Thorsten Meyer | ThorstenMeyerAI.com | February 2026
Executive Summary
McKinsey projects $3–5 trillion in global agentic commerce revenue by 2030. Morgan Stanley estimates $190–385 billion in US e-commerce spending by AI agents alone, capturing 10–20% of market share. The agentic AI retail market: $60.43 billion in 2026, growing at 29.29% CAGR to $218.37 billion by 2031. 23% of Americans made an AI-assisted purchase in the past month. 1 in 5 Cyber Week 2025 orders involved an agent.
A parallel web is being built — not for human browsing, but for machine-readable commerce. Google, Shopify, Walmart, Target, and 20+ partners launched the Universal Commerce Protocol (UCP) at NRF 2026. Visa introduced the Trusted Agent Protocol with 10+ partners. Mastercard is deploying Agent Pay across Latin America. The Linux Foundation’s A2A protocol has 150+ supporting organizations. The infrastructure layer for autonomous agent transactions is being built in real time — by the largest payment networks, retailers, and technology platforms on earth.
The strategic question is no longer whether agents will transact. It is who controls the rails, who owns the identity layer, and who captures the margin when agents replace the shopping cart.
| Metric | Value |
|---|---|
| Agentic commerce revenue (2030, global) | $3–5T (McKinsey) |
| US agent e-commerce spending (2030) | $190–385B (Morgan Stanley) |
| Agent share of US e-commerce (2030) | 10–20% (Morgan Stanley) |
| Agentic AI retail market (2026) | $60.43B |
| Agentic AI retail market (2031) | $218.37B (29.29% CAGR) |
| Americans: AI-assisted purchase (past month) | 23% |
| Cyber Week 2025: agent-involved orders | 1 in 5 |
| Black Friday 2025: AI-assisted purchases | 1 in 6 |
| UCP partners (Google/Shopify) | 20+ (Walmart, Target, Etsy, Wayfair, Adyen, AmEx) |
| A2A supporting organizations | 150+ (Linux Foundation) |
| Visa Trusted Agent Protocol partners | 10+ (Oct 2025) |
| Mastercard Agent Pay | Launching LatAm/Caribbean 2026 |
| OECD unemployment (Dec 2025) | 5.0% (stable) |
| Youth unemployment (OECD) | 11.2% |
| OECD jobs: high automation risk | 27% |
Building Scalable eCommerce Platforms with Saleor: The Complete Guide for Developers and Engineers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
1. The Parallel Web Is Being Built
The internet was designed for humans browsing pages. The parallel web being constructed in 2026 is designed for agents executing transactions. The difference is architectural, not incremental.
What Is Changing
| Layer | Human Web | Agent Web |
|---|---|---|
| Discovery | Search engines, browsing | Machine-readable catalogs, API-first access |
| Decision | UI comparisons, reviews | Structured data, preference-matched optimization |
| Transaction | Shopping carts, manual checkout | Tokenized credentials, delegated authorization |
| Payment | Card-on-file, manual entry | Network-issued tokens, programmable rails |
| Identity | Username/password, cookies | Cryptographic identity, scoped permissions |
| Trust | Brand recognition, reviews | Protocol-verified agent credentials |
This is not a feature upgrade to existing e-commerce. It is a new commerce infrastructure layer where the primary customer is a machine acting on behalf of a human — and the machine never sees a webpage.
The Protocol Landscape
| Protocol | Owner | Function |
|---|---|---|
| UCP (Universal Commerce Protocol) | Google + Shopify + 20+ partners | End-to-end agentic commerce standard |
| A2A (Agent-to-Agent) | Linux Foundation (150+ orgs) | Agent coordination and communication |
| MCP (Model Context Protocol) | Anthropic | Structured access to tools, data, systems |
| AP2 (Agent Payments Protocol) | Google Cloud | Secure agentic payment authorization |
| Trusted Agent Protocol | Visa + 10+ partners | Bot vs legitimate agent verification |
| Agent Pay | Mastercard | Agentic commerce payment rails |
| x402 / ERC-8004 | Crypto ecosystem | Machine-to-machine micropayments |
Seven major protocols launched or expanded in the past six months. The infrastructure race is live.
