S6E8 Zebra Technologies ZONE 2026 Conference Recap

[00:00:00] Teaser

[00:00:01] Ricardo Belmar: 80 apps, one handheld, zero patience. Retail's frontline is drowning, until now.

[00:00:08] Casey Golden: Ricardo just got back from Zebra's ZONE 2026 conference, and he brought receipts.

[00:00:14] Ricardo Belmar: Two exclusive interviews, Zebra's CTO Tom Bianculli and mobile computing chief James Poulton, on AI that runs entirely on the device. No cloud, no tokens, no waiting.

[00:00:25] Casey Golden: Plus why workers are drowning in 80 apps in the first place, the super app fixing it, and real-time translation, any language, live on a device.

[00:00:36] Ricardo Belmar: From the warehouse to the sales floor to your front door, this is what pragmatic AI for the frontline actually looks like.

[00:00:44] Casey Golden: A special bonus episode of the Retail Razor Show. Stick around and keep it sharp!

[00:00:49]

[00:01:01] Show Intro

[00:01:01] Ricardo Belmar: Welcome everyone to a special bonus episode of season six of the Retail Razor Show, the top-ranked indie business, management, and marketing podcast on Goodpods, and the original show in the number one indie podcast network for retail.

[00:01:14] I'm Ricardo Belmar

[00:01:15] Casey Golden: And I'm Casey Golden.

[00:01:16] And I have to say, Ricardo, you are practically glowing through the microphone right now. clearly been someplace fun!

[00:01:23] Ricardo Belmar: I have. So early last week, I had the privilege of being invited as one of the official media guests at Zebra Technologies' ZONE 2026 customer conference at the Gaylord Opryland in Nashville. It was two days of executive interviews, keynotes, demos, customer conversations, and yes, I have plenty of receipts from this one.

[00:01:44] So I came back with two really interesting video interviews. We're gonna share those with everybody today in this episode, and a notebook full of highlights from the sessions and other conversations I had on site with so many people.

[00:01:55] Casey Golden: I'm very jealous, but my, my current life [00:02:00] revolves around a two-and-a-half-month-old puppy, so I don't get to go anywhere.

[00:02:04] I can't even go on a walk yet. I've got questions!

[00:02:10] Ricardo Belmar: Fire away!

[00:02:12] Casey Golden: Big one. First, what did Zebra announce, and what's the headline coming out of Zone?

[00:02:19] Ricardo Belmar: Well, the headline, for me is that Zebra is no longer just telling a devices story. And I think we all know Zebra Technologies from all kinds of handheld devices and, and retail stores everywhere. But they're now telling, what I would call a frontline platform story. So there are three big software announcements, anchored at the event.

[00:02:38] So the first one is called Zebra Nucleus. It's a single pane of glass for configuring, managing, deploying every device in the portfolio. Mobile computers, scanners, printers, RFID readers, all these used to live in four different management tools. Now they're all unified under

[00:02:56] Casey Golden: A single pane of glass?

[00:02:57] Ricardo Belmar: Yeah.

[00:02:58] Casey Golden: What am I looking at?

[00:02:59] Ricardo Belmar: [00:03:00] One interface, right?

[00:03:00] So instead of having to deal with four systems, you just got one, no matter what kind of device it is. So I thought that was a pretty good, pretty good opener, to kick things off. And second is a product called Workcloud IO. The IO is for integration and orchestration. It connects to over 250 enterprise applications and intelligently routes work to the right person.

[00:03:19] So this is more of like a communication and messaging system. It's like a, a master layer on top of all of those other enterprise apps you have going out to the store being routed through one, again, one single interface for the frontline worker. And then the third one is called Workcloud BI, and this one's all about delivering role-based dashboards.

[00:03:37] So store associate versus a store manager versus a regional district manager. They all get a different dashboard, but it's tailored to what actually they're looking for.

[00:03:46] Casey Golden: So management, orchestration, and intelligence. I would kind of like to go redo my retail career with dashboards.

[00:03:55] Ricardo Belmar: Yeah, yeah.

[00:03:56] Casey Golden: That could kind of be fun.

[00:03:57] Ricardo Belmar: Yeah, right, right. Exactly[00:04:00]

[00:04:00] Casey Golden: So what does any of this mean for everyone on the sales floor? Because we, we're advocates for the frontline workers on this topic on the show. This is the topic as of late. What does it mean?

[00:04:12] Ricardo Belmar: Yeah. Yeah, so that, that's where it gets really interesting, I think. Zebra shared some numbers that h- honestly, you're, you're gonna be floored by these, and then probably, you're gonna say, "Well, yeah, of course." So they had data that shows that retailers are loading all these frontline workers' handheld, handheld devices with over 70 to 80 different apps.

[00:04:32] yeah, so like seven zero, eight zero, 70 to 80 different apps. But when-- And then if you think about, so they, when they analyze the usage, of course, only about a dozen of them actually get used regularly. You know, it's a classic case of the 80/20 rule playing out.

[00:04:47] Casey Golden: I mean

[00:04:49] Ricardo Belmar: yeah. So, so they've built what they're referring to as a super app.

[00:04:53] They a- they call it Sync. Basically covering, you know, 99% of whatever that frontline worker's gonna need across all these 70 to [00:05:00] 80 apps. It, it takes care of communication tasks, shifts, clock in, clock out, e- even micro-learning applications. So all those things that used to be individual apps are getting routed through this super app they've introduced, and then that's the only one of your, your frontline store worker has to interface with

[00:05:14] Casey Golden: That's great. I mean, that kind of goes back to the whole narrative that, in retail we're, infamous for having, 60 systems and we use about 10, 15, maybe 20% of each one of them. Having someone have to log in to more than five a day,

[00:05:32] Ricardo Belmar: Hmm. Yeah

[00:05:34] Casey Golden: that's so much friction

[00:05:36] Ricardo Belmar: Oh, yeah. Yeah. It's a huge human cost, right? From that. So one of their executives, Suresh Menon, he leads the software business, he shared a McKinsey stat related to that, in his session, said that, frontline workers spend 30% of their time just looking for the information they don't have, is a lot of time.

