The pallet industry is no stranger to hard work and tight margins.
What is changing now is visibility. A new generation of AI-powered technology is making the shop floor measurable in real time. Zira, an AI vision platform that has been growing fast in pallet recycling and manufacturing facilities across North America Pallet Enterprise editor Chaille Brindley recently sat down with Zira CEO Elhay Farkash to discuss the company’s technology, real-world results and what it takes for a pallet company to get the most out of AI.
Pallet Enterprise: What are the top use cases you’re seeing Zira deployed for in the pallet industry?
Elhay Farkash: Everything we do comes down to one thing. Turning the shop floor into something you can see and act on in real time. We see three core use cases driving immediate value. The first is operational shop floor visibility. This makes operations live and real time so operators, floor managers and owners can see exactly what is happening at any given moment. This allows teams to track performance against targets, and fixing problems immediately instead of discovering them after the shift or the next day.
The second use case is pallet grading and sorting for recycling operations. We use AI vision to replace manual inspection, and we do it extremely fast. Depending on the configuration, inspection takes place in under 45 milliseconds, sometimes as little as 15 milliseconds per classification. At 15 milliseconds, the line never has to stop. It runs at full speed, and the AI makes accurate sort decisions the entire time, routing pallets according to the customer’s specific grading requirements.
The third use case is quality control for new pallet manufacturers. This is what we call the No Bad Palletssystem. Every pallet coming off the line is captured, measured, and validated for quality, protruding nails, cracks, missing boards, misalignments, board spacing and other parameters. Quality is built directly into the production process, so there is certainty that every pallet leaving the facility meets high standards. This also creates full traceability and gives both operators and customers confidence in every shipment. We are seeing the industry move toward a no bad pallet standard. Every pallet documented, validated, and traceable.
Pallet Enterprise: There are a lot of misconceptions about AI in manufacturing. What do you hear most often, and how do you respond?
Farkash: Most systems tell you what happened. We tell you what is happening right now. The biggest misconception is that AI is disruptive. This mindset suggests that AI is complicated, expensive, requires a team of experts and means tearing apart your existing operation. That is simply not how we think about what we do at Zira. Our system is designed to be as simple to use as a cell phone. It is extremely capable and built to deliver value immediately, without requiring major changes to your infrastructure. The second misconception is the fear that AI is here to replace people. We are not trying to replace managers or shop floor workers. We want to give them better tools so they can reach higher levels of throughput and yield than was previously possible.
There is also a flip-side misconception: that you can simply drop AI into your operation and keep doing everything exactly the same way. You do need to adapt your processes to take full advantage of what the system can do. I compare it to learning to ride a bicycle. You can ride with training wheels, but if you learn to ride with two wheels and proper gears, you become far more efficient. The technology can be dropped in without disruption, but getting the most out of it requires a process adjustment. And that adjustment is well worth making.
Pallet Enterprise: Walk us through what the onboarding process looks like for a new Zira customer.
Farkash: Onboarding is a structured process built on the millions of data points and best practices we have already accumulated. We don’t just install technology. We bring a proven way to operate with it. The first step is defining the customer’s objectives. What is the primary thing you need to achieve? Is it throughput? Yield? Grading accuracy? That goal shapes everything else.
From a timing perspective, basic functionality like counting starts within one hour of installing a camera. If a customer is deploying our full inspection booth, which comes with pre-built grading models, they have grading and sorting capability from day one. A full deployment across a facility typically takes between one hour and two weeks depending on the number of cameras and the complexity of the environment.
After initial deployment comes the continuous improvement phase, which is really the ongoing heart of the relationship. That might mean experimenting with pallet flow direction, adjusting camera positioning relative to equipment, or refining the sort categories the system routes to. We now allow customers to start with a single camera rather than requiring a minimum of four, so they can experience the value firsthand, understand the system’s capabilities, and expand from there.
Pallet Enterprise: Tell me about the inspection booth technology itself. How does it work?
Farkash: We offer three levels of our inspection system. The full system uses seven cameras in a multi-camera configuration. This provides a complete view of any pallet, whether block or stringer, up to 60 inches in size. It makes classification decisions in 15 to 45 milliseconds depending on pallet type, without ever stopping the line. The system can handle 21 different pallet classifications and identifies everything from board condition to sticker presence to structural integrity.
For operations that don’t handle block pallets, we offer a four-camera system that covers the full inspection. We also designed the system to work at both ends of the line. Positioned at the beginning, it grades and routes incoming pallets to various departments, repair, dismantling and direct fulfillment. Positioned at the end, it verifies quality post-repair and confirms things like sticker presence. Some customers install systems at both locations to capture visibility across the entire value chain.
Pallet Enterprise: How has Zira’s technology evolved over the past year?
