Physical AI in Supply Chains: How Real-Time Intelligence Maximizes Visibility and Reduces Loss

Summary

Wiliot creates digital identities for physical assets using battery-free, RF-powered Bluetooth sensors the size of a postage stamp that continuously measure location, temperature, humidity, and light exposure without manual scanning.
The Wiliot Physical AI platform processes sensor data at scale using Databricks, enabling real-time detection of mis-shipments at dock doors, automated shrink detection, and tracking of reusable assets such as pallets that would otherwise be lost in the supply chain.
Walmart is using Wiliot across every distribution center to transform how product flow is monitored, with organizations achieving up to 12x ROI and one use case identifying $260 million in potential loss from idle pallet dwell time.

Physical AI in Supply Chains: How Real-Time Intelligence Maximizes Visibility and Reduces Loss

Supply chain leaders lack real-time visibility into the location and condition of goods across complex, distributed networks — driving shrink, inefficiencies, and missed revenue.
Physical AI introduces a new paradigm by continuously capturing real-world data from physical assets, enabling a live, accurate view of operations. The Wiliot Physical AI platform, powered by Databricks, processes this data at scale to deliver real-time intelligence and automated decision-making.
Organizations are achieving up to 12x ROI by improving inventory accuracy, reducing loss, and unlocking end-to-end visibility. Leading retailers like Walmart are using this approach to transform how they monitor and optimize product flow.
Attend this video to learn how to operationalize Physical AI in your supply chain and achieve continuous, scan-free, real-time visibility — so you can respond instantly, reduce waste, and drive ROI from day one.
Talk By: Amir Khoshniyati, Vice President, Wiliot ;

Chapters

FAQs

What is Physical AI and how does Wiliot apply it to supply chains?

Physical AI continuously captures real-world data from physical assets—such as pallets, crates, and inventory—using battery-free Bluetooth sensors and converts it into a live, accurate view of supply chain operations. Wiliot's platform, powered by Databricks, processes this sensor data at scale to enable automated decision-making without manual scanning.

How do Wiliot's sensors work without a battery?

Wiliot's sensors are postage-stamp-sized devices powered by ambient RF energy, eliminating the need for batteries or manual activation. They continuously measure location, temperature, humidity, and light exposure, creating a persistent digital identity for every tagged physical asset throughout its supply chain journey.

How is Walmart using Wiliot's Physical AI platform?

Walmart has deployed Wiliot across every distribution center to achieve real-time visibility into product flow. Use cases include detecting mis-shipments at the dock door in real time and tracking reusable assets such as pallets and plastic crates that previously disappeared in the supply chain.

What is the economic impact of supply chain brownouts that Wiliot addresses?

This video cites 1 billion hours of manual supply chain reconciliation per year industry-wide and $260 million in potential loss identified from idle pallet dwell time in a single use case. Organizations using Wiliot's Physical AI platform have achieved up to 12x ROI by improving inventory accuracy and reducing loss.