[00:05:56] Uh, and then to make it even, even worse, half of all the errors that [00:06:00] happen are because those employees just didn't understand what the standard operating procedure was. So what does that tell you? I mean, it tells me, you know, it's not a hardware problem, right? This isn't about having the right device.

[00:06:10] It's not about giving mobile devices to people. That, that's only a part of the equation. This is really about letting people actually do their job!

[00:06:17] Casey Golden: Oh my God. Like, that stat for errors in SOPs is wild. I mean, I can think of how many hours I've wasted looking for information that doesn't exist.

[00:06:28] Um, I mean, granted, probably part of that isn't, was in Microsoft Outlook.

[00:06:33] Um

[00:06:34] Ricardo Belmar: Yes. 'Cause email is where information goes to die. I always say that.

[00:06:37] Casey Golden: Right? Um, but that is wild, 30%. So what was the b- anything surprise you? Did... Was there something said that you didn't expect to hear? Anybody drop a big bomb on stage?

[00:06:53] Ricardo Belmar: there were a couple things. So the f- the first one is all about language. So here's another super interesting stat. So almost [00:07:00] 68% of frontline workers in the US speak a language other than English at home. In fact, one executive o- on stage, they mentioned they had a single distribution center where 18 different languages were being spoken. 18.

[00:07:13] Casey Golden: That's right

[00:07:13] Ricardo Belmar: Yeah. So y- and, and nearly half of frontline mistakes end up being traced back to a language miscommunication. So Zebra's answer for this, this is one of, I think, the cooler things they introduced, was real-time translation via AI into pretty much every layer, , even live voice. So, you know, you speak English, I hear Spanish, vice versa.

[00:07:31] It's all happens right on that mobile device. , Actually, I had lunch with, Erin Vigil from Burlington, , one of those days, and she mentioned, keeping in mind that, Burlington's one of those rare retailers where there's no e-commerce. It's all, every, all sales are in store.

[00:07:45] So o- of course, that means, you know, store associates are it, right? I mean, that that's where the, the rubber meets the road. That's where they, they service the customer because there's no e-com taking care of that. But she mentioned, yeah, they have-- they, they see this every day in their stores. And they [00:08:00] said that translation capability was one of the biggest wins for them across the store associate base, and that, that was like a no-brainer for them to turn on and a big reason for adoption.

[00:08:10] Casey Golden: I mean, I couldn't imagine anybody asking me to do my job in French. I mean, I went to an American university in Paris for a reason. 25 years of like lessons wasn't even enough. So that's really cool that even like 18 languages in like one store, I mean that's that's pretty rad. To be able to support that,

[00:08:31] I really- I'm really digging it. That's making somebody not just a shift better, but that just improves quality of life. It feels so much more accepting, I'm sure as well at work, the culture.

[00:08:44] Ricardo Belmar: Right. It's such a practical application for AI, I think.

[00:08:46] Casey Golden: Yeah.

[00:08:47] it's not a flashy AI silverware.

[00:08:51] Ricardo Belmar: Yeah.

[00:08:53] Casey Golden: What, what about surprise number two?

[00:08:55] Ricardo Belmar: So that would have to be the word tokenless. If I had a dollar for every time someone [00:09:00] mentioned tokens in the context of AI at this event, whether it was running out of tokens, exhausting your token budget, wanting to be tokenless, I mean, let's just say I'd have enough dollars to buy a year-long supply of cloud tokens.

[00:09:11] So Zebra's entire, strategy here with AI is being built on running things locally on that device. The model's hosted on the device, the inference happens on the device. They're not sending it back to the cloud. No tokens are being consumed in all, in all these applications. So they had two-- They talked about two new devices in their mobile computer range.

[00:09:31] I think it's the TC501 and 701 that are the newest ones. They have built-in neural processing units. If you think about it, these are starting to look like, like the laptops that you and I are using, in that handheld device, for the, for the frontline worker. And the timing, of course, was just classic because that same week, what were all the news headlines about?

[00:09:48] You had, Uber and e-even Microsoft, reporting that they had to back off on their internal, AI spending because the token costs were blowing up everybody's budget. So Zebra basically said, "Well, what if your [00:10:00] AI bill was zero?"

[00:10:01] Casey Golden: Well, I have some, some people's AI bill is zero 'cause they're not using it.

[00:10:07] Ricardo Belmar: it. That might be a

[00:10:10] Casey Golden: But okay, okay. I know. Okay, so tokenless is going to be the buzzword bingo card probably for the rest of, what? Probably for the next, like, six months at least. , But that, it actually makes sense.

[00:10:22] Um, I mean, literally you had people saying like, "Please stop saying thank you." Um, so what, how does that impact, like, the speed versus running in the cloud?

[00:10:34] Ricardo Belmar: Yeah, that's an interesting part, and I think everybody m- maybe misses, this aspect of it. So, think about frontline use cases. You know, you need answers in tens of milliseconds, right? At, at most. I mean, you're the associate helping the customer. And I actually heard them talking about this, and I thought back in my head to an example, like, over a decade ago, in, in a past job where I was talking to someone about a reason for, why you needed lots of bandwidth and network capacity at a store because you had [00:11:00] someone doing clienteling on a tablet.

[00:11:02] And I would tell everybody, "Well, role play," right? "You're the associate, I'm the customer. Count three seconds while you're standing there waiting for an answer." I mean, it feels like an eternity. And what customer wants to stand there and wait when nothing is happening? And so now it's like the example is think about how long you wait for a response from ChatGPT after you hit send.

[00:11:20] Casey Golden: I mean, ChatGPT does give me an answer faster than Claude, but the answer is not... The Claude answer is kind of worth the wait.

[00:11:29] Ricardo Belmar: Yeah.

[00:11:29] Casey Golden: if I was standing in person with someone, I would not have that

[00:11:32] Ricardo Belmar: But you wouldn't have that patience, right? So, so that's why we're talking milliseconds, not like half a second, uh, to wait. So, but if it's on the device, it just works in the flow of what's happening.

[00:11:41] Casey Golden: That's great. so before we get into the interviews,

[00:11:45] Ricardo Belmar: Yeah

[00:11:46] Casey Golden: what would you say the best moment of the event was? Keynote, demo, hallway conversation, water coolers,

[00:11:53] Ricardo Belmar: Yeah. Yeah. Well, here, here's...

