Enterprise Agentic Commerce: How SAP, Oracle, Magento, and Shopify Plus Are Preparing for the AI Agent Economy

Executive Summary

Enterprise B2B commerce is approaching a structural discontinuity. By 2027, Gartner projects that 30% of B2B procurement decisions will be initiated or negotiated by AI agents — not humans. These agents will browse catalogs, request quotes, negotiate pricing, approve purchase orders, and reconcile invoices across ERP systems. The $20 trillion global B2B commerce market is about to be rewired around machine-to-machine commerce.

We analyzed five enterprise commerce platforms — SAP Commerce Cloud, Oracle Commerce, Adobe Commerce (Magento), Shopify Plus, and WooCommerce Enterprise — across seven dimensions critical to AI agent readiness: API architecture, structured data maturity, ERP integration depth, approval workflow programmability, search/discovery APIs, B2B-native feature coverage, and MCP/agentic protocol support.

Our findings reveal a market bifurcated between platforms built for human shoppers being retrofitted for agents (the majority) and a small cohort of platforms whose architecture is genuinely agent-first. The gap between "can integrate with AI" and "was built for AI agents" is wider than most enterprise buyers realize — and it will define winners and losers over the next 36 months.

Key Findings for Enterprise Decision-Makers

1. SAP Commerce Cloud leads on ERP-native integration — but its agent readiness depends heavily on SAP BTP and the speed of S/4HANA migration. The IDoc-to-agent bridge is the critical path.

2. Oracle Commerce has the strongest AI-native search and recommendation stack — CX Unity and adaptive search give it a head start for product discovery agents. ERP integration via Oracle Integration Cloud is capable but complex.

3. Adobe Commerce (Magento) is architecturally slowest to adapt — deep legacy coupling makes B2B workflow programmability the hardest among the five. Sensei AI is catalog-focused, not procurement-focused.

4. Shopify Plus is the dark horse — B2B wholesale native, GraphQL-first API, and the fastest pace of AI feature shipping. Its weakness is ERP depth; strength is agent-to-checkout speed.

5. WooCommerce Enterprise + MCP offers the most flexible bridge — open-source + headless + MCP protocol creates an adaptable agentic layer that proprietary platforms struggle to match.

Part 1: The B2B Commerce AI Landscape — Why Enterprise Is Different

It's tempting to view "AI commerce" as a B2C problem — AI assistants recommending sneakers or comparing phone prices. That framing misses the largest economic opportunity by an order of magnitude. B2B commerce is fundamentally different, and those differences make it both harder and more valuable for AI agents to penetrate.

What Makes B2B Commerce Structurally Different for AI Agents

B2B commerce has five structural characteristics that B2C AI integrations never need to address:

Complex procurement workflows: Purchase orders, multi-level approvals, budget codes, cost centers — a single B2B transaction may require 6–12 approval steps before a purchase order is generated. An AI agent must navigate this workflow graph, not just recommend a product.
Negotiated pricing: B2B pricing is rarely list-price. Contract pricing, volume tiers, customer-specific catalogs, and quote-to-cash cycles mean the "price" an AI agent sees depends on who's asking. The system must authenticate the buyer organization and apply the correct pricing matrix before the agent can act.
Multi-vendor RFQ/RFP processes: Enterprise buyers don't click "add to cart." They issue RFQs to multiple suppliers, compare quotes, and negotiate terms. AI agents operating in this environment must be able to generate, submit, and compare RFQs programmatically.
ERP entanglement: B2B commerce doesn't end at checkout. It flows into inventory management, accounts payable, supplier portals, and financial reconciliation. An AI agent that can't reach into the ERP is an AI agent that can't close a B2B transaction.
Reordering logic: A significant portion of B2B volume is replenishment — reordering raw materials, components, or supplies on a schedule. AI agents that understand consumption patterns, lead times, and safety stock can create more value than any product recommendation engine.

The implication: A platform that handles B2C AI commerce well doesn't necessarily handle B2B AI commerce at all. The architectural requirements are fundamentally different. A GraphQL product endpoint that works perfectly for a DTC brand is useless to an enterprise buyer whose transaction requires SAP IDoc-level integration, three approval signatures, and a contract price lookup.

