MSI Surfaces Boosts Inventory Accuracy with AI Drones | Manufacturing Engineering

These days, drones are used for everything from aerial photography and spectator shows to product delivery. When the largest distributor of countertops, flooring and wall tile in North America felt it needed a better handle on its vast amount of wares, the company found another use for them: inventory tracking.

MS International Inc. (MSI Surfaces) tapped Corvus Robotics Inc. to supply artificial intelligence (AI)-powered drones in a move that ultimately led to better inventory accuracy, higher customer satisfaction, and lower operational costs.

Corvus Robotics was launched in 2017 in Boston, before moving to Mountain View, Calif., in 2022. It views itself as a mission-driven company that uses robotics and AI to improve supply chain efficiency and economic productivity throughout the globe. The company builds its own products at its home base in California. It sells AI-driven robotics, software and data-driven insights to companies and governments to provide better inventory visibility.

Improved inventory visibility is exactly what MSI was after when it enlisted Corvus Robotics’ help. MSI was founded in 1975. The company now has about $3 billion in annual revenue and more than 400,000 global employees. MSI has grown to become a leading supplier of a range of hardscaping products in North America. Based in Orange, Calif., the company has more than 45 showrooms and distribution centers across the U.S. and Canada. Its product portfolio spans everything from quartz, luxury vinyl tile, turf, porcelain and natural stone, imported from some 37 countries across six continents.

Keeping Track

MSI also has four major warehouses and distribution centers, which measure more than 200,000 sq ft (18,581 sq m). The facilities are filled with tens of thousands of products that may need to be shipped to customers at a moment’s notice.

Before MSI deployed Corvus Robotics drones, employees would have to scour the racks and peek into various locations to physically take stock and compare it to what was in the company’s inventory management system. That was the primary process for finding missing material. The setup also carried the risk that the company would come up empty for certain material and not be able to fill an order.

“We used to have a team of two to four employees who would drive through the racks and look at every location, then compare what’s physically in the location to what we’re expected to have in the system,” explains Matthew Zucker, team lead for operations at MSI Surfaces. “That process could take weeks, potentially longer if those employees were pulled at any point to fulfill orders. So to cut down on the time it’d take us to do a full inventory audit and to reduce the number of missing materials, we knew we had to improve our inventory management processes, which ultimately lowered the amount of delayed orders for our customers.”

With MSI’s previous system, inventory accuracy was estimated to be about 80%. A physical inventory (PI) audit was done annually, and the cycle counting speed was about 3-4 weeks. Now, inventory accuracy is touted at 99%, there is no PI labor and PI frequency is weekly, while cycle counting speed is 20 times faster.

With the drones, the company can get more accurate systematic and physical information. That’s reduced discrepancies in the data. Moreover, with the improved accuracy along with the faster cycle counting, the company estimates it saved $110,000 in labor reallocation and $1.25 million in recovery of lost inventory.

How It Works

At first, MSI tried to develop its own drone. While this approach didn’t provide the value or effectiveness the company sought, it did pave the way for an external partner. “From that standpoint, we were already comfortable with the idea of using drones,” says Reggy Soenarso, MSI’s senior vice president of operations. “And we wanted to vet it further with a company (Corvus Robotics) that actually developed and manufactured their own drones for this exact scenario.”

However, MSI still didn’t go all in initially. It started with a proof-of-concept unit at its main distribution facility in Orange. “Once we had the data to prove that the drone was beneficial for us, we then expanded to several other locations in New Jersey, Georgia and Texas,” Soenarso says. The company ended up choosing Corvus because it provided a plug-and-play system without requiring any major infrastructure changes, Soenarso notes.

So how does it work? One might assume that it involves significant setup time, but Corvus says that’s not the case. The company was conducting fully autonomous flights in MSI’s warehouse within the first day. The total setup includes a landing pad and software.

“Each drone is equipped with 14 cameras designed with custom optics to see and understand the environment using an AI, large-world model,” explains Jackie Wu, Co-founder and CEO at Corvus Robotics.. “All compute for the AI, large-world model runs onboard the drone in real-time, without needing a ‘trunk full of GPUs’ for autonomy like a self-driving car would.”

Depending on the size of the facility, how many pallet positions need to be scanned, and how often inventory needs to be audited, anywhere from one to six drones can be deployed. The drones can fly at any time, during operating or non-operating shifts, through either very narrow—50-inch (1,270-mm) width—or wide aisles. A typical “mission” lasts about 20 minutes with scanned positions ranging from 200 to 500 per mission, according to Corvus. The variance depends on distance traveled, barcode placement and the number of barcodes captured per location (there could be multiple pallets/SKUS stacked in one location).

No stickers, beacons, reflectors, or WiFi for localizing inside a warehouse are needed, he adds. Instead, Corvus One drones autonomously navigate off the company’s large-world AI model running onboard, which uses a learning-based, neural network approach to autonomy to understand what inventory is where, what to look for, and how to interact with the environment. It’s inventory management on autopilot.

The lightweight drones fly at walking speed without disrupting workflow or blocking aisles and can ascend to avoid collisions with people, forklifts or robots as needed, according to Corvus. Moreover, the company adds, advanced barcode scanning can read any barcode symbology in any orientation placed anywhere on the front of cartons or pallets.

After the drone flies, all the data is downloaded to a dashboard that can be viewed by MSI personnel. They can pull up any flight and see the scan of any location, what the drone scanned and what was expected to be in place based on the company’s warehouse management system.

The system automatically identifies and provides notification of any discrepancies. If a discrepancy is found, Corvus categorizes the type and provides a suggested resolution, which an MSI associate can access by logging into the system and viewing a video of the location in which it occurred. The associate then updates the warehouse management system to record the variance for that location, and decides what physical steps, if any, should be taken to address the issue—such as fixing the barcode if it is blocked from view or damaged, or moving or swapping pallets of inventory to the correct location.

Droning On

So, what advice would MSI have for other companies considering drones for inventory management? “I would say, keep an open mind and be willing to accept that the root cause of inventory-related issues may not be as simple as they originally seemed when you only have a few anecdotal data points,” Zucker says.

Data from the drones has also unlocked insights into other upstream and downstream processes and control gaps in the warehouses. While the drones “primary” value is to help MSI identify and fix immediate problems, ”but the real power comes now that we have the data and time to ask the question ‘why is it wrong today and how do we prevent it from being wrong again tomorrow?’” Zucker says. That requires full buy-in from both upstream and downstream operations, he adds, noting that inventory is not a standalone function, it’s a full-warehouse one.

Zucker also stresses that while there is a perception that robots, such as drones, are more reliable because they never call in sick, for example, the reality is that no emerging technology is perfect and there will be some bumps in the road.

“Our experience has been that Corvus is a great partner to work through those bumps with together to achieve a shared, positive outcome,” Zucker says. “Automation does not mean perfection, and perfection is not required to make an impact.”

Lights Out, Upgrades Added

The commercial version of Corvus One debuted in 2021. Today, more than 40 warehouses across the U.S. use the technology to track inventory, according to Corvus Robotics.

In October 2024, the company launched an upgraded version of the system that allows drones to operate in a lights-out distribution center without any added infrastructure. The upgrade was supported by an $18 million Series A and seed funding led by S2G Ventures and Spero Ventures.

Corvus is now in the process of replacing older versions of the drones with the units at no cost to customers. MSI and several other companies are already using the upgraded systems.

For more information about MSI Surfaces, visit www.msisurfaces.com or call 800-490-0060. For more information about Corvus Robotics, visit www.corvus-robotics.com or call 833-558-0812.