While large-scale AI deployments are largely non-existent throughout the retail industry, AI use cases are rising every year. According to research from KPMG, 73% of consumer and retail CEOs expect to spend 10% to 20% more on AI in the next 12 months. Also, 64% expect AI to be a top investment priority for their business. The infatuation with AI in retail is understandable. Top decision-makers envision more streamlined business operations such as automated checkouts, real-time inventory adjustments, and automated loss prevention.
Even as the industry works to build and scale these competencies, retail organizations still have much work to do to establish what’s necessary to deploy AI effectively. In fact, the industry has already seen certain companies dial back AI initiatives for core retail functions. For certain major retailers, AI technology miscounted inventory, mislabeled products, and/or even struggled to process orders. What these retailers are finding is something that has always been true for any enterprise looking to scale AI. Without the right network and data foundation, large-scale AI implementation is impossible.
Clearing the Retail Data Bottleneck
When AI tools fail, the catalyst is rarely in the AI technology itself. Although AI technology continues to evolve — from generative AI to AI Agents and even physical AI — there remains a pretty universal truth: AI is only as good as the data it’s trained on. When modern retail operations struggle with AI, it’s often due to legacy platforms and siloed databases. This translates to very fragmented data with very poor data quality. Before any AI implementation, retail enterprises must spend a sufficient amount of time preparing data lakes and integrating data pipelines, so AI has a good basis from which to provide value.
Supporting AI with Network Infrastructure at the Edge
Once data is ready, the network must follow. Today’s retail applications are more distributed than they’ve ever been, no longer sitting in a single environment. They stretch over physical stores, edge systems, and cloud platforms. This means a lot of store-related data must travel across enterprise locations to maintain successful business operations. For example, hours of high-definition video must be able to travel throughout an entire retail ecosystem. AI is usually connected to this use case to help with loss prevention, traffic analysis, and shelf visibility.
If enterprises are forced to send all of that information back to a central cloud, it risks unnecessary latency and a potential rise in operational costs. To provide the real-time benefits associated with AI, retailers need to process high-volume data at the edge. As a result, data travel time decreases and network response times accelerate.
Retail enterprises should also leverage network solutions that can bring 5G and SD-WAN together to relieve potential network pressure as AI solutions scale. 5G allows businesses to better connect their stores, pop-up shops, warehouses, and vehicles. Moreover, it allows retailers to more easily support AI-powered mobile operations and real-time analytics, especially as these functions move outside the four walls of one store. When it comes to orchestrating multi-WAN environments, enterprise network administrators can use SD-WAN’s network traffic steering capabilities to manage network traffic based on business priority. In addition, network administrators can apply the same policy across sites. A cellular-centric SD-WAN solution can provide added value such as aggregating multiple wireless connections to improve both throughput and resiliency, bringing more wired attributes to wireless WAN networks. Together, these solutions provide flexibility and control that relieve network pressure while also creating a network foundation that limits disruptions to critical AI-driven retail workloads.
Properly Scaling AI in Retail
Although retail CEOs say they plan to spend more on AI in 2026, it’s still early days for large-scale AI use cases. This means retailers don’t have to rush in preparation for more AI implementation. It will be important to take a measured approach to AI readiness. First, create the right data foundation with clean data lakes and clear, connected pipelines. Next, don’t put every AI use case in the same network zone. Segmentation — or placing POS systems, security cameras, and employee tools in different network zones — lessens the chance for security breaches or performance issues.
Retail enterprises must also ensure each AI use case is under a consistent network template, centralized network policy, and pathway to scalable provisioning. This will allow easier and more responsible AI deployment as AI use cases expand. Link diversity will also play a role in ensuring AI tooling remains online. By leveraging network solutions that can intelligently connect to the best cellular link, or network solutions that can aggregate links to support more traffic, retail enterprises will remain prepared to support data-intensive AI workloads.
Finding True ROI and Leaving Hype Behind
Retailers still have the opportunity to learn from the cloud hype cycle of a few years ago. After the big cloud migration, many enterprises were forced to participate in cloud repatriation. They realized a measured, systematic approach to the cloud allows businesses to feed off its benefits while protecting proprietary data and other sensitive information close to home.
The enterprise has the chance to learn a similar lesson and take a measured approach to AI adoption. Whether it’s generative AI, agentic AI, or physical AI, the retail enterprises that win with AI will be the ones that have strategically leveraged a network that keeps AI tooling always-on, business always-running, and customers ultimately happy.