The Agent Readiness Gap

Across all five platforms we analyzed, we assessed readiness across seven dimensions on a 1–5 scale:

Dimension	SAP	Oracle	Adobe/Magento	Shopify Plus	WooCommerce
API Architecture	4	4	3	5	4
Structured Data Maturity	5	4	3	4	3
ERP Integration Depth	5	5	2	3	2
Approval Workflow Programmability	4	3	3	4	3
Search/Discovery API	3	5	4	4	2
B2B-Native Features	5	4	4	4	2
MCP/Agentic Protocol Support	1	1	1	2	4

The most striking finding: no platform scores above 3 on MCP/agentic protocol support — except WooCommerce, which achieves it through the open-source ecosystem, not through the platform itself. This is the gap that will define 2027–2028 vendor strategies.

Part 2: Platform-by-Platform Analysis

SAP Commerce Cloud

Agent Readiness: High

AI Capabilities Today: SAP embeds AI through its Business AI portfolio — Joule (the generative AI copilot integrated across SAP applications), AI-assisted product content generation, and intelligent catalog management. SAP Commerce Cloud benefits from the broader SAP AI ecosystem, including predictive replenishment and demand forecasting that uses machine learning models trained on S/4HANA data.

ERP Integration: This is SAP's structural advantage. Commerce Cloud is designed to run alongside S/4HANA, with real-time order orchestration, pricing synchronization, and inventory visibility. The OData API layer — exposed through SAP Cloud Integration and API Management — provides structured, RESTful access to commerce data. For AI agents, this means product data, pricing (including contract pricing from SAP SD), and inventory levels are accessible through versioned, documented APIs.

Agent Readiness Assessment: SAP's strength is data depth, not agent accessibility. The OData APIs are powerful but designed for B2B integration middleware, not autonomous AI agents. An agent querying SAP Commerce Cloud would need to understand SAP Business Objects, navigate complex OData entity relationships, and handle SAP-specific authentication flows. The critical gap: no native AI agent protocol. SAP's strategy appears to route agent interactions through Joule (SAP's own copilot) rather than opening the commerce layer to third-party agents. This is a vendor-lock strategy — effective for existing SAP accounts, limiting for multi-platform enterprises.

Recommended for: SAP-centric enterprises running S/4HANA who can commit to the SAP BTP ecosystem.

Oracle Commerce

Agent Readiness: Medium-High

AI Capabilities Today: Oracle Commerce has the most mature AI-native search stack of any platform analyzed. Oracle Adaptive Search uses machine learning to personalize product results, ranking, and recommendations in real time. The Oracle CX Unity data platform centralizes customer data across commerce, marketing, and service — giving AI agents a unified view of buyer accounts, entitlements, and contract terms. Oracle's recommendation engine has been tuned over a decade of enterprise deployments.

ERP Integration: Oracle Commerce integrates with Oracle Fusion Cloud ERP (and on-prem E-Business Suite) through Oracle Integration Cloud. Pre-built adapters handle order-to-cash, procure-to-pay, and inventory synchronization. The REST APIs are well-documented, and Oracle's OAuth 2.0-based security model is agent-compatible. For enterprises running both Oracle Commerce and Oracle ERP, the integration path is the most turnkey in the market.

Agent Readiness Assessment: Oracle's advantage is its AI-native search and personalization — the product discovery layer where most AI agents will start. An agent looking for "industrial-grade stainless steel valves, ANSI Class 300, with 2-week lead time" can get a structured response from Oracle Commerce's search API. The limitation is on the transactional side: approval workflows, purchase order generation, and contract price negotiation are handled in Oracle ERP, not Oracle Commerce. An agent must cross the commerce-to-ERP boundary for every transaction — Oracle Integration Cloud makes this possible but adds latency and complexity.

Recommended for: Oracle ERP enterprises prioritizing AI-powered product discovery with strong ERP alignment.

Adobe Commerce (Magento)

Agent Readiness: Medium

AI Capabilities Today: Adobe Commerce delivers AI through Sensei — Adobe's AI/ML platform — focusing on product recommendations, Live Search (powered by Adobe Sensei), and visual similarity search. Product recommendations can be deployed across category pages, product detail pages, and the cart. Live Search provides typo-tolerant, faceted search with merchandising rules. These are solid B2C-and-light-B2B features but do not extend to procurement-grade AI capabilities.

