Have you ever walked into a store and thought about how AI and compute are everywhere? Folks may not know this, but beyond chatbot LLMs, AI in retail is a multi-billion-dollar industry that impacts us just about every day. After hearing about it in many presentations over the last year, I started looking. Then I started asking.
I eventually asked the AMD folks about this since I am a big AMD EPYC 8004 fan, and the EPYC 8005 is coming. If you did not know this, a large amount of AI computing in retail locations is done on CPUs instead of GPUs. Over the years, we have seen demos of video analytics on CPUs because oftentimes, CPUs have enough performance that you do not need to add the expense and power of a GPU.
Someone on the call with AMD said they knew someone in Arizona. Quickly thereafter, I met Josh Ruggiero of Ruggiero’s Ace Hardware at one of their Arizona stores. It turns out, AMD made the introduction, but Josh used STH for some of his back-end IT research. We have to say this is sponsored by AMD, and also thank Ruggiero’s Ace Hardware for giving us special access to their store so we can show off some of the neat places AI is being used today.
Before we get started, I just want to point out that “retail” is a very large space. Some food providers, for example, are doing applications to check order accuracy before the food goes in the bag. Others, like the family business I visited in Lamay, Peru, which serves roadside cuy, would be on the other end of the spectrum, where the nice woman was the one who checked.
This is important because, while folks talk about all the cool things retail AI can do, realistically, the industry spans a wide spectrum. The Ruggiero’s Ace Hardware we are visiting is far from the woman, advanced in years, selling roadside cuy, but it is also not a 100% robot-driven space ship of a retail establishment. Like many retailers, it is somewhere in between, but I thought a hardware store was something many of our STH readers could relate to.
A Retail Location Has Compute Servicing Many Different Parties
For most of our readers, retail locations are simply places where we purchase goods and services. That is far too simplistic of a view, however. Instead of just customers, that retail location also has employees and vendors who staff and service the business. I certainly worked in a retail location when I was younger, and many of our readers have as well. Also, there are operators or owners of the locations.
Going beyond on-site computing, there are often online experiences, and all the computing that happens in a data center. For example, a large retail chain may do marketing analytics and craft personalized offers from a data center. Payment card processing also goes back to data centers, where there is an enormous frontier push to use AI to validate transactions.
The Parking Lot Is Already Computing
AI with retail location starts well before you even walk in the door. Even in the parking lot, cameras track how often certain spaces are used, how much of the lot is actually in use at any given time, and whether cars have been sitting there for extended periods.
Other retailers take this further with traffic analytics. They track how many people walk by the storefront versus how many actually enter the store. That conversion rate is an important metric to monitor, and you can do it pretty easily with AI these days. The video analytics compute infrastructure running these applications usually runs on in-store CPUs rather than sending all the video streams to the cloud.
Next, let us get into the store.
Every day I ask myself: when will Patrick finally lose it and post grilled rodents on STH? Today is that day!