Full transcript

[00:08] I'm the last session on a technically one of the last days. So, I'm going to try to keep this engaging the best I can with the headphones and especially in this live environment and I'll also test you guys a few times through this session to make sure you're paying attention as we run through everything.
[00:22] So, as Matt mentioned, I'm Air Koshi. I'm vice president at Willot. Um, I have a very exciting and hopefully very quick presentation for you all really around physical AI and supply chains. Just so I
[00:34] get a lay of the land before we start, raise of hands. How many are end customers here? Okay, a few of you. How many are on the software side or you're a partner in the data bricks ecosystem?
[00:47] Okay. And how many of you didn't vote? >> Okay, couple of you. Okay. So today's presentation I'm going to go over the concept of what physical AI is. I'm going to go into what will as a business
[00:59] and I'm going to talk around how we've partnered with data bricks and I actually have a example of the real world what we're doing with a customer that was actually spotlighted in an earlier session today and their name is
[01:11] Walmart which many of you and all of you should should know who they are. Before I start and really jump into it, um I really want to start with the problem that we see from Willia's perspective in the supply chain and that's really
[01:24] around supply chains overall have a lot of brownouts. What we mean by brownouts mean it really means that they're flying blinds. There's some data points up here. I'm not going to spend too much time going
[01:36] granular on them, but they have to do with items that are in shrink. They're lost. So visibility is completely out in the supply chain. They have to do with reusable assets. So many times when we're tracking assets as an
[01:48] organization, we actually have our own assets that those assets go into. We use them in reusable plastic crates, sometimes pallets that are reused. Uh we've seen that many times these are lost throughout the process and once
[02:00] they're lost, they're gone in the abyss. And this is a major hurdle for uh supply chains overall. Revenue loss is is inevitable and then there's a lot around manual scanning. So
[02:13] we have a we have a data point here 1 billion hours. You can multiply this by whatever those working hours are for each employee and and it really translates quite a bit. So high level
[02:25] there's a lot of different facets to supply chains and the areas where there's brownouts but I wanted to call this out to lay the land around what we do as a business and how we achieve it.
[02:37] How many of you are familiar with Willot as a starting point? Okay, couple couple hands in the audience. We are a physical AI platform. What that means is that we take the real
[02:50] world, the physical world, we create digital identities and through our platform, we ingest that data and we make sense of it and we give you actionable insights. How do we do it? We
[03:02] do it with a pixel size of a postage stamp that I'm holding up. This is a Bluetooth tag without a battery. We harvest energy off of ambient waves. We have a capacitor that acts like a
[03:15] sponge. It energizes this tag and then it sends a signal via 2.4 GHz of where the asset is. It gives you insights into the sensing behind it. So it tells you the condition
[03:27] it's in, temperature, humidity, even gives you exposure to light. So if it's sitting in a package, that package has been opened any time throughout the supply chain. We can give you visibility that it's been tampered with. What's
[03:40] interesting here around the ecosystem that we've created, everybody wins within that ecosystem. That's why I started with the question, who's in the data bricks ecosystem? Are you a software partner? We have a play for everybody within that. You could be a
[03:52] tag manufacturer, produce these tags. So, a lot of the RFID manufacturers, they produce these on the same equipment that they're right now running as a standard process. On the network side, I
[04:05] uh I was made aware there's some network folks in the audience today. It's T-Mobile, a few others. We uh we work within the gateway ecosystem. So, transmitting this data when the tags are energized and they're repurposed to a
[04:18] network device as a middleman. And then on the platform side, we work as an ingestion engine, but we're not the endall. We we are the source of truth, but we can pass that on to any ERP, WMS,
[04:30] TMS in the process. Now a concept that I really wanted to dive into here. How many are aware of the concept of internet of things?
[04:42] I think we should have most of the audience have their hands up with internet of things. I was never a believer in internet of things and I'll tell you why. When this concept was created for me it was more around internet of
[04:55] devices. Many of you are aware we started with on premise hardware. On premise went to the cloud. We had a lot of keynotes, a very notable one from Satia from Microsoft where it
[05:07] was uh mobile first, cloud first messaging. It was a it was an era really around devices. The the devices were going to the cloud. Many of us today now have multiple
[05:19] phones, we have tablets, we have laptops. It was a foundation for these devices to come together. The concept of internet of things really didn't give rise until co where everything went
[05:31] contactless. QR codes were prominent. Credit cards started to transition into contactless with NFC, a subset of RFID. All of this gave rise to the era that
[05:44] we're sitting in which is called physical AI. And this this slide highlights it very well that we are giving giving evolution in the technology stack from barcodes and QR
[05:56] codes which require line of sight. They're reasonably affordable when you look at scale but security and capabilities are not at an all-time high as you move up the ladder like you have
[06:09] with some set of RFID but where you get with also Willots's pixels in RFID. Many of the things that we've seen is really around traceability. Of course, there's aspects of sensing when you get
[06:21] into the active state, but you have to change the infrastructure and that infrastructure is typically pretty expensive. So, when you're looking at a supply chain solution, you can start with dock doors, you can start at entry
