The Agent-to-Agent Economy: What Happens When Bots Start Billing

In January 2026, Google CEO Sundar Pichai announced the Universal Commerce Protocol at the National Retail Federation conference, signaling that AI-completed purchases had officially arrived (Ekamoira, January 2026). Within weeks, Visa announced that hundreds of secure agent-initiated transactions had been completed across its partner ecosystem, predicting millions of consumers would use AI agents to complete purchases by the 2026 holiday season (Visa, 2025). McKinsey projected that agentic commerce could generate $3 trillion to $5 trillion globally by 2030 (McKinsey, October 2025).

The agent economy isn't a future scenario. It's the present. And it's growing at a pace that makes the early e-commerce boom look leisurely.

From Assistants to Autonomous Actors

The shift happening right now is more fundamental than most people realize. We're moving from AI that recommends to AI that transacts.

The first wave of AI in commerce was advisory — chatbots that helped you find products, compare prices, read reviews. The agent was an assistant. You still clicked "Buy."

The second wave, unfolding now, is autonomous. As McKinsey describes it, agentic commerce features AI that "anticipates consumer needs, navigates shopping options, negotiates deals, and executes transactions" (McKinsey, October 2025). The agent doesn't just suggest — it discovers, evaluates, negotiates, pays, and confirms. The human sets parameters and steps back.

But there's a third wave that's already happening in the B2B and developer space: agents buying from other agents, with no human in the loop at all. Your inference agent needs compute, so it buys GPU time from a compute provider's agent. Your data pipeline agent needs fresh training data, so it purchases from a data vendor's agent. These are pure machine-to-machine transactions — discovered, negotiated, executed, and settled programmatically.

Morgan Stanley estimates that agentic shoppers could represent $190 billion to $385 billion in U.S. e-commerce spending by 2030 (Ekamoira, January 2026). Deloitte-based estimates suggest up to 30% of global e-commerce transaction value could be influenced by agentic AI by 2030 — approximately $17.5 trillion in gross merchandise value (Getnet/Santander, January 2026).

The Infrastructure Is Being Built — Without Tax

The commerce infrastructure for agents is materializing rapidly. Consider the protocols and standards that launched in just the past year:

Model Context Protocol (MCP) from Anthropic enables agents to read data and interact with external tools through a standardized interface.

Agent-to-Agent Protocol (A2A) enables agents to exchange capabilities, status, and context through standardized protocols regardless of vendor or architecture.

Agent Payments Protocol (AP2) from Google is a payment-agnostic protocol enabling autonomous AI agents to make verifiable purchases on behalf of users.

Universal Commerce Protocol (UCP) from Google establishes direct integrations between agents and merchants for product discovery, availability, pricing, and checkout.

Visa Intelligent Commerce provides tokenization and trust infrastructure specifically for agent-initiated transactions, with over 100 partners building in its ecosystem (Visa, 2025).

J.P. Morgan, Stripe, and other payment processors are actively developing agent-compatible payment rails (J.P. Morgan Payments; PYMNTS, February 2026).

Notice what's missing from this stack? Tax. There's a protocol for agent identity. A protocol for agent payments. A protocol for product discovery. A protocol for trust and tokenization. But there is no protocol for tax compliance. No standard for how an agent determines its tax obligation, collects the correct amount, routes the tax portion to the right reserve, and reports the transaction for filing.

Every other layer of the agent commerce stack is being built. The tax layer is a blank space.

Why Tax Can't Be an Afterthought

In the early days of e-commerce, tax was exactly that — an afterthought. Online sellers operated for years without collecting sales tax, and states spent decades trying to catch up. The 2018 Wayfair decision was the culmination of that catch-up effort, finally allowing states to require remote sellers to collect tax based on economic presence rather than physical presence (South Dakota v. Wayfair, Inc., 2018).

The lesson from e-commerce should be clear: ignoring tax obligations during a commercial revolution doesn't make them go away. It creates a massive liability that compounds until enforcement catches up. And enforcement always catches up.

The agent economy is making the same mistake, but at dramatically higher velocity. E-commerce grew over decades. The agent economy is scaling in months. Forrester predicts that by 2026, one in five sellers will be responding to AI-powered buyer agents (CommerceTools, January 2026). Gartner forecasts that 40% of enterprise applications will embed task-specific AI agents by 2026, up from less than 5% in 2025 (Salesmate).

Every one of those agent-to-agent transactions has a tax obligation. Sales tax. Use tax. 1099 reporting. Nexus tracking. And virtually none of it is being captured.

What the Agent Economy Needs

The tax layer for agent commerce requires several things that don't exist in traditional tax infrastructure:

Protocol-native compliance. Tax calculation must be embeddable in agent-to-agent protocols — a standard API call that any agent can make before settling a transaction, regardless of framework, language, or payment rail.

Real-time classification. Agents transact in milliseconds. Tax classification can't require human review or manual lookup. The system must automatically determine whether a transaction is a digital service, digital good, or general service, and apply the correct state rules.

Bilateral awareness. In agent-to-agent commerce, both sides have tax obligations. The seller may owe sales tax. The buyer may owe use tax. The system needs to track both perspectives — not just one.

Identity without friction. Agents need a way to communicate their tax-relevant identity — jurisdiction, entity type, exemption status — without disrupting the transaction flow. This is the "network" component: a registry of agent tax profiles that other agents can query.

Scale-ready filing support. An agent making 10,000 transactions per day across 40 states generates filing obligations that no human can manage manually. The compliance data must accumulate automatically and be export-ready for every jurisdiction.

This is exactly what AgentTax provides. One API endpoint. One call per transaction. Both seller and buyer perspectives. Automatic classification, nexus checking, rate application, 1099 tracking, and dashboard logging. It's the tax layer the agent commerce stack is missing.

The Opportunity Window

Right now, we're in a rare moment where the infrastructure is being built but the standards haven't hardened. Google's UCP, Visa's Intelligent Commerce, and the A2A protocol are all in their early stages. Partners are being selected. Integration patterns are being established.

The agent economy will have a tax compliance layer. The only question is whether it's bolted on retroactively — creating the same mess that Wayfair tried to clean up in e-commerce — or built in from the beginning as a native protocol component.

Every AI builder shipping agents that buy or sell commercially should be thinking about this now. Not because a state is going to audit your agent tomorrow. But because the transactions are accumulating, the thresholds are being crossed, and the enforcement apparatus is getting smarter every quarter.

The bots are billing. It's time to make sure they're billing correctly.

AgentTax is the tax compliance protocol for the agent economy. One API call per transaction. Get your free API key →

Sources:

• McKinsey, "The Agentic Commerce Opportunity," October 2025

• McKinsey, "The Automation Curve in Agentic Commerce," January 2026

• Visa, "Visa and Partners Complete Secure AI Transactions," 2025

• Ekamoira, "What Is Agentic Commerce? 2026 Guide," January 2026

• Getnet/Santander, "Getnet Outlines Agentic Commerce Strategy," January 2026

• PYMNTS, "AI Agents Are Becoming the New Power Brokers in Digital Commerce," February 2026

• J.P. Morgan Payments, "Agentic Commerce: The Future of AI-Powered Shopping"

• CommerceTools, "7 AI Trends Shaping Agentic Commerce in 2026," January 2026

• Salesmate, "AI Agent Adoption Statistics by Industry (2026)"

• DevPro Journal, "NRF 2026: The Rise of Agentic Commerce," January 2026

• South Dakota v. Wayfair, Inc., 585 U.S. ___ (2018)