“The parallel web does not replace the human internet. It runs alongside it — and it transacts faster, cheaper, and at scales no human shopping experience can match.”
cryptographic identity tokens
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
2. Strategic Consequences: Who Wins the Agent Commerce Stack
The parallel web creates three structural power shifts that enterprise and public-sector leaders must understand now.
Power Shift 1: From UI to API
When agents bypass the browsing experience entirely, every dollar invested in website design, brand display, and checkout optimization faces a different ROI calculus. The competitive surface shifts from “how does our site look?” to “how machine-readable is our catalog?” Merchants that cannot be discovered by agents are invisible to agent-mediated commerce.
Power Shift 2: From Brand Loyalty to Protocol Access
Agents optimize for structured criteria — price, availability, delivery speed, return policy — not brand affinity. Brand matters when humans browse. Protocol access matters when agents transact. The merchant that integrates UCP first captures agent traffic before the merchant that has the better logo.
Power Shift 3: From Payment Networks to Payment Intelligence
Visa, Mastercard, Google, and Shopify are not building payment rails — those exist. They are building payment intelligence layers: delegated authorization, tokenized credentials, programmable spending controls, and Know Your Agent (KYA) verification. The margin is shifting from transaction processing to transaction governance.
| Shift | From | To |
|---|---|---|
| Discovery | SEO, ads, brand | Machine-readable catalogs, API access |
| Loyalty | Brand affinity | Protocol integration, data quality |
| Payment margin | Transaction processing | Transaction governance, intelligence |
| Trust | Reviews, ratings | Cryptographic verification, KYA |
| Checkout | Human conversion funnel | Delegated agent authorization |
“The most important customer experience investment in 2026 is not your website redesign. It is your machine-readable product catalog.”
THE INVISIBLE STORE: Mastering the Universal Commerce Protocol (UCP): The Blueprint for the AI-Agent Economy and the High-Ticket Strategy to Sell When No One is Browsing.
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
3. OECD Constraints and Public-Sector Implications
The parallel web is not just a commerce story. It operates within the same labour market and social context that constrains every AI deployment.
Labour Market Context
| OECD Signal | Value | Agent Commerce Implication |
|---|---|---|
| Unemployment (Dec 2025) | 5.0% (stable) | No labour surplus — retail transition must be managed |
| Youth unemployment | 11.2% | Entry-level retail/commerce roles face direct agent substitution |
| Jobs: high automation risk | 27% | Commerce workers disproportionately represented |
| Agent market (2026) | $60.43B retail | Scale already material; displacement is active |
| Agent e-commerce (2030) | $190–385B US | Structural shift, not marginal efficiency gain |
Retail and commerce roles — cashiers, customer service, order processing, logistics coordination — are concentrated in the 27% of OECD jobs at high automation risk. Agent commerce does not eliminate these roles overnight, but it structurally reduces the transaction volume that requires human involvement. The transition is uneven: entry-level and customer-facing roles absorb displacement first.
Public-Sector Concerns
- Consumer protection. When an agent makes a purchase on a consumer’s behalf, who is liable for errors, unauthorized transactions, or preference mismatches? Current consumer protection frameworks assume human agency at the point of purchase.
- Competition and market access. If UCP or AP2 become dominant protocols, smaller merchants without technical capacity to integrate face structural exclusion from agent-mediated commerce. The parallel web could concentrate market access, not democratize it.
- Tax and regulatory visibility. Machine-to-machine transactions at speed and volume may outpace current tax collection and regulatory monitoring infrastructure. Governments optimized for human-speed commerce face a compliance gap.
- Digital identity sovereignty. Agent identity protocols (SPIFFE, X.509, Visa Trusted Agent) are being defined by private companies. Governments that do not participate in identity standard-setting lose visibility and control over agent-mediated transactions within their jurisdictions.
