Step 1: Goal Setting

A user gives a high-level instruction to their personal AI agent. Rather than "search for headphones," the prompt is "Find me noise-cancelling headphones under $300 with great battery life. I need them by Friday." This is the only direct human input required.

Step 2: Agent Discovery

The buyer agent discovers merchant agents using protocols designed for this purpose. In January 2026, Google launched the Universal Commerce Protocol (UCP), an open standard for agentic commerce that works across the entire shopping journey - from discovery and buying to post-purchase support. UCP is compatible with existing industry protocols like Agent2Agent (A2A) for agent coordination, the Agent Payments Protocol (AP2) for secure payments, and the Model Context Protocol (MCP) for tool integration.

Step 3: Data Exchange

Merchant agents expose structured, machine-readable product data using UCP's standardized methods. As of March 19, 2026, Google has expanded UCP capabilities with features such as "multi-item cart support, real-time catalog access (including inventory and pricing), and identity linking to preserve loyalty benefits and logged-in experiences". UCP uses a "capability declaration + RESTful API + standardized data model" architecture, secured through HTTPS and OAuth 2.0.

Agents can assess factors beyond simple specs, including organizational data like sustainability practices and core values, when evaluating offers.

Step 4: Negotiation and Selection

This is where A2A commerce proves its value. Agents can engage in dynamic pricing negotiations, bundle products, apply loyalty discounts, and compare trade-offs across multiple merchants simultaneously - all in milliseconds.

Step 5: Transaction

Payment is handled through emerging protocols. Beyond Google's AP2, a major alternative is the Agentic Commerce Protocol (ACP) , co-developed by OpenAI and Stripe. ACP was announced in September 2025 and provides an open standard that allows AI agents to discover products, initiate checkouts, and complete purchases without leaving the conversation. It has been implemented as a production-ready reference by both OpenAI and Stripe.

Step 6: Fulfillment and Post-Purchase

The merchant's agent handles fulfillment, sends confirmation, tracks delivery, and manages returns or support requests - all agent-to-agent. The user is only notified when action is required.

The Protocol Wars: Google, OpenAI, Stripe, and the Race for Standards

The battle for agentic commerce supremacy is being fought over protocols - the "language" that agents use to communicate.

Google's Approach: UCP as the Orchestrator

Launched at NRF 2026 in January, UCP establishes a "common language for agents and systems to operate together across consumer surfaces, businesses, and payment providers". Rather than requiring unique connections for every individual agent, UCP aims to enable all agents to interact easily. Agents and businesses can "pick and choose specific extensions of the protocol that suits their needs". It is currently supported by major partners including Shopify, Etsy, Wayfair, Target, Walmart, Home Depot, Lowe's, Poshmark, and Reebok, with Ant Financial International also on board. UCP has been released under the Apache open-source license.

OpenAI and Stripe's Approach: ACP for Simplicity

Stripe and OpenAI took a different approach, focusing on the core transaction. The result: ACP functions effectively as the "checkout API for AI," defining how the agent discovers products and constructs the transaction. Merchants integrate once and can sell through AI agents while retaining full control over branding and fulfillment. ACP is also an open standard, and its documentation has been made available as open source. PayPal has since announced it will adopt ACP, bringing product catalogs of tens of millions of merchants into ChatGPT.

The landscape is complex. Between October 2025 and April 2026, more than ten payment protocols launched to let AI agents spend money on their own. The ultimate outcome is unknown - either one protocol will win, or they will interoperate. Google has positioned UCP as compatible with both ACP and AP2, perhaps aiming to become the connecting layer.

Real-World Examples and Case Studies in 2025–2026

These are not theoretical. Several major implementations are already live.

Anthropic's "Project Deal" Marketplace

In perhaps the most revealing experiment, Anthropic created a classified marketplace where AI agents represented both buyers and sellers, striking real deals for real goods and real money. The pilot included 69 employee participants. In December 2025, Claude interviewed people about which belongings they might want to sell and what they might be willing to buy. The autonomous agent‑to‑agent negotiations resulted in completed transactions.

Instacart Inside ChatGPT

In December 2025, Instacart became the first grocery partner to launch an end-to-end shopping app inside ChatGPT. Shoppers can build carts and complete checkout directly within a chat, with orders fulfilled through Instacart's delivery network. Behind the scenes, it leverages ACP technology for transactions. OpenAI has since expanded similar partnerships with Walmart and Target.

JD Sports Tests Direct AI-to-Checkout

At NRF 2026 in January, JD Sports announced a test connecting AI‑led shopping interactions to existing e‑commerce infrastructure, including pricing, inventory, fulfillment, and payments.

