Jul 25, 2021New technologies emerge all the time, often wowing us with visions of futuristic applications that are as real as concept cars at motor shows—they look great, but it will be years until the rubber hits the road, if it ever does. To be effective, a new technology needs to be real-world tested to provide proof of concept before it can be adopted, let alone displace existing technology solutions. Go too fast, and there is a risk that the potential benefits of the new technology will be lost in an expensive and failed implementation.
One well-known example comes from RFID, with Walmart's missteps nearly 20 years ago in an ill-fated early adoption of the then-new technology across its supply chain, complete with pushback from suppliers, technical problems and databases that couldn't handle the volume of data generated. This case demonstrated the folly of a "too much, too soon" approach, though since then, RFID has of course found much success across many industries all over the world.
The Rise of the Machines
Artificial intelligence (AI) and machine learning (ML) are very much in focus for many different industries, with boundless applications promising untold benefits to businesses and consumers alike. In one ambitious implementation a few years ago, Amazon Go showcased a combination of AI, ML, computer vision and sensor fusion technologies to create a so-called "just walk out" store with no check-out. A tech-heavy retail store—complete with myriad cameras, microphones, projectors, motion detectors and volume-displacement sensors, all connected to an inventory-management system—promised a frictionless shopping experience.
At the time, it was posited that such a solution would be more cost-effective than RFID, due to the high cost of tags, especially for low-margin consumer goods. Of course, the cost of RFID tags continues to fall, but Amazon Go did showcase the potential of various AI and ML technologies to track the movement and sale of products in an innovative, low-touch solution. With the potential for AI and ML solutions to do it all at the point of purchase, where does that leave RFID?
A Holistic View
Besides being very costly to establish and maintain, the Amazon Go retail store that utilized AI and ML provided very limited scope, considering the entirety of the supply chain that works to get products into the store in the first place. From production to distribution to warehouses and finally to stores both real and virtual, RFID provides end-to-end coverage for complete visibility and efficient inventory management with a combination of tags, RFID readers and customized software.
Beyond the store environment, AI and ML are making inroads into warehouses, including the use of drones and robots. While computer vision relies on adequate lighting to work efficiently, the use of RFID can help overcome insufficient lighting as part of an integrated solution. Combining RFID with AI and ML can also improve warehouse operations through the analysis of all kinds of data—from stock and people movements to predictive restocking to improving layouts for efficient product movement. AI and ML are not likely to displace RFID across the supply chain, but early implementations prove it can be a useful addition to RFID solutions.
Cutting-Edge Cuts of Beef
In an industry-first, Kilcoy Global Foods, located in southeast Queensland, has implemented AI with RFID to combat issues of incorrectly labeled beef products being shipped to markets all over the world. The new system includes photos from various angles being taken of RFID-tagged primal cuts of beef as they are made, with image reader stations placed strategically throughout the production facility monitoring their progress towards shipping. Over time, the system learns to visually identify individual cuts of meat.
"We have 10,000-plus primals going past these cameras every single day," says Kilcoy's Nigel Adler, "so it makes it easy for the system to start to learn what a primal looks like." In effect, RFID teaches the AI system which piece of meat is which, ensuring the correct labels are on the meat as it moves through the facility. The system also monitors cartons of meat as they are assembled, preventing them from being shipped with incorrect quantities and/or types of tagged items inside.
Brains and Brawn
AI and ML promise to bring a lot of smarts and sheer data-crunching capability to supply chain management. RFID is currently doing the heavy lifting of end-to-end tracking and undoubtedly improving the efficiency and accuracy of inventory movement and management right across the supply chain. Where AI and ML will provide the most impact, at least in the short term, will be to complement and enhance existing RFID technology solutions. This will take the form of improving accuracy and further reducing human error, such as in the Kilcoy example.
Beyond that, there is immense scope to implement AI and ML to effectively analyze and utilize the sheer amount of data produced by RFID solutions. Automated 24-7 number-crunching will provide untold insights and shed light on previously unseen patterns, ensuring better informed business decisions based on real-time data. When it comes to RFID, artificial intelligence and machine learning, it's not so much a question of either/or, but rather deciding on an appropriate mix of all three.
Pete Reinke is the CEO of Ramp RFID.