[06:33] and exit points, but scalability is a problem, especially when we look at brownouts and some of those problems I highlighted on the first slide. Now, enter Williet into this new era of what we call physical AI. We hit the
[06:47] capabilities that you see with the barcode. So, this pixel that I'm holding up, many of you can come by afterwards. We can show you some of the samples. The pixel has an antenna like an RFID tag. So, we cover the same capabilities you
[06:59] see here. But on the conversion side, we also convert via barcode. So, we can work into existing infrastructure. So as we get into some of the use cases, especially the one with with data bricks and Walmart, you'll see that a lot of
[07:12] the capabilities that were in the early stage era as we were getting into internet of devices into internet of things and giving rise to physical AI, they started within this foundation that we've carried over into the Williet
[07:24] Pixel, but we've went one step further with those sensing capabilities with an all-in-one. Right now price point wise we're roughly about 10 cents for a pixel and future gens are going really aggressively towards five cents. So
[07:37] we're covering this whole technology ladder but with additional capabilities. Okay, how does it all work with data bricks? Why are we here and why are we co-presenting? So we've really brought
[07:51] together a four-step process within the pixels themselves and the way that they work within the data bricks arena. Step one is ingesting this data. So the pixels will stay on a product. They will
[08:05] energize. They will tell you association where that product is, what it's doing, the condition it's in. We'll ingest that data to the Willia cloud. When it's within the data, that's where the engine
[08:17] starts to run. We start to process with our own algorithms, start to do our own levels of ingestion. Data bricks comes in as a additional layer there. And then very proud of the
[08:29] partnership because the output here is where the steps three and four come in. We have a delivery stage that's really built around execution. It's built around the module that you've seen with
[08:41] Genie, how it's embedded with what we call Ether, the the Williet engine. And then ultimately, it gives the end destination around our end customers on how we ingest that data and what to do
[08:54] with it. So that can sit on the Williet platform with data bricks assistance or we can be a bridge that tunnels that data to any of your ERP WMS's or wherever your custom engines are on the
[09:06] software side. So it's a four-step process but the partnership has came together really really well with uh with data bricks.
[09:18] >> One thing that we've done here you've seen a QR code I have one last slide after this. We've created a ebook. It's almost like a novel with the level of detail behind it, what goes into the pixels, what this engine looks like, how does it work within a customer like
[09:31] Walmart. Um, I encourage you all to scan that and I'll show it again on the next slide. But it a lot of the ROI that we've proven with Walmart and some of our additional customers are built into that ebook. So with Walmart, where we
[09:45] are, we've made this public last October. We are right now uh in every one of their distribution centers at the pallet level and by the end of this year we're on track to have every US location over
[09:58] 4600 stores outfitted with Willian's technology with data bricks assistance. And what we're doing at the pallet level at each one of these DCS we're associating the pallet to a pixel. We
[10:12] recognize that pallet is going out the door. If it goes to the wrong truck before that truck leaves the dock door, we can flag it so they remove it and reduce any kind of cost that for misout
[10:24] routing that uh that shipment. And then we track the journey of that pallet to the store. Everything from the location to the condition that it's in with the temperature. We found during this process that for every 30 minutes that a
[10:38] pallet is sitting idle, its dwelling at a at a back of store or at a dock door, it equates to $260 million in loss from Walmart. So, a very simple 10-cent
[10:51] solution going on a pallet and recognizing that it went through the right door. It left the facility, it's auto received at the location within Walmart, and it's sending event triggers that that pallet needs to be unpacked
[11:04] and put into a refrigerator or freezer is a massive amount of savings for a very simple lift technology. So, we're on track right now through that. Uh we have some really good call
[11:16] outs and you'll also see that in the ebook with uh with Greg and and the progress that we've made with that team. Um there's three real main uh benefits that we've found from this uh solution. So the temperature side on the
[11:29] perishable, it's the mis shipments, mis routing for the trucks, and last but not least, it's really the man-hour and the automation around it. We don't require Walmart employees backup store anymore to scan
[11:42] items. We've automated that process and the next stage of where we're going with this partnership is to evolve from the pallet side and move to the case side and for us it it evolves. It goes to 4x
[11:55] when what we're doing on the pallet side. So sky's is the limit for this partnership with data bricks and I'm really looking forward hopefully to collaborating with many of you where you sit in the ecosystem to continue to build on this partnership and and many
[12:08] others. And what we've seen from this so far with that massive number that I outlined on loss 260 million plus is that ROI is less than one year. So we see this as a common thread right now in
[12:22] all the different engagements that we're working with.
[12:34] got a lot of inquiries on uh LinkedIn prior to starting. You can scan this, reach me and my team and I will follow up with you. And also really encourage you to download that ebook. A lot of the presentation that I I went through today
[12:46] very quickly, the details are really within that ebook. And then we also have a lot of examples on the technology and the ROI that it's creating as well. So there's a lot of value that you can all benefit from.

Learn more about the Databricks Data and AI platform.

The information provided herein is for general informational purposes only and may not reflect the most current product capabilities or configurations.