Albertsons and the AI Shopping Assistant

In November 2025, Albertsons Companies launched an AI shopping assistant with agentic commerce capabilities including budget optimization, in‑store aisle location, and voice integration.

B2B: Real Production Deployment

In April 2026, GreenCore Solutions announced the full production release of 30 deterministic AI Orderability agents operating across Microsoft Azure AI Foundry and Google Cloud A2A Enterprise. These AI agents interface directly with buy-side AI systems inside retail grocery ERP environments including SAP S/4HANA, Oracle Fusion, and Microsoft Dynamics 365, handling procurement and ESG decisions. The agent traffic reached 1.96 million calls that month, delivering a sustained 6:1 AI-agent-to-human ratio.

The Quiet Revolution: Asian Markets

Asia Pacific is leading adoption. According to Deloitte, 29% of consumer businesses in Asia Pacific report they are already adopting agentic AI, and this is expected to surge to 76% within two years.

The Opportunities - Why This Market Is Exploding

The growth isn't hype. It's driven by real value for everyone in the ecosystem.

For Consumers: Time, Money, and Personalization

Massive time savings: Agents work tirelessly while you sleep.
Better deals: Your agent monitors hundreds of merchants and negotiates dynamically.
Hyper-personalization: The agent builds deep context over time - your size, style, budget, delivery preferences, brand affinities, even your calendar.
Proactive shopping: Your agent anticipates needs before you articulate them.

Consumer adoption is following early patterns. According to NRF data, while 72% of consumers still shop in physical stores, 41% already use AI assistants for product research, 33% for reviews, and 31% for deal searching. A Riskified study from Q1 2026 found that 61.5% of consumers have used AI tools for product discovery. Strikingly, at NRF 2026, Tapestry Inc. revealed that 57% of people today trust AI advice more than they trust advice from family and friends.

For Merchants: New Channels, Higher Conversion

The invisible shelf: Merchants sell inside AI applications without requiring customers to visit a website. Google calls this the "invisible shelf" - alongside physical and traditional digital shelves.
Higher conversion rates: AI-driven checkout dramatically reduces cart abandonment.
Rich data exchange: Merchants gain structured insight into consumer intent and decision criteria.
Dynamic pricing optimization: Merchant agents adjust pricing in real time based on demand signals.

The Challenges and Risks - What Could Go Wrong

For all its promise, agentic commerce faces significant hurdles.

The Liability Vacuum

This is perhaps the most pressing risk. A recent analysis describes a "$385 billion liability vacuum" - a reference to the entire projected market size - noting that "layering agentic commerce on top of existing infrastructure without resolving liability turns a scaling challenge into a structural one". When an AI agent makes a bad purchase - whether fraud, misuse, misunderstood intent, or simple error - the merchant currently carries the risk. In an autonomous environment, the audit trail becomes the merchant's primary defense.

Disputes in agentic commerce "will become less about whether a transaction happened and more about proving how it happened". Legal issues extend to new frauds and scams, whether website terms of service are enforceable against agentic transactions (if the user has never agreed to them), and privacy.

Security and "Know Your Agent"

Traditional fraud models rely heavily on static credentials and behavioral data. Agentic transactions often pass with "far less context" than traditional flows. This has led to a new requirement: "Know Your Agent" (KYA) procedures - verifying the identity, authority, and integrity of the AI agent.

In May 2026, Experian launched "Agent Trust," a solution that combines identity checks, payment authentication, and network enforcement for agent-led commerce. The company's products already help clients avoid an estimated $15–19 billion in fraud losses annually.

Interoperability and Fragmentation

With multiple competing protocols - UCP, ACP, AP2, MCP, and others - the risk of fragmentation is real. If the ecosystem becomes Balkanized, agents on one platform may not be able to interact with agents on another, limiting commerce.

Data Privacy

For agents to act effectively on your behalf, they need deep access to personal data - location history, purchase records, calendar, budget, health data. This concentration creates obvious privacy risks, and regulators are taking notice.

Regulation: The CMA Takes the Lead

On 9 March 2026, the UK's Competition and Markets Authority (CMA) published guidance on complying with consumer law when using AI agents. The guidance addresses risks like AI agents facilitating coordinated pricing outcomes "by learning and reacting to other AI agents in a concentrated market" without explicit communication. The Digital Regulation Cooperation Forum has also weighed in, highlighting that under the Consumer Rights Act 2015, agentic AI providers must still meet minimum standards of quality, performance, and fitness for purpose.

Trust Gap

Consumer hesitation is real. The same Q1 2026 Riskified study found that 55% of consumers are not comfortable with AI agents making purchases on their behalf. While a Radial survey found that 58% of consumers are open to using an AI agent, only 6% have actually done so.