ERP Integration: This is Adobe Commerce's structural weakness for enterprise B2B. Magento has no native ERP. Integration with SAP, Oracle, NetSuite, or Microsoft Dynamics requires third-party middleware (e.g., Celigo, MuleSoft, Boomi) or custom development using Magento's REST/SOAP APIs. The B2B module (introduced in Magento 2.2) adds company accounts, shared catalogs, requisition lists, and quote management — but programmatic access to these features through clean, modern APIs remains inconsistent.

Agent Readiness Assessment: Adobe's acquisition of Magento was a content-and-experience play, not an ERP play. The platform is strong on front-end personalization but weak on back-office integration — exactly the reverse of what B2B AI agents need. The GraphQL coverage for B2B features (purchase orders, approval rules, requisition lists) is incomplete compared to the B2C GraphQL surface. An AI agent trying to submit a purchase order through the Magento API would need to navigate a mix of GraphQL, REST, and SOAP endpoints with inconsistent data models. This is the platform most in need of an agentic middleware layer.

Recommended for: Mid-market B2B with simpler procurement, or enterprises willing to build custom agent middleware.

Shopify Plus

Agent Readiness: High

AI Capabilities Today: Shopify has been the fastest-moving platform on AI. Shopify Magic (2024) introduced AI-generated product descriptions, AI-powered image editing, and AI customer service (Sidekick). In 2025–2026, Shopify expanded into AI-powered commerce with automated merchandising, AI search tuning, and AI-assisted checkout optimization. Shopify's pace of AI feature shipping exceeds every other platform on this list.

B2B Features: Shopify Plus's B2B wholesale channel (launched 2022, expanded significantly through 2025) provides company profiles, customer-specific catalogs and pricing, quantity rules, payment terms, and purchase order-based checkout. All B2B features are accessible through the Storefront API (GraphQL) and Admin API (GraphQL + REST). This is the cleanest API surface of any platform: if it exists in the Shopify admin, it exists in the API. Shopify Functions enable custom business logic at the edge (discounts, shipping, payment), which could be routed for agent-driven decisions.

Agent Readiness Assessment: Shopify Plus is the most agent-ready commerce platform on API architecture alone. The GraphQL-first design means AI agents can query exactly the data they need — no over-fetching, no undocumented endpoints, no SOAP legacy. The B2B APIs are comprehensive and versioned. The gap? ERP depth. Shopify doesn't own the ERP layer. For Shopify, the ERP is always third-party (NetSuite, Acumatica, Dynamics). An AI agent can complete the commerce transaction (quote → order → payment) on Shopify natively, but cannot reconcile that transaction into the ERP without a separate integration step. For some enterprise buyers, this is acceptable. For others — particularly in manufacturing and heavy industry — it's a dealbreaker.

Recommended for: Speed-focused enterprises; DTC brands expanding to B2B; wholesale-first businesses needing fast agent-to-checkout.

WooCommerce Enterprise

Agent Readiness: Medium (Adaptable)

AI Capabilities Today: WooCommerce itself has minimal built-in AI. The WordPress ecosystem provides plugins for AI-powered product recommendations, AI search, and AI content generation — but these are point solutions, not platform-native capabilities. Where WooCommerce excels is extensibility: the REST API is comprehensive, headless deployments (via WPGraphQL or custom REST endpoints) are straightforward, and the open-source architecture means nothing is locked behind proprietary data silos.

B2B Features: WooCommerce B2B depends on plugins — B2BKing, Wholesale Suite, or custom development — which add company accounts, tiered pricing, minimum order quantities, request-a-quote, and purchase order-based checkout. This plugin dependency is both a weakness (no single source of truth for B2B APIs) and a strength (you can build exactly the API surface your AI agents need). Headless WooCommerce deployments using React/Next.js as the front end are increasingly common in enterprise contexts.

MCP/Agentic Protocol: This is where WooCommerce diverges sharply from proprietary platforms. The MCP (Model Context Protocol) ecosystem — tools, resources, and prompts that let AI models interact with external systems — can be layered onto WooCommerce's REST API with minimal friction. Shop2LLM and similar tools already provide llms.txt, JSON-LD product schema, and MCP-compatible product search endpoints for WooCommerce. No other platform on this list has a functioning open-agentic bridge ready today. WooCommerce does, because the open-source community built it.

Recommended for: Enterprises that value API flexibility over turnkey integration; open-source-first organizations; companies building custom agentic commerce layers.