[00:11:54] Casey Golden: after party

[00:11:57] Ricardo Belmar: Yeah. Well, I'll, I'll give you an example from, [00:12:00] the keynote. It was futurist Jonathan Brill. He had this whole metaphor going about the octopus organization, and the example was, two-thirds of an octopus's neurons are in its arms, not its head. So the arms explore and decide on their own while the octopus's brain coordinates.

[00:12:17] So his point was, an AI-empowered frontline team should work the same way? They're pushing intelligence and decision-making to the edge of the organization at the frontline, not waiting for it, not letting it get hoarded at headquarters. It's distributing the, the knowledge. You know, who doesn't talk about distributing the knowledge in, in these examples?

[00:12:35] When you think about it, it's exactly how all this AI technology should enable things to happen, and I, I think it's what we've been talking about across so many episodes of the show about how you empower and augment those store teams versus trying to rely on AI to replace them for, for any reason. And then another example was, they called it the power journey demo.

[00:12:54] There was basically a tour they, they gave us that kind of made all these things a little more tangible. They walked us through a [00:13:00] product's entire life cycle from factory to warehouse to store shelf to being delivered to a customer's doorstep. And every single step, they, they demo like how it relied on AI and RFID to stitch it all together as efficiently and, and error-free as possible, whether it was computer vision checking the assembly quality, at the factory, whether it was, readers scanning 1,300 RFID tags a second from 100 feet away as products being received at the warehouse, or a worker getting step-by-step guidance in their native language, whether it was for the assembly or picking and packing and shipping.

[00:13:36] It was the whole thing end-to-end laid out in that, a single demo u- using all these technologies stitched together

[00:13:42] Casey Golden: Interesting. So an o- an octopus org chart and a 1,300 tags a second magic wand

[00:13:51] Ricardo Belmar: Yeah.

[00:13:51] Casey Golden: Sounds impressive.

[00:13:52] Ricardo Belmar: Mm-hmm

[00:13:53] Casey Golden: So let's talk about these two interviews. What do we got on deck?

[00:13:58] Ricardo Belmar: Yeah, so first one, was, Tom [00:14:00] Bianculli. He's Zebra's ChiefTechnology Officer. He's the big picture guy. So you'll hear him talk through and wa-walk through all the announcements. We'll get into what I think is maybe the most important idea out of the event. We sort of touched on it already, this, this notion of pragmatic AI for the front line, making things more practical.

[00:14:16] Tom gives a healthcare analogy, which even though not retail, it's so interesting. He talks about, you know, giving nurses back 20% of their day. So you'll hear about that one. He talks about taking a, a 40% manual receiving dock process and turning that into "snap a single photo workflow," and it's done.

[00:14:32] So we also touch on the future of wearable technology, where that fits in, what he calls AI first computing. It's a very aspirational conversation, which I guess makes sense. We're talking to the CTO, right?

[00:14:42] But every example, he's got good, you know, good grounded examples, on something that's actually shipping. So we're not, we're not talking about pie in the sky stuff. It's all what's available now.

[00:14:51] Casey Golden: So I know I've been patient long enough. Patience is not my virtue.

[00:14:55] So before we dive in, we've got two quick [00:15:00] favors for our awesome audience. If you are getting value from the show, and I like to think that you are, hit like and subscribe in your favorite podcast player and on your YouTube channel so you never miss an episode.

[00:15:16] And if you really wanna help us out, leave a five-star rating and review on Apple Podcasts, Spotify, and Goodpods. It genuinely helps more retailers find the show, and that just keeps moving us forward.

[00:15:29] And then should we listen to the first interview?

[00:15:33] Ricardo Belmar: Yeah, let's do it. Here's my conversation with Tom Bianculli, CTO of Zebra Technologies, recorded live on site at Zebra's ZONE 2026 conference in Nashville.

[00:15:43] Interview - Tom Bianculli, CTO Zebra Technologies

[00:15:48] Ricardo Belmar: Hi, Tom. Thanks for joining me here today!

[00:15:50] Tom Bianculli: Hey, Ricardo. Great to be here. Thanks for having me!

[00:15:53] Ricardo Belmar: Why don't you give us a quick rundown of your role at Zebra and give us a little bit of an introduction to what you do there and then [00:16:00] maybe tell us a little bit more about some of the great things that were introduced at the ZONE event.

[00:16:03] Tom Bianculli: Yeah, sure. So, Tom Bianculli, Chief Technology Officer here at Zebra, and I've got the great fortune of working really closely with our customers. As you know, we do business with about 85% of the Fortune 500. And you know, we're right in their frontline operations, so we've got some strategy people. We've got lots we're doing, as many are, around AI. We're doing some real unique things with AI specifically for the frontline that we'll be able to talk about. And then, we're always looking out at how do we redefine the experience for those frontline workers? So we have an entire UI/UX design team that helps us do that.

[00:16:36] So, that's kind of where my focus is, and at ZONE, we're super excited by the announcements. We have a number of things on, on AI for the frontline. Something we call Workcloud WIO, which is integration and orchestration capability to be able to deploy these systems faster and connect real-time data from the edge with legacy-type ERP systems to extract insights and, and deploy those.

[00:16:59] We've got [00:17:00] Workcloud BI, which provides role-based personalized dashboards and information. You know, if you're a frontline worker restocking that shelf, or you're the store manager or regional manager, district manager, all the way up to corporate we're deploying unique dashboards into our customers' operations to help them run with the agility they need to deliver on their business.

[00:17:20] And then the, the one other update that's worth mentioning was in our press release for ZONE is around Nucleus, which provides a single portal, if you will, or pane of glass for being able to configure, manage, and deploy our devices. So just as, our portfolio of RFID readers and mobile computing and scanning and printing has grown up over the years those configuration tools have been, you know, unique to the particular part of the portfolio.

[00:17:45] And what we heard from a lot of our customers and, and certainly many of our partners is it would be great to learn one way of doing that and then be able to enjoy that commonality across the whole portfolio, and that's really what Nucleus is all about.