agent payment authorization devices
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
4. The Infrastructure Layers: What Is Being Built Now
Layer 1: Machine-Readable Commerce
| Component | What It Means | Who Is Building It |
|---|---|---|
| Structured product catalogs | Agent-parseable inventory, pricing, availability | Shopify (UCP), Google, Walmart |
| API-first checkout | No UI required for transaction completion | UCP, AP2 |
| Real-time inventory feeds | Agents see live availability, not cached pages | Major retailers via UCP integration |
| Preference-matched recommendations | Structured criteria, not visual merchandising | Agent-side optimization |
Layer 2: Payment Rails
| Component | What It Means | Who Is Building It |
|---|---|---|
| Tokenized credentials | Network-issued tokens replace stored card data | Visa, Mastercard, Adyen |
| Delegated authorization | Agents authorized to transact within parameters | AP2, Trusted Agent Protocol |
| Programmable spending controls | Per-agent budgets, category limits, approval thresholds | Payment networks + enterprise |
| Know Your Agent (KYA) | Verification that agent is legitimate, not bot | Visa Trusted Agent Protocol |
Layer 3: Identity and Trust
| Component | What It Means | Who Is Building It |
|---|---|---|
| Cryptographic agent identity | SPIFFE/X.509 certificates per agent | Linux Foundation, enterprise |
| Scoped permissions | Agent can only act within defined parameters | A2A protocol, enterprise governance |
| Action logs and audit trails | Every agent transaction traceable | UCP, enterprise compliance |
| Consent flows | Human approval/override mechanisms | Protocol-level requirements |
Layer 4: Coordination and Governance
| Component | What It Means | Who Is Building It |
|---|---|---|
| Agent-to-agent communication | Agents negotiate, coordinate, transact | A2A (Linux Foundation, 150+ orgs) |
| Policy enforcement | Spending limits, prohibited categories, compliance | Enterprise governance layers |
| Exception routing | Escalation to human when agent encounters edge case | Protocol-level + enterprise |
| Cross-protocol interoperability | UCP + A2A + MCP + AP2 working together | Consortium-driven standardization |
“The parallel web is not one protocol. It is four infrastructure layers — discovery, payment, identity, and coordination — all being built simultaneously by different consortiums with different incentives.”
5. Enterprise Playbook: Preparing for Agent-Mediated Commerce
For Retailers and Merchants
| Priority | Action | Timeline |
|---|---|---|
| 1. Machine-readable catalogs | Implement UCP or equivalent structured data | Q1–Q2 2026 |
| 2. API-first checkout | Enable transaction completion without UI | Q2 2026 |
| 3. KYA integration | Distinguish legitimate agents from bots | Q2–Q3 2026 |
| 4. Agent analytics | Track agent-mediated vs human traffic, conversion, basket size | Q3 2026 |
| 5. Pricing strategy review | Account for agent comparison optimization | Ongoing |
For Payment Networks and FinTech
| Priority | Action | Timeline |
|---|---|---|
| 1. Tokenization infrastructure | Network-issued tokens for agent transactions | Active |
| 2. Delegated authorization APIs | Enable scoped agent spending permissions | Q1–Q2 2026 |
| 3. Fraud detection for agents | Agent-specific fraud patterns vs human patterns | Q2 2026 |
| 4. KYA verification services | Sell agent identity verification as service | Q3 2026 |
| 5. Machine-to-machine settlement | Real-time settlement for agent-speed transactions | 2026–2027 |
For Enterprises (Buyer Side)
| Priority | Action | Timeline |
|---|---|---|
| 1. Agent procurement policies | Define what agents can purchase, spending limits, categories | Q1 2026 |
| 2. Identity and credential management | Scoped agent identities, not shared credentials | Q2 2026 |
| 3. Audit and compliance infrastructure | Every agent transaction logged, reviewable, auditable | Q2 2026 |
| 4. Budget governance | Token-level or transaction-level spending controls | Q2–Q3 2026 |
| 5. Vendor protocol assessment | Evaluate UCP, AP2, A2A readiness of supply chain | Ongoing |
6. Practical Actions for Leaders
1. Audit your machine-readability now. Can an AI agent discover your products, check availability, and complete a transaction without touching a webpage? If not, you are invisible to the fastest-growing commerce channel. 1 in 5 Cyber Week orders already involved an agent.
2. Integrate with at least one major protocol by Q3 2026. UCP (Google/Shopify), AP2 (Google Cloud), Trusted Agent Protocol (Visa), or Agent Pay (Mastercard). Protocol access is the new market access. Merchants outside these protocols lose agent-mediated traffic.
3. Build Know Your Agent (KYA) capability. Distinguish legitimate purchasing agents from bots and scrapers. Bot-based payment attacks are surging. The identity verification layer is security infrastructure, not optional.
4. Establish agent spending governance. For enterprises deploying purchasing agents: per-agent budgets, category restrictions, approval thresholds, and full audit trails. The governance infrastructure from agentic operations (article #39) applies directly to agentic commerce.