Future Outlook and Predictions (2026–2035)

Based on current data and projections, here is a realistic timeline.

Short-Term (2026–2027): Pilots and Early Production

Expect continued pilot programs, limited production deployments, and protocol refinements.
Consumer adoption is projected to increase from 19% in 2025 to 46% by the end of 2026.
Gartner predicts that 33% of organizations will adopt agentic AI by 2028, up from less than 1% in 2024.

Medium-Term (2027–2030): Mainstream Adoption

By 2028, Gartner predicts that AI agents will handle 90% of all B2B purchases - over $15 trillion in annual spend.
By 2030, 20% of digital commerce transactions will be agent-executed (Gartner).
Morgan Stanley projects agentic GMV at 10–20% of U.S. e-commerce by 2030.
McKinsey projects $1 trillion in the U.S. and $3–5 trillion globally.

Long-Term (2030–2035): Autonomous Economies

Gartner predicts that machine customers will directly influence or participate in $30 trillion worth of purchases by 2030.
Tietoevry expects the value of e-commerce initiated by agents to reach €191 billion by 2035.
Over 50% of CEOs plan to have a strategy within two years to deal with machines as part of buying or selling (Gartner).
By 2028, an estimated nine billion IoT devices will have the potential to become "machine customers."

How to Prepare for the Agentic Era

Whether you are an individual, a business, or a developer, here are concrete actions.

For Individuals

Start using AI agents early. Experiment with ChatGPT with Instacart or Gemini with Google Shopping.
Set clear preferences and constraints. The quality of agent output depends on the quality of your instructions.
Review agent transactions regularly. Catch mistakes early and train your agent to do better.
Understand liability. Know who is responsible if your agent makes a bad purchase.

For Businesses

Make product data agent-friendly. This is the single most important action. Gartner's research emphasizes that "early movers with complete product data can gain superior positioning in recommendation and transaction workflows". Ensure product catalogs are structured, complete, and machine-readable. Use APIs and structured data formats.
Integrate agentic protocols. Implement support for UCP, ACP, or both. Work with your e‑commerce platform provider - Shopify, BigCommerce, WooCommerce, Adobe Commerce - to enable agentic integrations.
Deploy a merchant agent. Build or adopt a front-end agent that can respond to buyer queries.
Focus on the "invisible shelf." Google suggests CPG leaders "treat your product data as your new packaging" because in agentic commerce, product data wins.
Prepare for liability. Review legal and fraud frameworks. Consider implementing KYA procedures and robust audit trails.
Build trust. Consumers must trust that their data is protected and that agents act in their best interest. Trust is the most valuable asset retailers can protect in the agentic era.

For Developers

Familiarize yourself with A2A frameworks. Explore UCP, ACP, and MCP documentation.
Experiment with agent orchestration tools. Build simple agents that can search, compare, and transact.
Contribute to open protocols. The ecosystem is being built in the open. Your contributions shape the future.
Design for security and privacy. Built with KYA, encryption, and granular consent.

Conclusion - The Invisible Shelf Is Open

We are in the earliest stages of the agentic commerce revolution. Projections of $3–5 trillion by 2030 are not speculative hype - they are based on demand that is already being expressed and infrastructure that is already being built.

The protocols are live. The pilots are delivering. The merchants are integrating. And the road ahead - while challenging - is being paved in open standards, not closed silos.

The question is not whether agentic commerce will transform how we buy and sell. The question is who will be ready when it does.

Let's Talk - What Do You Think?

Would you trust an AI agent to make purchases on your behalf? What limits would you set?

Is your business ready for agentic commerce? What is your biggest barrier - data readiness, technical integration, or leadership buy-in?

Which protocol do you think will win - Google's UCP or OpenAI's ACP - or will they coexist?

The comment section is open. Let's have a conversation about the future of shopping.

Share This With Your E‑commerce and Product Teams

Tag your Chief Digital Officer. Share this in your company Slack. Post it on LinkedIn with the caption: "The invisible shelf is open. Is your brand on it?"

FAQ

Q: What is the difference between Agentic Commerce and A2A Commerce?

A: Agentic Commerce uses AI agents to research, compare, negotiate, or execute purchases. A2A Commerce is the subset where buyer agents and merchant agents communicate directly without human involvement beyond initial instruction.

Q: Is agentic commerce actually happening in 2026?

A: Yes. Live deployments include Instacart inside ChatGPT, JD Sports test, Anthropic's Project Deal, and GreenCore Solutions' 30 AI agents in B2B procurement.

Q: How big will the market be by 2030?

