Definition
Agentic Commerce
Agentic commerce is an economic model in which autonomous AI agents negotiate, execute, and manage commercial transactions on behalf of human users or organizations, shifting e-commerce from manual, user-driven processes to automated machine-to-machine transactions.
Updated
What It Is
Think of the difference between a human steering a car and a self-driving one. Traditional e-commerce is the human: you browse, compare, click, and pay. Agentic commerce is the self-driving car: an AI agent on your behalf finds options, evaluates them against your preferences and budget, negotiates terms, and completes the transaction — often without you touching a button.
The key shift is delegated authority. Instead of AI merely recommending products (what the industry calls assisted commerce), agentic commerce lets AI agents execute purchases autonomously within guardrails you set in advance: spending limits, approved merchants, risk thresholds.
How It Works
An agentic commerce transaction has several moving parts:
- Intent interpretation: The agent parses what you want — not just keywords, but constraints like budget, timing, and quality preferences.
- Negotiation and selection: The agent evaluates options across sellers, sometimes negotiating with other AI agents on the seller side.
- Policy and controls: Pre-defined guardrails — spending caps, merchant allowlists, fraud monitoring — keep the agent within bounds.
- Payment settlement: The transaction settles through payment rails that support machine-speed settlement, including card networks, bank transfers, stablecoins, or smart contracts.
The Stack That Enables It
Three layers of infrastructure make agentic commerce possible:
Identity and trust. Before an AI agent can spend your money, the system needs to verify it is acting on your behalf. Visa’s Trusted Agent Protocol (TAP), co-developed with Cloudflare, is one cryptographic standard for verifying agent identity and blocking malicious bots.
Authorization and controls. Programmable spending rules let businesses and consumers set limits directly in the agent’s code. Mastercard’s Agent Pay for Machines uses network tokenization to link AI agents to specific spending limits instead of exposing raw financial data.
Settlement rails. The payment itself can route through traditional card networks, bank-to-bank transfers, or programmable stablecoins. Stripe’s Agentic Commerce Protocol (ACP), co-developed with OpenAI, provides an open standard for these flows and is being adopted by major AI platforms.
Why It Matters
McKinsey projects global agentic commerce could reach $3 trillion to $5 trillion by 2030. The economic shift is significant: when machines buy from machines, the friction of human checkout — passwords, form fields, approval waits — disappears. That unlocks always-on, high-frequency, low-value transactions that would be impractical for humans to execute manually.
But the scale also raises hard questions about accountability. If an AI agent makes a bad purchase, who is liable? The user who set the guardrails, the agent’s developer, or the payment network? Regulatory frameworks like the GENIUS Act are beginning to address digital payment oversight, but agentic commerce sits at the intersection of payments, AI governance, and consumer protection — a space where standards are still being written.