Farkash: The pace of change has been rapid. One of the biggest advances has been on the human interface side. Our team has built visual tools that make it much easier for people on the shop floor to understand and act on what the system is seeing. The technology was always strong, but making it accessible and intuitive for operators who are not technology experts has been a major focus.
On the detection side, we started with counting and have progressively expanded what the system can identify. We now detect 21 classifications on pallets, along with lumber features. We can see boards, shiners, protruding nails, cracks, and distances down to the sub-millimeter level, all from a standard 2D camera image. There is no scanning technology required. This is possible because of the billions of pixels of data we have accumulated over two years of real-world deployments. The AI continuously trains and improves our models.
We also just released a mobile capability that lets customers photograph a stack of pallets with a tablet or phone and receive an immediate AI analysis of pallet sizes, colors, source, mix, and overall condition. This replicates what a trained human would assess when walking up to a load, but with the added ability to measure things that the human eye can only estimate from a distance.
To put the pace of this in perspective: when we talk about ‘legacy cameras’ internally at Zira, we mean cameras from 18 months ago. Our next-generation cameras came out six months after that. Compare that to a legacy ERP system, which might be considered legacy after 15 or 20 years. That is the speed at which this technology is evolving, and it is one of the reasons why our software-as-a-service model matters so much. Our customers’ models improve continuously rather than being static.
Pallet Enterprise: How does Zira handle data ownership, and what does the relationship between customer data and your AI models look like?
Farkash: In our model, the data belongs to the customer and the models belong to Zira. We use customer data to improve our models, and those improved models are then deployed across all our cameras. Every customer benefits from what we learn from the collective data. But we never sell customer data, and customers have full visibility into everything they’ve accumulated from day one. If they ever want to remove their data or start from scratch, that’s entirely their right.
There is an important distinction here compared to something like a traditional machine controller. A Siemens controller, for example, is a one-time purchase. You pay for it, it does what it does and that’s it. Zira is a living service. The models keep evolving with every pallet that passes through. That continuous learning is exactly why we can do things our competitors cannot, including grading at the speeds we achieve.
We also have some customers who build proprietary custom models on our platform for specialized use cases. Those models belong exclusively to those customers. We cannot replicate or sell them to anyone else, even if another customer asks for the same capability. Those customers pay for their own model development, and what they build is entirely theirs.
Pallet Enterprise: What kind of results are your customers actually seeing?
Farkash: In most cases, customers see a return on investment within the first one to three months, driven by higher output and better utilization of existing labor and equipment. For example, across our customer base, we consistently see 20 to 30% throughput increases on existing equipment. This is achieved on existing equipment with no new machinery required and with the same workforce. For most operations, that translates directly into higher output per shift and lower cost per pallet.
Georgetown Pallet is a great published example. Its previous peak was 700 pallets per shift and its typical production was around 500 pallets. Today, 900 pallets per shift is their standard, and they reach 1,000. That is a step change in productivity on the same machines with the same workforce.
Beyond throughput, quality is becoming an increasingly important driver for AI adoption. We see the demand for higher pallet quality coming from pallet users across food and beverage, high tech, and other sectors. This need to improve quality, we see it on both the new pallet side and the used pallet side. Quality is emerging as one of the clearest competitive differentiators for pallet producers. This selling point provides a compelling reason to invest in AI beyond the throughput and yield benefits alone.
Pallet Enterprise: What would you say to a pallet company operator who has been watching AI from the sidelines but hasn’t pulled the trigger yet?
Farkash: I understand the hesitation, but the barrier to entry is no longer there. You can install one camera, and within one hour, you are operating with real-time data you did not have before.
That is the starting point. Everything else builds from there.
The fundamental shift is not technology, it is mindset. The pallet industry has always been operationally intensive. What AI adds is visibility: the ability to see everything that is happening right now, not just what happened at the end of the day. That shift from ‘what happened today’ to ‘what is happening right now and what can I do about it’ is genuinely transformative. Once you can see everything in real time, you can act in real time. You can change the outcome before the shift is over.
What I would also say is this: what Zira and companies like us are doing in the pallet world right now is genuinely cutting-edge. I participate in roundtables with leaders across many industries, and the level of AI adoption the pallet industry has achieved is ahead of where a lot of other sectors are. A plant with older machinery that puts Zira on top of it can achieve the visibility, efficiency, and quality levels of a fully automated modern facility. This can be done for a fraction of the investment of putting in all new equipment. That is the power of what is available today. The question is no longer if this will happen. It is whether operators choose to lead with it or fall behind it. The ones moving now will define what this industry looks like over the next five years.
Editor’s Note: Zira’s platform is currently deployed in pallet recycling and manufacturing facilities across North America and expanding globally. The company offers scalable solutions starting from a single camera and expanding to multi-camera inspection booths capable of classifying any pallet type at full production line speed. For more information, visit www.zira.us, email sales@zira.us or call (650) 701-7026.