[00:17:56] Ricardo Belmar: Yeah, and I think that, that sounds like such a great great update on [00:18:00] that 'cause it really does make a difference. I, I can imagine for any, any of your customers, right, to have that single interface to be able to deploy and configure all the, all the devices that they're using.

[00:18:09] Why don't you tell us a little bit more from the perspective, if you were talking to one of those frontline workers, whether that's a frontline worker at a, a distribution center, warehouse, or in a retail store, and they're, they're used to working, of course, they have the devices, they're using them hands-on.

[00:18:23] But in terms of like what, what's new and how you're layering on new AI-based capabilities, how would you describe to them what the benefits are gonna be to their typical workday?

[00:18:32] Tom Bianculli: Yeah. So, you know, the way we, we think about that benefit is really returning time to that frontline worker to do more value-added tasks as opposed to, you know, rote tasks. So if you think about, you know, one of my favorite examples is in the healthcare space, where we're partnered up with a company around ambient listening that we have optimized that runs down on our device, and if you think about the nurse or the clinician, they spend twenty-five to thirty percent of their time doing documentation, right?

[00:18:59] And [00:19:00] I've never met a nurse that went to school to, to do documentation. They went to school to make patients better, to provide better care, to get people healthy fast. And so we're able to return that significant amount of time, like in the order of twenty percent of their workday to them to spend time in the areas that, their passion is focused on and where they went to school for, and then be able to do that ambient listening in the background that does the charting, that creates the notes, that enables for a seamless shift handoff from one provider to the other, you know, without having to be burdened with all that documentation.

[00:19:34] Similarly, you know, if you move into another completely different workflow, like the receiving dock of a retailer or the receiving dock of a warehouse receiving all of that merchandise and then getting it entered into the inventory system of that particular company requires a lot of manual operation, manual barcode capture.

[00:19:52] I was... You know, we're working with a large customer that has, out of all the manifests they have when they of shipments they receive, [00:20:00] which is about four thousand a week, they have about forty percent of them that arrive with no barcode on the manifest, so they have to manually enter all this information.

[00:20:08] And now we're using AI in the hands of that frontline worker, so the message to them, back to your question, is, rather than manually entering all that information, how'd you like to just take a picture of the manifest? Even if it doesn't have the barcode, we'll capture all that information, we'll parse it, we'll understand, what's the part number, what's the description, what's the quantity field, who's the ship from address, and we'll automatically enter that all into the SAP system without you having to go field by field, keystroke by keystroke.

[00:20:36] And, that's super empowering, super exciting. It allows that frontline worker to get more done more accurately and spend time on things, back to the healthcare analogy, that are just more value add for them.

[00:20:47] Ricardo Belmar: Right. Right. And that's a huge, huge time saver on, on, I think on any, anyone in those roles. So when you look at the technology itself one of the, the messages I, I heard a lot during ZONE was the importance of what you're doing in edge [00:21:00] computing and the ability on some of your newer devices to have that edge capability to do some local AI, local LLM processing.

[00:21:07] Can you talk a little bit more about the, the significance of that?

[00:21:10] Tom Bianculli: Yeah, so, we are really focused on allowing these workloads to run on the edge, and we, we've been using this term of tokenless processing because many of our customers are, many of our customers are looking at the ROI associated with running these workloads in the cloud which has significantly more cost than being able to run it in an optimized way right down on the edge.

[00:21:30] That's one. The other one is just from a security perspective lots of IT departments are adopting the philosophy of consuming the data that's generated as close to the source as possible. So you don't have to worry about security if the data's never leaving the device. If it's all running down to the device, then you don't have to worry about kind of that secure channel.

[00:21:50] So sec- so tokenless from a cost point of view, security's a big one, and then the other one that, caught me a little bit off guard, to be honest, was, was around the latency side of things. I just didn't... [00:22:00] You know, I hadn't realized that you think about the cloud, you think it's relatively fast, but when you think about some of these use cases like receiving or picture proof of delivery at, you know, someone's front door, if you're gonna make a go, no-go decision, this these AI capabilities need to operate in tens of milliseconds.

[00:22:17] So you can't have a second, two second, three second delay, like maybe we're all used to in a ChatGPT kind of interface. It goes off and thinks and comes back, right? And when it's thinking, the meter's running on the cost, and it's also, it's also actually slowing down your workflow if you're in these frontline workflows as opposed to speeding them up.

[00:22:37] So that's where the, running it on the edge, security, really low latency allows you to leverage it in the workflow, in line with the workflow, and save on those costs as well.

[00:22:48] Ricardo Belmar: Yeah. And y- you've also talked about this concept of physical AI in the space. So I have to believe, you know, with the benefit of doing more of that compute on the e- at the [00:23:00] edge particularly, you know, while as you were just talking about the latency involved and understanding your space, I mean, that time difference, that latency difference has to be a huge benefit to really be able to leverage the power of this technology to really understand the space that you're in and what you're looking at.

[00:23:13] Tom Bianculli: Yeah, yeah, totally, Ricardo, yeah, so from a physical AI perspective, I think so much of the what comes to people's mind around physical AI is, is the robotics, the physical automation side of things. But if you really unpack the definition, and this isn't, you know, the, just the way Zebra's looking at it, it's the way the industry looks at it, there's kinda two parts to physical AI.

[00:23:30] One is the physical movement, transport, robotics, you know, everything from application-specific robots to humanoids. And the other side of it, though, is creating a digital twin of the physical space. It's really taking the instrumentation of an environment, doing that in a, like an ambient sort of way so a, a person is not having to capture the information, it's just being autonomously collected.

[00:23:54] And so, to some of the examples we've been talking about, computer vision's a big one down in the handheld device, but we're also looking at [00:24:00] infrastructure-based computer vision, and then combining that with things like a domain-specific knowledge. So, to build that out a little bit further, think about shelf intelligence, where I've got computer vision running down on the device, much like I spoke earlier about snapping a picture of that manifest and it extracts the appropriate information.

[00:24:19] Snap a picture of a shelf, and it it finds all the stock-outs, it identifies which products are located potentially in the wrong position. Are any of the shelf edge labels incorrect in terms of what they're saying is there from a product point of view or from a pricing point of view? And it allows you to do that compliance in what would've been maybe tens of minutes in just a few minutes.