Part 3: The AI Agent Integration Challenge

An AI agent performing a B2B purchase is not like a chatbot answering a FAQ. It's a multi-step, stateful workflow crossing commerce, ERP, and financial systems. Understanding how agents must navigate these steps reveals the deepest architectural requirements of each platform.

What a Complete B2B Agent Transaction Looks Like

Consider an AI agent tasked with reordering 500 units of a specific industrial component for a manufacturing facility:

Authenticate the buyer organization — The agent must present credentials that identify the buying organization, not just an individual user. This requires company-account-level authentication, not just customer login.
Resolve contract pricing — The agent must look up the contract (not list) price for this buyer. This may require querying a separate pricing engine or ERP module. The product exists in the commerce platform; the contract price exists in the ERP. The agent needs both.
Check inventory and lead time — Real-time ATP (Available-to-Promise) data, not cached catalog data. For large orders, this may trigger a backorder workflow or suggest an alternative supplier.
Validate purchase order requirements — Some organizations require a PO number before an order can be placed. The agent must generate a PO number (from the ERP numbering scheme) or prompt the user to provide one.
Navigate approval workflow — If the order exceeds a threshold ($5,000, $50,000, etc.), the agent must route it for approval. This requires understanding the organization's approval matrix (by amount, department, cost center).
Place the order — Submit the order to the commerce platform, receive an order confirmation number, and log the transaction in the buying organization's procurement system.
Confirm ERP reconciliation — The order must appear in the ERP's purchase order register. The agent should verify this reconciliation has occurred before considering the transaction complete.

Steps 1, 2, 3, and 6 are possible on most enterprise commerce platforms today. Steps 4, 5, and 7 require ERP-level access that few platforms expose to external agents. This is the agentic commerce bottleneck.

The Three-Layer Agent Architecture

Layer 1: Commerce API — Product search, catalog access, cart/order submission. Available on all five platforms in varying quality.

Layer 2: Workflow Engine — Approval routing, quote management, contract lookup. Partially available on SAP and Shopify Plus; fragmented on Oracle and Magento.

Layer 3: ERP Bridge — Purchase order creation, inventory ATP, financial reconciliation, supplier management. Only SAP and Oracle offer native ERP bridges; Shopify Plus and Magento require third-party middleware.

An AI agent that can only access Layer 1 can recommend products but cannot close transactions. Full agentic commerce requires all three layers.

How Each Platform Handles These Steps

Agent Step	SAP	Oracle	Adobe/Magento	Shopify Plus	WooCommerce
Buyer Auth	✓ Native	✓ Native	✓ B2B Module	✓ Company Accounts	⚠ Plugin
Contract Pricing	✓ SAP SD	✓ CPQ Cloud	⚠ Shared Catalogs	✓ Price Lists	⚠ Plugin
ATP Inventory	✓ S/4HANA	✓ Fusion SCM	✕ No Native	✕ No Native	✕ No Native
PO Generation	✓ ERP Native	✓ ERP Native	✕ No Native	⚠ Draft Only	✕ No Native
Approval Routing	✓ Flexible WF	⚠ Approval Rules	⚠ Approval Rules	✓ B2B Logic	⚠ Plugin
Order Placement	✓ Commerce API	✓ Commerce API	✓ GraphQL/REST	✓ GraphQL	✓ REST API
ERP Reconciliation	✓ Auto	✓ Auto	✕ Manual/3rd	✕ Manual/3rd	✕ Manual/3rd

The pattern is clear: SAP and Oracle win on completeness; Shopify Plus wins on API quality; Magento and WooCommerce depend on the ecosystem to fill gaps. The question for enterprise buyers is whether API quality (speed to agent integration) matters more than ERP completeness (depth of agent capability).

Part 4: ERP + AI Agent Bridge — How Agents Connect to the Back Office

ERP is the gravitational center of enterprise B2B. No purchase flows through a commerce platform that doesn't eventually land in an ERP. For AI agents to operate autonomously in B2B commerce, they need a bridge between the commerce layer (where products are discovered and orders placed) and the ERP layer (where orders are validated, fulfilled, and reconciled).

SAP IDoc: The 800-Pound Gorilla

SAP's IDoc (Intermediate Document) format has been the lingua franca of enterprise integration for 30 years. Billions of purchase orders, invoices, and delivery notes flow through IDocs daily. The question is not whether AI agents will need to interact with IDocs — they will — but how.