5. Engage in identity standard-setting. Agent identity protocols are being defined now by Visa, Google, Linux Foundation, and others. Organizations and governments that do not participate will accept standards designed by others — and the sovereignty implications are real.
| Action | Owner | Timeline |
|---|---|---|
| Machine-readability audit | CTO + Commerce | Q1 2026 |
| Protocol integration (UCP/AP2) | CTO + Digital | Q2–Q3 2026 |
| Know Your Agent (KYA) | CISO + Commerce | Q2 2026 |
| Agent spending governance | CFO + Procurement | Q2 2026 |
| Identity standard engagement | Legal + CTO + Policy | Ongoing |
What to Watch
Whether UCP becomes the dominant agentic commerce standard. Google, Shopify, Walmart, Target, Etsy, Wayfair, Adyen, American Express, and 20+ partners create significant gravity. The question is whether competing protocols (x402, Mastercard Agent Pay) coexist or fragment the market — and what that means for merchants forced to support multiple standards.
Government participation in agent identity standards. The Trusted Agent Protocol, SPIFFE, and KYA frameworks are being defined by private consortiums. Governments that engage now shape the rules. Those that wait accept them. The sovereignty implications of agent identity — who can verify, who can transact, who can monitor — are as significant as the sovereignty implications of digital currency.
The transition impact on retail workforce. $190–385 billion in agent-mediated US e-commerce by 2030 (Morgan Stanley) structurally reduces human transaction touchpoints. With 27% of OECD jobs at high automation risk and retail workers disproportionately represented, workforce transition planning is a near-term policy requirement, not a long-term scenario.
The Bottom Line
$3–5 trillion in global agentic commerce by 2030 (McKinsey). $190–385 billion in US agent e-commerce (Morgan Stanley). 1 in 5 Cyber Week orders. 23% of Americans purchased via AI. 7 major protocols in six months. 150+ organizations behind A2A. 20+ partners behind UCP. 27% of OECD jobs at high automation risk.
The parallel web is not a prediction — it is infrastructure under construction by the largest payment networks, retailers, and technology platforms on the planet. The organizations that build machine-readable commerce, integrate with agent protocols, and govern agent transactions will capture the margin of the next commerce era. Those that wait for the human web to remain sufficient will discover that their most important customers are machines they never learned to serve.
The next trillion-dollar commerce platform will not have a homepage.
When agents become the primary shoppers, the competitive advantage is not the best website — it is the best machine-readable catalog, the most trusted agent identity, and the most governed transaction rail.
Thorsten Meyer is an AI strategy advisor who notes that the phrase “build it and they will come” assumes “they” are humans who can appreciate your font choices. Agents cannot. More at ThorstenMeyerAI.com.
Sources
- McKinsey — $3–5T Global Agentic Commerce by 2030
- Morgan Stanley — $190–385B US Agent E-Commerce by 2030 (10–20% Share)
- Mordor Intelligence — $60.43B Agentic AI Retail 2026, $218.37B by 2031 (29.29% CAGR)
- Ekamoira/Industry Reports — 23% Americans AI-Assisted Purchase; 1 in 5 Cyber Week
- Google/Shopify — Universal Commerce Protocol (UCP) Launch, NRF 2026
- Linux Foundation — A2A Protocol, 150+ Organizations
- Visa — Trusted Agent Protocol, 10+ Partners (Oct 2025)
- Mastercard/Axios — Agent Pay, Latin America/Caribbean 2026
- Google Cloud — Agent Payments Protocol (AP2)
- Anthropic — Model Context Protocol (MCP)
- CoinDesk/Crypto Ecosystem — x402, ERC-8004 Machine Payments
- PYMNTS — Tokenization, Delegated Authorization, KYA
- Citibank — Three Foundations: Identity, Money Flow, Interoperability
- Linux Foundation — Agentic AI Foundation (Anthropic, Block, Google, Microsoft, OpenAI)
- OECD — 5.0% Unemployment, 11.2% Youth (Feb 2026)
- OECD — 27% Jobs at High Automation Risk
- SAP — Agentic Commerce for Retailers (NRF 2026)
- Microsoft — Agentic Commerce as New Front Door to Retail (Feb 2026)
© 2026 Thorsten Meyer. All rights reserved. ThorstenMeyerAI.com
