Retailers can put Edge AI to work and benefit from its capabilities, writes Marc Del Vecchio, Retail/ Staff Solutions Manager, Supermicro
Edge AI technology is endlessly flexible, and economical solutions can be created to suit retailers of any size to help improve operations and enhance the customer experience in a range of different ways.
Industries as diverse as manufacturing, finance, and healthcare are being transformed by AI, and retail is no exception. Retailers similarly collect vast amounts of data, and AI excels at mining reams of data for actionable insights. Nearly every retailer relies on video, and AI is adept at tirelessly monitoring and analysing video content. Finally, interactions with users – customers – can be enriched by adopting chatbots and other digital assistants based on AI large language models (LLMs).
LLMs are the means by which AI programmes understand natural language – how your smart speaker understands you when you issue verbal commands.
In the beginning, it took the formidable processing capabilities resident in giant data centres to run AI workloads. If these data centres together with the world’s vast communications network constitute the Cloud, then the network Edge is defined by the devices that are endpoints of this network – smartphones, smart home gadgets, security cameras, factory robots, point-of-sale (POS) terminals and much more. In recent years, the number of requests for AI processing coming from the Edge has grown explosively.
In recent years, it has also become possible to run some AI workloads in computers located at the network Edge. Edge AI has been possible for some time, but is it advisable? As time and technology innovation have progressed, the answer is now unequivocally yes.
Edge AI is faster and more efficient than ever before. Data does not have to make the round trip to and from some distant computing centre, so network latency is avoided entirely and results are more immediate. Edge AI represents minimal usage of network resources, translating to cost savings all around. It is more energy efficient than cloud computing. Finally, data generated and used at the Edge stays at the edge, serving the interests of both data security and data privacy.
The benefits of AI deployed at the Edge apply whether the retailer is a chain with multiple outposts, has only a few locations, or operates only one store. Retail works the same regardless of the scale, with shelves to stock, customers to serve, and a variety of electronic systems for inventory management, POS, customer service, and security.
Putting Edge AI to work
Edge AI is the means for meeting a variety of retailers’ needs and improving a range of their business operations and practices.
For retailers, the emphasis on loss prevention is only getting more acute. Loss encompasses innocent mistakes made at checkout counters as well as criminal activity, but Capital One recently calculated that US stores lost $45 billion to theft alone in 2024 and projected that shoplifting could cost US retailers over $53 billion in 2027. Surveys of retailers in Europe over the years reveal that loss is a global and widespread problem.
Edge AI is invaluable for analysing data from security cameras for identifying suspicious behaviour on retail floors, at POS stations, and in stockrooms. This capability can be augmented with visual language models (VLMs), which are similar to LLMs, only optimised for video. VLMs allow AI systems to interpret actions captured on video. Edge AI can produce results in real-time or close to it, giving retailers the option to take immediate remedial action.
AI is also a valuable tool for inventory management. Inventory control can be more precise. This is especially the case when working with stock that can be tracked with RFID tags or Bluetooth Low Energy transponders.
It is also possible to implement systems that detect when shelves need to be restocked and then alert staff.
Edge AI can also be valuable for enhancing customer service. This could include adopting chatbots based on LLMs to provide product information or even make product recommendations. It is possible to use video to monitor shoppers’ browsing activity and then present relevant promotional offers on in-store kiosks or screens.
Inherent in the concept of applying AI is the concept of systems integration. This is because an AI solution is made up of many individual components: different types of software and hardware working together. The retail market has been computerised in a piecemeal fashion, and retailers often have separate systems for POS, signage, inventory management, purchasing, and so on.
The interrelated needs common to most retailers, then, are for edge computing, integrated computing, and the application of AI.
These needs can be met with Edge computing systems optimised for AI workloads.
To the Edge and back
There is demonstrable value in having a dedicated Edge AI system in any single physical outlet. Of course, larger organisations with many locations can aggregate data from all of their locations and use AI to get enterprise-wide insights.
Some retailers have connected their inventory and ordering systems, making it easier to manage the flow of products.