SAP's strategy is to route this through BTP (Business Technology Platform) and the SAP Integration Suite. The path for an AI agent is:

// AI Agent → OData API → SAP Commerce Cloud → CPI → IDoc → S/4HANA
POST /sap/opu/odata/sap/API_SALES_ORDER_SRV/A_SalesOrder
// Commerce order is created → triggers CPI iFlow → generates ORDERS05 IDoc → 
// S/4HANA processes via Standard Purchase Order processing

This pipeline works today for system-to-system integration. An AI agent that can construct an OData-compliant JSON payload can create a sales order in SAP Commerce Cloud, which will be serialized into an IDoc and processed by S/4HANA. The technical capability exists. The missing piece is an AI-friendly abstraction layer. Raw OData with SAP's entity model is not a developer experience designed for large language models to navigate autonomously.

NetSuite SuiteTalk: SOAP's Unlikely Resilience

NetSuite's SuiteTalk API — a SOAP-based web service — powers more B2B integrations than any REST API in the mid-market. While SOAP is architecturally dated, NetSuite's decision to maintain a comprehensive, well-documented SOAP API means that AI agents can interact with NetSuite through structured, schema-validated XML messages — ironically more predictable for LLMs than some REST APIs with fluid JSON schemas.

// SuiteTalk: AI Agent → SOAP Request → NetSuite → Purchase Order created
<soap:Body>
  <add xmlns="urn:purchases_2021_1.transactions.webservices.netsuite.com">
    <record xsi:type="PurchaseOrder">
      <entity internalId="BUYER_ID"/>
      <tranDate>2026-06-20T00:00:00</tranDate>
    </record>
  </add>
</soap:Body>

The key insight: schema-validated XML is LLM-friendly. The strict XSD definitions mean agents can't hallucinate field names — the schema rejects invalid requests. This makes SuiteTalk a surprisingly viable agentic bridge for the mid-market, especially when combined with RESTlet endpoints for lighter-weight queries.

The MCP Protocol: The Missing Universal Bridge

The Model Context Protocol (MCP), introduced by Anthropic in late 2024 and rapidly adopted for AI-tool integration, represents the most promising architecture for ERP-agent bridging. MCP defines three primitives:

Tools: Functions AI models can invoke — like "searchProducts," "createQuote," "submitPurchaseOrder"
Resources: Structured data the model can access — like product catalogs, price lists, order history
Prompts: Reusable templates for specific workflows — like "requestQuote" or "approveOrder"

An MCP server sitting between an AI agent and an ERP system could abstract away the complexity of IDocs, SuiteTalk, and REST APIs into a clean, LLM-optimized interface. The agent would call submitPurchaseOrder(productId, quantity, costCenter) without needing to know whether the implementation hits OData, SOAP, or a GraphQL endpoint.

As of mid-2026, no major commerce platform offers native MCP support — a striking gap given the protocol's adoption in the AI tool ecosystem. The exception is WooCommerce, where the open-source community (led by plugins like Shop2LLM) has built MCP servers that expose product data and search functions to AI models through the standard protocol.

Prediction: MCP Will Become the Standard ERP-Agent Bridge by 2028

The ERP integration middleware market (Boomi, MuleSoft, Celigo, Workato) is already exploring MCP-native connectors. Once these platforms ship MCP servers for SAP, Oracle, and NetSuite, the ERP-agent bridging problem will shift from "how do we build it" to "how do we secure and govern it." Enterprise CTOs should begin evaluating MCP as an architectural requirement for any agentic commerce initiative starting in 2027.

Part 5: Decision Framework — Which Platform for Which B2B Scenario

There is no universal "best platform." The right choice depends on your B2B commerce model and the speed at which AI agents will enter your procurement flow. We've mapped the five platforms to three archetypal B2B scenarios:

Scenario A: Transactional B2B (High Volume, Low Complexity)

Characteristics: Catalog-based purchasing, standard pricing, quick reorder cycles, minimal approval workflows. Think office supplies, MRO (maintenance, repair, operations), wholesale distribution.