[00:24:39] And then back to the infrastructure side of things, think about cameras that are looking at a dock door or a cross-dock facility, and that camera is connected to a domain-specific, what's called vision language model, that understands what a good dock should look like. Like, where should the pallets be?

[00:24:56] Which doors should be having trucks leaving in order to [00:25:00] hit the downstream SLAs that those destinations have with their customers? And it's, it's monitoring that overall environment and then orchestrating the right actions to be dispatched to the right people in order to ensure, optimized performance overall.

[00:25:16] So that becomes really, really interesting when the idea of physical AI from a data collection digital twin perspective married up with this domain-specific vision or large language models. You could think about it as like someone that's maybe a 20-year veteran in a specific space, like the warehouse or dock receiving space, that could see everywhere within that facility all the time.

[00:25:44] So they can see everywhere all the time, what would they be dispatching and doing different? And that's what some of this AI is allowing us to do now, which is pretty cool.

[00:25:53] Ricardo Belmar: Yeah, I think that's pretty fascinating. And I, I guess as I think about it, that also really lends itself, the, like the analogy you use [00:26:00] you had this senior associate that has all the experience of what it should be like and how it, things should be done. If you're a new employee in that scenario, how, how are you benefiting?

[00:26:09] Like, are you, you, enabling new things that help them with their training, with helping with learning how to proceed on these tasks in, in, in a way that they couldn't do before that might make their, their day easier?

[00:26:19] Tom Bianculli: Yeah, that's, that's, that's a great one, and we see that especially with the customers we serve in many, many industries like retail, quick-serve restaurants, hospitality and frankly now, you know, I mean, manufacturing, warehousing. The turnover rates, you know, the attrition rates are in the 60, 70,

[00:26:35] Ricardo Belmar: Now turnover's

[00:26:35] Tom Bianculli: like in QSR, it might be, like, 80%.

[00:26:38] And so what's what some of the industry analysts now have been talking a lot about is the time to competency. So if you've got a new person coming in, how do you get that person up to speed as, as quickly as possible, number one, and then number two, how do you reduce the attrition? And we're seeing this help with both of those areas.

[00:26:57] So if you're new to the job don't [00:27:00] think about learning as, you know, an event-based, you know, I'm gonna now... Now, I'm gonna go take three hours of training or, or learning in the back room. We can actually inject microlearning right into the workflow. So if you know, maybe there's certain questions you're asking of these agents on the device, or there's certain workflows that you're struggling with, we'll identify that and then deploy down the right microlearning module that may just take two or three minutes.

[00:27:27] You know, think about it as like the TikTok of learning, right? You just get, like, a quick video that allows you to become, you know, more proficient in that specific area that's curated just for you. And with that kind of capability, along with removing kinda these rote tasks and making it easier to capture that monotonous information that's also creates more empowerment, more job satisfaction communication, collaboration skills we have on the device, so it's easier to message and communicate with your peers.

[00:27:56] And through those sorts of areas, we're actually looking and [00:28:00] are reducing the attrition rates. So you kinda have both sides of this, right? You know, increase people's satisfaction so they don't leave . And if they do leave and you're replacing them, allow them to get up that curve, much faster and in a much more dynamic and engaging sort of way

[00:28:15] Ricardo Belmar: Yeah, I think that that's a huge benefit for, for so many of these industries where they, especially where they have that high, high turnover rate.

[00:28:22] So maybe to close things out a, a little bit, if you look ahead to where things are headed, right? You have all these technologies, you're putting all of this capability in the hands of frontline teams.

[00:28:31] Obviously, the, the rate of change has been so fast the last couple years. We used to always ask people, right, "What does it look like five years out?" By now you can barely even wonder what it is five months out. Um, but if you were to look out, a year or two from now, how do you see this evolving and, and, and additional capability being delivered to the front line through what Zebra's doing?

[00:28:49] Tom Bianculli: Yeah, so we, we, we see a lot more of what we're calling the skills and capabilities that come down on our, let's call it, like traditional existing form factors. But what you're gonna [00:29:00] see from us even in the next just even one year is an expansion to more and more wearable technology. And the idea there really is, I mean, we invented that category in the enterprise space with you know, some of the largest transportation logistics providers and couriers out there where they, they wanted to use wearable technology so they could be moving goods like parcels, but be able to keep their hands, hands free, but have the technology right there on their body to be able to interact with that.

[00:29:25] Now, with AI, we're seeing that evolve to form factors where you may have a camera and an audio system that is seeing what you're seeing, it's hearing what you're hearing as well, and then it's able to cognitively understand that environment and help guide you through a situation. So really the computing paradigm becomes one where instead of the person having to interact or initiate the interaction with the compute, like we do today with applications, typically, you, you pick up your phone, you start the interaction.

[00:29:59] If we've [00:30:00] got... Exactly, and if, but if we've got AI and these form factors that are wearable, it can be seeing what's happening and let's say it tells me to pick an item from a certain location, it sees my hand reach out and go to pick that item. If I pick the wrong quantity or I pick from the wrong location, it'll be able to detect that and then give me a nudge.

[00:30:20] You know, it's like, "Hey, you're about to make this mistake," and prevent what could be a very costly downstream cost. So, lots of our customers are excited about this kind of closing the loop on the workflow, which

[00:30:32] is you send a task out or you send a pick order out, and you now hope it got done right, right?

[00:30:39] And you just deal with the error rates that you have today in terms of downstream customer satisfaction impact, net promoter scores, the cost of, remediating the situation. Whereas if we can catch it in the moment then we can prevent that mistake from ever happening, which is better for the employee, better for the customer, lower cost.

[00:30:57] And you can... You know, if you let your imagination run on [00:31:00] this, you can imagine both what we call on body, you know, wearables that are looking at maybe picking kinds of workflows. But you can also imagine fixed infrastructure that might be looking down, let's say, at a manufacturing work cell. And maybe I have a sub-assembly I'm working on, and I have all these bins from which I need to pick parts. And the SOP for assembling that item is known by the large language model. So it knows this part should go first, that part should go second, the torque wrench should be hit a certain torque when I put this screw in. And so we can send all of that, the video data, the torque data, and these are real examples.

