Retailers rarely lose on strategy. They lose on execution—and execution, almost always, comes down to data moving at the wrong speed, between the wrong systems, at the wrong time.

Imagine a mid-size fashion retailer preparing for a major promotional event. Their pricing engine has the right markdown logic. Their e-commerce platform is ready. But the inventory system—still running on a nightly batch feed—doesn't reflect the afternoon's sell-through data. By the time the AI-driven reorder recommendation fires, the stockout has already happened online. A competitor with real-time inventory visibility captures the sale instead.

This is not a hypothetical edge case. It is the daily reality for retailers managing disconnected integration landscapes—and it is the single biggest reason why AI investments in retail fail to deliver their potential.

Why integration is retail's AI bottleneck

AI models are only as good as the data they can access, and only as fast as the systems that carry it. Retail environments are among the most complex in enterprise IT: point-of-sale systems, e-commerce platforms, loyalty engines, ERP, warehouse management, supplier networks, and an expanding portfolio of SaaS applications—all generating data, all needing to talk to each other in real time.

Fragmented integration such as point-to-point connections, siloed middleware, batch pipelines refreshing every few hours does not just slow AI down, it quietly degrades it. Personalization arrives after the moment has passed. Demand sensing becomes pattern-matching on yesterday's data. Dynamic pricing is neither dynamic nor a competitive advantage when every competitor's system updated before yours did. AI models are only as intelligent as the data flowing into them, and stale data produces confident answers to questions the business stopped asking hours ago.

The road to AI in retail is, first and foremost starts with integration.

What retailers are actually trying to solve

Retail integration challenges cluster around four persistent pain points:

• Real-time omnichannel consistency: Customers expect the same inventory truth whether they are browsing online, checking the app, or asking a store associate. When pricing and stock data flow through disconnected pipes at different speeds, consistency breaks down—and with it, customer trust.
• Supply chain agility: Tariff volatility, shifting supplier relationships, and unpredictable demand require integration that can propagate change signals instantly—from a supplier update in an ERP system to a replenishment decision in a warehouse management system without manual intervention in between.
• AI-powered personalization at scale: Delivering personalized offers in real time requires unified customer data: transaction history, browse behavior, loyalty status, and current basket—all available to the AI engine at the moment of interaction.
• Ecosystem complexity: Modern retail runs on an ecosystem of partners, marketplaces, logistics providers, and fintech services. Each requires a reliable, governed integration that can scale without becoming a maintenance burden.

These are not new problems. What is new is the expectation—from customers, boards, and competitive markets—that they get solved now, not over a multi-year roadmap.

What connected retail looks like in practice

Consider another scenario: a grocery retailer operating a buy-online-pickup-in-store (BOPIS) service. A customer places an order at 9:00 AM. In a fragmented environment, that order touches six systems—e-commerce, OMS, WMS, ERP, loyalty, and in-store fulfillment— each with its own integration pattern, latency, and error handling. The 30-minute pickup promise becomes a 90-minute reality.

Now imagine the same retailer with an event-driven integrated processes. The moment the order is placed, an event fires across all connected systems simultaneously. Stock is reserved, the picker is notified, the loyalty points are staged, and the customer receives a real-time status update—all within seconds. The 30-minute promise holds.

This is exactly what Elkjop Nordic, one of Europe's largest consumer electronics retailers, built with SAP Integration Suite. Elkjop processes 5 million messages daily through SAP Integration Suite, including Advanced Event Mesh, enabling real-time alignment of in-store pricing with local promotions, live order status updates, and omnichannel order fulfillment. The result: online orders fulfilled for in-store pickup within 30 minutes. They also reduced integration complexity significantly—supporting 500 integrations more efficiently by streamlining tenants. According to Mirko Adamovic, Team Lead for Integration at Elkjop: "We can now distribute events that help us deliver a unified retail experience where customers can order online and pick up in a store within just 30 minutes." ¹

The capability that made this possible was not just technology—it was the architectural shift from batch-based, siloed connectivity to a real-time, event-driven integration fabric.

Legacy modernization: the foundation that unlocks AI

Before retailers can pursue AI ambitions, most face a more pressing challenge: a legacy integration landscape that was never designed for speed or scale. Many are still running solutions built for a different era of IT—capable enough for their original purpose, but not for real-time retail or AI workloads.  Examples include:

• Others have accumulated a patchwork of point-to-point connections between ERP, WMS, and e-commerce platforms, each built by a different team at a different time to solve a specific problem, with no shared data model and no central visibility.
• Middleware tools that were state-of-the-art a decade ago—on-premise ESBs, file-based EDI hubs, custom-built API gateways—now sit at the center of critical retail processes but cannot be easily upgraded without business risk.
• Supplier connectivity still runs on batch EDI cycles that settle overnight, at the exact moment that real-time supply chain decisions demand intraday data.
• In-store systems talk to headquarters through scheduled extracts rather than live event streams. And when a retailer has grown through acquisition, the integration landscape often reflects that history directly: three ERPs, two loyalty platforms, and five different middleware tools that were each the right answer for the business unit that bought them but collectively form a fragmented estate that no single team fully understands.

Harrods, the iconic London luxury department store, faced this exactly.² Migrating from SAP Process Orchestration to SAP Integration Suite while maintaining business continuity, Harrods took a phased hybrid migration approach—running both platforms simultaneously during the transition. The result: 60% of existing content reused from the Enterprise Services Repository, and a 40% reduction in total cost of ownership across the integration landscape.

For LL Flooring³ (formerly Lumber Liquidators), the migration was equally structured: 200+ interfaces moved from SAP PI/PO to SAP Integration Suite, integrating 23 systems including CRM, ECC, Gateway, and Fiori. The outcome was a modernized integration backbone capable of supporting a Customer 360 view — a prerequisite for any meaningful AI personalization initiative.

Both stories illustrate the same truth: integration modernization is not a prerequisite that competes with AI investment. It is the investment that makes AI possible.

The AI layer: what SAP Integration Suite enables next

SAP Integration Suite is not standing still while retailers modernize. The platform is building the AI capabilities that transform integration from infrastructure into intelligence, for example:

• Joule-powered integration flow generation (on roadmap)⁴—allows developers to describe an integration in natural language and generate the flow automatically, cutting build time from days to hours. For a retailer onboarding a new marketplace partner or spinning up a seasonal promotion pipeline, this changes the economics of integration development entirely.
• APIs as MCP servers enables SAP-managed APIs to be consumed directly by AI agents—meaning that a retail AI agent can query live inventory, trigger replenishment, or check supplier availability without a custom integration build. It turns the integration layer into an AI action layer.
• Joule conversational analytics for API Management (on roadmap) gives retail operations teams natural language access to integration performance data: which flows are healthy, which APIs are under load, where latency is creeping up—all without a dashboard or a data analyst.

Together, these capabilities describe a future where integration is not a project that precedes AI—it is the substrate through which AI operates continuously.

The competitive reality

Retailers who invest in integration consolidation and modernization today are not just reducing IT cost—they are building the real-time data fabric that AI requires to work. The retailers who will win the next five years of omnichannel competition are, in large part, those who are making this shift.

Companies struggling to manage fragmented integration landscapes will not be slow to adopt AI. They will be slow because their data moves too slowly, their systems talk too rarely, and their integration teams are too busy keeping legacy pipes running to build anything new.

The path forward is clear. It starts with integration.