Best Fit: Shopify Plus

Fastest agent-to-checkout pipeline
GraphQL API is agent-optimized
B2B wholesale channel handles catalogs and price lists
Lowest integration overhead for basic B2B workflows
Weakness: ERP reconciliation requires middleware
Best for: Speed and API quality over ERP depth

Also Consider: WooCommerce + MCP

Most flexible agentic layer via open-source
Lower TCO for medium-volume scenarios
Headless deployment options with modern front-end
MCP protocol support through community plugins
Weakness: No native B2B features without plugins

Scenario B: Contractual B2B (Medium Volume, High Complexity)

Characteristics: Negotiated contracts, customer-specific pricing, complex approval chains, multi-year agreements. Think industrial components, raw materials, specialized manufacturing inputs.

Best Fit: SAP Commerce Cloud + S/4HANA

Contract pricing native in SAP SD module
Approval workflows programmatically configurable
ERP reconciliation is automatic, not bolted-on
ATP inventory visibility at enterprise scale
Weakness: Heavy implementation, SAP ecosystem lock-in
Best for: SAP-centric enterprises where ERP integration is non-negotiable

Also Consider: Oracle Commerce + Fusion ERP

Best AI-powered search for complex industrial catalogs
CX Unity provides unified buyer data
CPQ Cloud handles complex pricing scenarios
OIC provides pre-built ERP adapters
Weakness: Similar lock-in concerns as SAP

Scenario C: Marketplace B2B (Multi-Vendor, Platform Model)

Characteristics: Multiple suppliers, aggregated catalogs, buyer-supplier matching, RFQ-driven. Think B2B marketplaces like Alibaba, Amazon Business, industry-specific procurement hubs.

Best Fit: Adobe Commerce (Magento) or Custom

Multi-vendor architecture via extensions
RFQ workflow available with customization
Adobe ecosystem provides content + commerce
GraphQL coverage improving rapidly
Weakness: ERP integration is third-party; B2B APIs incomplete
Best for: Content-rich marketplaces with multi-vendor needs

Also Consider: Shopify Plus + Marketplace Apps

Simpler multi-vendor setup via apps
Clean API for supplier onboarding
Better out-of-box B2B features than Magento
Faster time-to-market
Weakness: Marketplace depth limited vs. Magento

Strategic Recommendations for Enterprise Buyers & CTOs

Based on our analysis, we recommend the following five actions for enterprise technology leaders evaluating their agentic commerce strategy:

Audit your API surface with an agent's perspective. Don't ask "do we have APIs?" Ask "can an LLM navigate our APIs?" Your REST endpoints may be documented for human developers but incomprehensible to an AI model. Conduct an agentic-readiness audit: try to complete a full purchase order flow using only API documentation as context. If your internal team can't do it, an external AI agent definitely can't.
Prioritize ERP integration over commerce platform features. The commerce platform is the agent's storefront — but the ERP is where value is created. A beautiful commerce API that can't create a purchase order in SAP is a glass door. Invest in the integration layer first: SAP CPI, Oracle OIC, MuleSoft, or MCP-based middleware. Commerce features can follow.
Demand MCP or equivalent agent protocol support from vendors. In RFPs and vendor evaluations starting today, ask: "What is your MCP roadmap? How will AI agents authenticate to and navigate your platform? Do you support structured agent tools and resources, or only REST/GraphQL endpoints?" Vendors who can't answer this question will be architecturally behind by 2028.
Build or buy an agentic middleware layer. The platform that scores lowest on our vertical integration assessment — WooCommerce — scores highest on agentic protocol support, precisely because the open ecosystem built the bridge that proprietary platforms haven't. If you're on SAP, Oracle, or Magento, you should be building an MCP server that exposes your commerce and ERP functions through clean, LLM-optimized tools. This is a 3–6 month build, not a year-long platform migration.
Start with replenishment, not procurement. The highest-ROI entry point for enterprise agentic commerce is automated replenishment — AI agents that monitor inventory levels, predict consumption, and place reorders within predefined parameters. This is a contained scope with measurable impact: reduced stockouts, optimized cash-to-cash cycles, and lower manual order processing costs. Full autonomous procurement (RFQs, negotiations, supplier selection) is a 2028+ play.

The 18-Month Window

Enterprise technology leaders have approximately 18 months before AI agents become a meaningful channel for B2B commerce. This is not a 5-year horizon. The infrastructure that supports agents — MCP servers, API gateways, structured data layers, ERP bridges — must be planned, budgeted, and built starting now.

The enterprises that will benefit most from agentic commerce are not the ones with the newest commerce platforms. They are the ones that invested early in making their existing platforms agent-accessible. The architecture is the advantage, not the vendor.