[00:31:36] We're doing this together with, um, a company called Tulip. Tulip Interfaces that is instrumenting that work cell a- with a combination of vision and these connected tools, and then it's able to detect those quality issues before, you know, they either arise downstream, which are more costly to address, or they never get caught, and it, the, it gets all the way out the door and, and it gets caught in the [00:32:00] field at the customer's site.

[00:32:01] So those are really powerful capabilities, and the the transformation you were saying over the next two or three years is I think that AI agents, these skills on the devices, domain-specific agents are all gonna add up to transforming what the form factor looks like and really transforming what the future of computing is.

[00:32:22] And we've seen this, you know, obviously when we went to desktop to laptops, laptop to mobile, mobile proliferated into tablets and so on. And the interesting thing about each one of those changes is that the previous generation of form factor never completely went away, right? Like, I'm doing this together with you on a laptop.

[00:32:40] You might be on a desktop. We're both using our mobile phones. And so I don't think, you know, this doesn't mean that the form factors we make today, like handheld devices or mobile phones, go away, but they're gonna be augmented with this sort of new type of sensing and cognitive capability that will be at our beck and call to get done whatever we need to get done, [00:33:00] whether that's in an industrial context or it's from a consumer point of view.

[00:33:03] Ricardo Belmar: That's a super interesting examples. I, I'm looking forward to see how this evolves, and I-- you can just imagine all the efficiencies to be gained and the cost savings that that means, and all those things trickle down the line, right? Which a lot of people don't often think about, when you-- we're looking at these back-end capabilities, but it trickles all the way down to whoever that end customer is and the benefits they're getting indirectly from all of this having been done further back the chain that just enabled it to happen correctly and efficiently at a lower cost.

[00:33:31] Tom Bianculli: Exactly. I mean, not to not to extend beyond your question, but I just wanted... What you were just saying made me think of a a use case that we announced at our recent earnings call on what we call picture proof of delivery for a large transportation logistics customer.

[00:33:44] And, picture proof of delivery is you drop the package at the front door, you take a picture. And what we do is we automatically verify the proper picture is taken. We remove, using AI, information that shouldn't be in the picture, like I- personally identifiable information or people or addresses, whatever it may [00:34:00] be.

[00:34:00] And to your point around scale, the interesting thing about this is we take that PPOD event from maybe taking three seconds to get done to closer to one and a half seconds to get done, and that's tens of millions of dollars in labor savings and also in reduced claims costs for these transportation logistics carriers because, if you're some of the larger ones, you might be delivering fifteen million packages a day.

[00:34:25] So fifteen million times a second or two is a lot of savings,

[00:34:30] Ricardo Belmar: that's a lot of time. Yeah.

[00:34:31] Yeah

[00:34:32] Tom Bianculli: exactly right. And we- that's why we're really excited to pragmatically apply AI to those kinds of use cases versus just talking about, "Hey, AI can do all this sort of stuff," kind of the whitewashing that goes out on out there with, with regard to this AI washing that goes on.

[00:34:47] So, and we think it's a really pragmatic types of use cases that are, that are super valuable. And to your point, every little percentage point adds up to massive savings.

[00:34:57] Ricardo Belmar: Exactly. Yeah, it all adds up. I love that framing of [00:35:00] them as pragmatic use cases of AI because it's not. We always get, everybody always gets enthralled with these fantastical things of what AI is going to enable you to do, and sometimes it's adding up all of these individual little tasks and moments that can be made so much better and more efficient that when you take them as a whole, it's, it's a really significant benefit.

[00:35:19] Tom Bianculli: Yep. That's what, that's our tagline is "better every day." A little bit better

[00:35:23] Ricardo Belmar: there we go. It makes sense, yeah. Yep. Yep. Well, well, Tom, thanks so much for, for spending this time. I appreciate all the, the insights. This is really a, a fascinating time to watch how all this technology evolves and gonna enjoy seeing how, how Zebra plays in it

[00:35:37] Tom Bianculli: Likewise, Ricardo, and look forward to catching up as we advance down the path here that we spoke about. So thanks for your time.

[00:35:43] Ricardo Belmar: All right. Thank you.

[00:35:50] Interview Recap

[00:35:50] Casey Golden: And we're back. Ricardo, I have so many notes, but the line that's going to stay with me is the nursing example.

[00:35:58] Nurses spending 25 to [00:36:00] 30% of their time on documentation and ambient AI on the device, giving them roughly 20% of their workday back, impressive. Nobody went to nursing school to do paperwork.

[00:36:13] Ricardo Belmar: Yeah. That's, that's basically the theme of the whole conversation, right? I mean, Tom kept coming back to this idea of returning time to the frontline worker. The receiving dock example, same story. Customer with 4,000 shipping manifests a week, 40% arrive with no barcodes. So instead of having to type everything in field by field manually, now a worker can just take a picture, and the AI parses it straight into SAP.

[00:36:35] Casey Golden: And so then it scales in ways that are almost hard to believe. The picture proof of delivery numbers, shaving a delivery confirmation from about three seconds to a second and a half sounds tiny until you multiply it by 15 million packages a day

[00:36:55] Ricardo Belmar: Yeah

[00:36:56] Casey Golden: tens of millions of dollars

[00:36:57] Ricardo Belmar: Yeah, e-exactly. I mean, it's-- That, that's where his [00:37:00] phrase I, I love comes from, the pragmatic AI. It's not fantastical stuff. It's just adding up all these thousands of little moments that are made better. I think that's how you build a strong ROI for AI. Instead of trying to go for these gigantic, massive transformational things up front, build up all these little moments that actually add up to meaningful dollar impact.

[00:37:20] And, and of course, it even matches Zebra's tagline, "Better every day," so there, there's also that. Did you catch his workforce angle too? Turnover rates, right, hitting 70%, even 80% in some industries. He talked about micro-learning, two-to-three-minute videos, that are pushed to you exactly when the AI-based system sees that you're struggling with a task and, and not able to get it done.