A: McKinsey projects $1 trillion in the U.S. and $3–5 trillion globally. Morgan Stanley projects $190–385 billion in the U.S. alone (10–20% of e-commerce). Bain projects 15–25% of U.S. e-commerce sales ($300–500 billion).

Q: What protocols enable this?

A: Google's UCP (Universal Commerce Protocol) and AP2 (Agent Payments Protocol); OpenAI and Stripe's ACP (Agentic Commerce Protocol); and Anthropic's MCP (Model Context Protocol) for tool access.

Q: What is the biggest risk?

A: Liability. When an agent makes a bad purchase, it's unclear who is responsible. Also, security, data privacy, consumer trust, and regulatory compliance.

Q: How can my small business prepare?

A: Start with product data. Ensure your catalog is structured, complete, and accessible via APIs. Integrate with protocols through your e‑commerce platform.

Q: Which countries are leading adoption?

A: According to Deloitte, the Asia Pacific region is expected to surge from 29% adoption today to 76% within two years.

Q: Are consumers ready to trust AI agents?

A: Adoption is growing - 61.5% have used AI for product discovery - but 55% are not comfortable with agents making purchases yet. A gap exists between openness (58%) and actual use (6%).

Research Sources

Google Cloud Blog – "The invisible shelf: How CPGs can win agentic commerce in 2026" (January 9, 2026)
Google Blog – "New tech and tools for retailers to succeed in an agentic shopping era" (January 11, 2026)
McKinsey & Company – "The agentic commerce opportunity" (October 2025)
McKinsey – "Are you - or your AI agent - shopping on Cyber Monday?" (December 2025)
Morgan Stanley – "A Novel Way to Shop Online" (February 17, 2026)
Stripe – "Developing an open standard for agentic commerce" and "Stripe powers Instant Checkout in ChatGPT" (September 29, 2025)
GitHub – agentic-commerce-protocol/agentic-commerce-protocol (September 2025)
Channel Engine – "Google's Universal Commerce Protocol and Merchant Center" (March 24, 2026)
Digital Commerce 360 – "Instacart ChatGPT app" (December 12, 2025)
Digital Commerce 360 – "JD Sports tests agentic commerce" (January 12, 2026)
Digital Commerce 360 – "What Anthropic, OpenAI and Google are each doing in agentic commerce" (April 30, 2026)
Anthropic – "Project Deal: our Claude-run marketplace experiment" (April 24, 2026)
TechCrunch – "Anthropic created a test marketplace for agent-on-agent commerce" (April 25, 2026)
GreenCore Solutions – "30 AI Agents for TreeFree Diaper® Launched" (April 29, 2026)
Deloitte – "Asia Pacific set to lead the agentic future of commerce" (March 30, 2026)
Riskified – "Q1 2026 Agentic Commerce Pulse" (April 27, 2026)
Edgar, Dunn & Co – "Acquirers Must Adapt to Agentic Commerce" (April 1, 2026)
Bain & Company – cited in TheFashionLaw (March 13, 2026)
NewtonX – "A $385B liability vacuum" (April 23, 2026)
National Law Review – "Agentic AI Commerce: Next Wave of Online Shopping and Retailer Risk" (April 20, 2026)
The Paypers – "Agentic commerce: risks, regulation, readiness" (April 22, 2026)
Rivero – "Agentic Commerce Needs a New Dispute Framework" (March 6, 2026)
IXOPAY – "Why Agentic Commerce Needs a Unified Trust Layer" (March 23, 2026)
Experian – "Experian launches Agent Trust for AI-led commerce" (May 1, 2026)
SISA – "Payment Intelligence Report – Fraud & Scam Landscape" (January 2026)
CMA UK – "Complying with consumer law when using AI agents" (March 9, 2026)
Cooley – "AI Agents and Consumer Law" (March 26, 2026)
Herbert Smith Freehills – "CMA focus on agentic AI" (March 25, 2026)
NRF – "Own the agentic commerce experience" (January 7, 2026)
Gartner – cited in OroCommerce and Netcore reports (2026)
Tietoevry – "A-commerce agents: all change for payments?" (October 7, 2025)
Entrepreneur – "The Way People Shop Has Quietly Changed Forever" (April 30, 2026)

Disclaimer

This article is for informational purposes only and does not constitute financial, legal, or professional advice. All market projections, timelines, and forecasts are based on publicly available research from sources including McKinsey, Gartner, Morgan Stanley, Bain & Company, Deloitte, and others as cited above. Actual outcomes may vary materially due to changes in technology, regulation, consumer behavior, and economic conditions. References to specific companies, products, or protocols do not constitute endorsements. Readers should consult qualified professionals for advice relevant to their specific circumstances. The views expressed are solely those of the author.