AI isn't enough: Why store infrastructure will determine which retailers scale successfully

AI has significant potential for retailers, but its success hinges on more than just the artificial intelligence itself. The infrastructure that supports AI-driven decisions in physical stores poses a substantial barrier to widespread implementation. While AI can swiftly recommend markdowns, predict stockouts, adjust picking routes, and identify patterns in store traffic, its value is contingent upon accurate execution within the physical store environment.

A Deloitte survey revealed that 75% of retail and consumer products executives consider AI a top strategic priority, yet only 16.5% are able to quantify the return they are receiving from it. The challenge lies in transforming isolated capabilities into consistent, business-wide operations.

Physical retail environments present unique constraints that software teams cannot simply abstract away. Stores are not cloud workloads; they contain thousands of devices, existing networks, and years of installed technology. The buildings themselves can impact connectivity, and as the number of connected endpoints grows, the importance of a reliable wireless foundation becomes increasingly apparent.

Retailers must consider whether their store estate can execute reliably at scale, taking into account factors such as walls, metal shelving, refrigeration, back rooms, loading areas, and legacy systems. Infrastructure changes that appear simple in theory can become costly and disruptive when applied across hundreds of store sites.

For example, inventory management, which relies on shelf sensors, handheld devices, picking tools, and accurate product information, demands a robust network capable of handling diverse data types. Similarly, computer vision, digital displays, and environmental monitoring introduce additional endpoints and data sources. These workloads have varying requirements for bandwidth, latency, and the management of a large number of battery-powered devices.

If a pricing decision cannot reach every store promptly, a replenishment signal may arrive too late, or each new application could require additional costly infrastructure layers, ultimately diminishing the value of the technology.

The challenge extends beyond individual pilots to the broader operational reality of scale. As technology teams add project-specific infrastructure to address coverage gaps, deployment economics become more complex. While the technology may work initially, the operational challenges increase significantly with scale. Retail technology teams must ask not only whether an application meets its business case but also what happens to the underlying infrastructure once the application succeeds.

A deliberate approach to connectivity, rather than treating it as a specification within each procurement, is essential. Different workloads require different wireless foundations suited to their specific needs. While enterprise Wi-Fi may remain the de facto transport layer for some applications, retailers should consider integrating a purpose-built wireless layer alongside the existing network, especially for large or difficult environments and as the number of retail IoT endpoints grows rapidly.

This approach should be deliberate, ensuring that each connectivity layer is scalable, reusable, and able to support the evolving demands of retail AI applications. Security must also be integrated into this decision-making process, with consistent management and open standards playing a crucial role in maintaining a manageable and adaptable architecture.