[00:37:39] He kinda referred to it as the TikTok of learning

[00:37:42] Casey Golden: You know, damn me, but it works.

[00:37:44] Ricardo Belmar: Yeah

[00:37:45] Casey Golden: You know? I have to say it works.

[00:37:49] So this addresses both sides of the turnover problem. Get new people competent faster, and make the job satisfying enough that fewer people leave.

[00:37:59] I don't wanna... Nobody [00:38:00] wants to log into 26 apps, let alone like five. That's empowerment, not necessarily surveillance. It's not feeling creepy. And the change management framing there, matters. Sounds like a really good impact.

[00:38:13] Side note, we're going to have a whole miniseries on Data Blades podcast episodes coming soon. Next month?

[00:38:22] Next month on this topic, so stay tuned. So where does that leave the second interview for us?

[00:38:27] Ricardo Belmar: Okay, so if Tom gives us the vision, that they have, second interview, James Poulton, he gives us the whole machinery around it. James runs Zebra's mobile computing business, so every handheld, tablet, wearable that, that they deliver. He's also delightfully opinionated, I guess is the way I would put it.

[00:38:46] Casey Golden: Is that

[00:38:46] Ricardo Belmar: Well, I open by asking him what's the most overhyped and most underrated thing in AI right now, and I think his answer is gonna surprise a lot of, people in the audience. He thinks the large language models are what's overhyped for the enterprise right [00:39:00]now.

[00:39:00] Casey Golden: Okay, you could be a little bit more controversial

[00:39:04] Ricardo Belmar: I'm not so sure.

[00:39:06] Casey Golden: I look forward to meeting him.

[00:39:07] Ricardo Belmar: Yeah, yeah.

[00:39:08] Casey Golden: data please.

[00:39:11] Ricardo Belmar: I mean, his argument's that frontline work doesn't need the vastness of the internet. It needs really small domain-specific models that actually speak, you know, their vocabulary. So he has this great analogy about voice picking. You know, warehouses having run for decades on a 25-word vocabulary that workers rely on to pick things all day long with it.

[00:39:30] We get into how multimodal data capture, whether it's barcodes, computer vision, RFID, location data, it all adds up to what, everybody wants, a digital twin of their environment. And it kind of ties into the, the Nucleus product launch that we talked about earlier and what that changes for the enterprise to manage everything in one place, and that whole tokenless strategy we talked about before, too, as a huge differentiator.

[00:39:52] So listen to his challenge to customers at the end. I think that's also a good one.

[00:39:55] Casey Golden: Small models, big impact. I'm a huge believer in that. I [00:40:00] don't want, I don't care about anything out industry. I think we need that concentration of knowledge. And I mean, I'm still very pro about vertical software, like everything being verticalized. I think maximize that coverage, in your space.

[00:40:17] So let's hear it.

[00:40:19] Here's Ricardo's conversation with James Poulton, SVP and GM of Mobile Computing at Zebra Technologies.

[00:40:27] Interview - James Poulton, SVP & GM Mobile Computing

[00:40:32] Ricardo Belmar: James, thanks so much for joining me for a quick conversation here on, everything that's happening around, the announcement Zebra's doing here this week at ZONE. Mm-hmm. Plus, uh, all, all things AI, because what else would we possibly talk about these days, right, if not AI? And maybe that's the best place to, to start.

[00:40:49] If I were to ask you, in, in your view, given everything that you guys have going into product and solutions now and with that in mind, when you look at the market, what do you think is the most overhyped versus the most [00:41:00] underrated things about AI and capabilities that are out there?

[00:41:04] James Poulton: That's a really good question. You know, I think, um- Specifically for the enterprise space, I would say large language models, I think, are being over-hyped.

[00:41:13] Ricardo Belmar: Hmm. Interesting.

[00:41:13] James Poulton: Yep. And the reason I say that- Yep ... is because when you think about how they've been trained- Mm-hmm ... and their knowledge-

[00:41:18] ...

[00:41:18] James Poulton: And you think about for the enterprise, there's...

[00:41:20] We're talking- Mm ... about domain-specific, really specialty- Yeah ... kind of knowledge. And I think that, you know, what's being under-hyped is the value of domain-specific- Mm ... open source models.

[00:41:31] Ricardo Belmar: Mm-hmm.

[00:41:31] James Poulton: Some of the stuff that we're doing now, the building into- Mm-hmm ... the mobile computers so that we can actually deliver very specific workflow improvements for our customers using very specific data that matters to them, using the vocabulary that they care about.

[00:41:44] Mm. These are all things that are really difficult to train in kind of a broad model.

[00:41:48] Ricardo Belmar: Yeah.

[00:41:49] James Poulton: Yeah.

[00:41:49] Ricardo Belmar: And how, um, how would you rate, Well, I guess maybe two things I'm thinking of. So, a- as, as long ago as maybe nine months ago, right? Which seems like an eternity- Yeah ... in AI. Uh, I, [00:42:00] I think maybe a little bit to your point, there, there started to be a lot of talk about small language models, uh, and how relevant they were, and, uh, it seemed like that always went in hand with increased edge compute capability.

[00:42:11] So how would you kind of rate that a- a- as importance of having edge compute to really enable you to do the things that you're doing for frontline workers?

[00:42:19] James Poulton: Yeah. That's a great question. So, so first of all, I think that, um, one of the first things that we need to do when we're working with our customers really establish, like, what is the data set- Mm

[00:42:29] that they actually need to have access to? And I'll give you, like, a really old... This dates, dates my, uh, dates me for how long I've been around in the business. But years ago, and even to this day, they do voice picking with ABC123 kind of alphanumeric- Mm-hmm ... and maybe a 25-word language. And you can pick an entire pick list for the entire day- Mm super efficiently. So when you start to think about how much data does a, does a worker really need, how repetitive- Yeah ... 'cause they're doing repetitive [00:43:00] tasks. Mm-hmm. So very often the data and the support they need, I think, becomes repetitive. So we're starting off with looking at running kind of a small language model, small models directly on the device.

[00:43:10] There is an idea though that suggests that if at a point in time you need more compute than what the mobile computer can provide, then you have the option to scale over or scale up- Mm ... to a different source, and that could be the cloud. Ideally, though, it would be more of, of an edge, of an edge solution, an edge box type of solution.