In multi-outlet operations, inventory and sales data from individual outlets can be cross-referenced, providing insight into how best to allocate stock. Data on single-store inventory and sales could be fed into an organisation’s supply chain management systems, integrating data from sourcing to sales. This can be accomplished from a centralised server with AI capabilities.
It is also possible to aggregate customer data and use AI to identify shopping trends that can be used to customise messaging from social media to in-store signage.
Right-sizing AI solutions
While most retailers have some basic needs in common with each other, the retail segment is far from homogeneous in terms of the nature of their operations, their relationships with customers, and their size, which can be quantified in many different ways, from annual sales to the number of outlets they have.
So, not only do retailers need Edge computing systems designed and built explicitly to run AI workloads, they need Edge solutions that can be right-sized for their applications and growth plans.
In the context of computing, right-sizing commonly refers to identifying the level of performance needed and balancing that against costs, while also considering return on investment (ROI). Naturally, it is always helpful to keep in mind the ability to upgrade – the possibility that business growth might necessitate adding computing resources at some point in the future.
Retailers looking to install an Edge AI system will want to consider the amount of processing power required and may wish to evaluate processor types, as some processors may be better suited than others depending on the different workloads.
To that end, retailers will want to work with AI systems suppliers who can provide options for the amount of compute power and memory, offer multiple connectivity configurations for the location, and provide support for future growth.
Retailers who plan to develop their own applications should also enquire about the availability of developer software development kits (SDKs). A useful SDK will already include modules that retailers can use so that they don’t have to build applications from scratch. Examples include applications that monitor foot traffic, track inventory, and analyse shopper behaviour. This makes AI development easier and reduces the time to market.
In the context of computing for retail specifically, right-sizing will also often refer to physical sizing. The retail category encompasses everything from big box stores with plenty of space that can be reconfigured as needed to tiny storefronts with barely any space to spare. With mid- to small-size locations, simply fitting a server can sometimes be a challenge.
There might be other factors involved as well. Temperature range is an example. A server placed in the back of a quick service restaurant (QSR) that will be subject to adverse environmental conditions, such as poor ventilation and system-destroying air particulates, will need to be configured differently from a server going into an expansive, temperature-controlled factory.
Successfully deploying and running AI applications at the Edge often requires specialised hardware. AI accelerators are the workhorses in this model, processing large amounts of data based on a pre-trained model. NVIDIA is the global market leader for these accelerators, with a broad range of AI-optimised CPU and GPU modules for workloads of any type and scale.
The NVIDIA Jetson Orin NX platform for embedded Edge and NVIDIA RTX PRO 6000 Blackwell Server Edition and NVIDIA H200 NVL GPUs for enterprise Edge, all have distinctive use cases based on the type and volume of data being processed, as well as deployment factors such as size and power consumption. Selecting the right hardware, therefore, includes identifying which AI accelerator is required and ensuring the server platform is compatible.
Developing AI applications can be a daunting and costly process. That is where a platform like NVIDIA Metropolis comes in. This end-to-end platform helps businesses fast-track the creation and deployment of intelligent video analytics by offering pre-trained models, SDKs, and optimised infrastructure.
For retailers, this means faster rollout of applications that monitor foot traffic, track inventory, and analyse shopper behaviour, without starting from scratch. By simplifying AI development and reducing the time to market, NVIDIA Metropolis helps companies lower costs while unlocking powerful insights from their video data, turning everyday surveillance into a strategic business asset.
It is worth noting that the retail and hospitality businesses can have significant overlap in terms of mission and even in terms of specific operations. Both aim to provide personalised service to customers. Many organisations that provide lodging also operate restaurants, retail shops, and mini-marts. All of the advantages of Edge AI that apply to retailers will also apply to the hospitality industry.
With over 25 years of experience in the IT industry, the last 10 of which was focused on AI, Marco has dedicated his career to advancing the fields of AI and computer vision, particularly within the retail and security sectors. His journey has been driven by a passion for leveraging cutting-edge technology to solve real-world problems and enhance operational efficiency.
There’s plenty of other editorial on our sister site, Electronic Specifier! Or you can always join in the conversation by visiting our LinkedIn page.