[00:43:29] Mm-hmm. So we think that that's a pretty, a pretty compelling model.

[00:43:34] Ricardo Belmar: Are there specific use cases, particularly in, in retail, you think lend themselves the most to that, that kind of configuration?

[00:43:40] James Poulton: You know, it's, it may not be as much use case centric- Mm ... although it will be. I, you know, as we get more and more into video, I mean, vision AI- Mm.

[00:43:47] Yeah ... is really becoming-

[00:43:49] Ricardo Belmar: And

[00:43:49] James Poulton: very data intense ... what's

[00:43:50] Ricardo Belmar: driving- Yeah. Right,

[00:43:50] James Poulton: right ... so much of the workflow improvements that we're seeing. But I mean, we obviously do fit for purpose devices. Mm-hmm. So not every one of our devices has the same platform. Yeah. Not every one of our devices [00:44:00] has the same amount of memory.

[00:44:01] Now, do we wanna limit our, our customers from being able to participate in AI because they've chosen the right device? Ideally not. So this is where the idea of being able to offload compute to an edge box- Mm ... for example-

[00:44:13] to allow the edge box to do all the heavy lifting, and then return the results back to the worker, all of our devices can kind of manage those kinds of things.

[00:44:20] So these are, these are the ways that we're looking at serving a broad set of customers- Mm ... broad set of devices across a broad set of use cases.

[00:44:28] Ricardo Belmar: . So one of the things that I, I think is a big takeaway for me in the, the new solutions you are announcing this week is how you're applying the kind of integration, orchestration layer to tie all the different types of devices that you have into...

[00:44:42] I guess I, I would describe it as maybe one cohes- cohesive platform or interface- Yeah ... into that platform. Can you talk a little bit more about, you know, how, what it is that you're announcing this week, what it is you're releasing, and what the benefits are to, to a retailer?

[00:44:56] James Poulton: Yeah, I mean, I think that...

[00:44:58] I'll give you my perspective on- Mm ... on [00:45:00] how this looks. I mean, in essence, our customers ultimately want a digital twin of the environment.

[00:45:06] Ricardo Belmar: Yeah.

[00:45:06] James Poulton: That's what they're looking for. Buzzword maybe- Right. ... but I mean, it's a reality, right? Yeah. And as a data collection company, who better- Mm ... to be able to provide that, that vi- that, that worldview of- Yeah

[00:45:17] of their environment, right? So but what's interesting is, is that, again, the data comes in a variety of different sources. The- Mm-hmm ... the data that becomes valuable is, you know, what does a product look like? Can we identify what a product is? You may wanna know what pricing is. Is the pricing correct? Do you wanna know the location of that price or where that product is?

[00:45:33] So, you know, when we start to stitch all that together, and I... that's where you get into the kind of that, our platform view of the world- that that becomes highly, highly valuable. Because now with that data, you can start solving all sorts of different problems. And, uh- I think that that's certainly as I think about how my portfolio contributes.

[00:45:53] We're a data collection company. We read barcodes, we see images, we capture video, we do location, we're reading RFID, we're [00:46:00] capturing BLE. Like, none of-- any and all of those things could be, you know, we call it multimodal- Mm-hmm ... data capture, and multimodal, data is what I think drives these models and drives context.

[00:46:11] Ricardo Belmar: let me ask this sort of a similar question I asked Tom earlier. A- as you think about the, all the new capabilities that you're enabling with this, if you're explaining that to a frontline worker, how do you sell them on the idea that this is gonna let you have a better environment, a better work environment, it lets you do things in a, in an enhanced way- Yeah

[00:46:30] versus, you know, where they're at now?

[00:46:31] James Poulton: You know, I think it would be. Maybe that's one of the things that's more overhyped. I don't know. But I think it would be a bit of a fallacy to say that our workers and frontline workers- Mm ... don't want tools.

[00:46:43] Ricardo Belmar: Mm. Yeah. I

[00:46:43] James Poulton: mean, imagine- Yeah ... if I took your computer away from you- Right

[00:46:46] your phone, and said, you know, just, "But, you know, do your job." Yeah. Exactly. It gets pretty hard. Yeah. So I think, and some of the studies that we've done suggest that access to computers and access to technologies actually improves the life of our- Mm-hmm ... mobile [00:47:00] workers. They feel happier. And the, and the reality is, as we get to more agentic

[00:47:05] Ricardo Belmar: AI

[00:47:05] James Poulton: from- Mm

[00:47:05] just kind of generative, but when we get to more agentic, we're gonna be taking multiple steps out of their workflow process. Mm-hmm. And the reality is, is that the things that humans are really, really good at are gonna be the things that humans continue to do. And then, you know, some of those less, you know, appealing parts of the workflow, we can automate with, with our devices, and we're already doing that now.

[00:47:27] Ricardo Belmar: Mm-hmm.

[00:47:28] James Poulton: I mean, we've got some really cool use cases for package delivery, for example. I don't know if Tom talked to you about, about our picture proof of delivery.

[00:47:36] Ricardo Belmar: Yeah.

[00:47:36] James Poulton: But this is... You know, it used to be a five or six-step process- Right ... depending on- Right ... on the worker and, and- Mm-hmm ... what they were delivering.

[00:47:44] Now, essentially with one action, they can capture all of that information and capture- Mm-hmm ... it really, really well. It's not that the mobile workers don't wanna think, but everybody wants to do a good job. So when you know that the system is judging and making a, the, the judgment of whether the [00:48:00] picture is, is clear, is it of the right quality?

[00:48:03] Do we have the right background? Mm-hmm. Are we seeing things we're not supposed to? Mm-hmm. If we are, let's remove them, you know. If we can make all of that, then that worker knows that for the few seconds that they're doing that part of the work, they're doing an excellent job.

[00:48:15] Ricardo Belmar: Mm-hmm.

[00:48:16] James Poulton: So I, I think it's, I think it's really, really positive for workers.

[00:48:20] Ricardo Belmar: How does this ability you have, as I'm understanding Nucleus now- Mm ... uh, you have this new ability to really see all the devices, right? As a single environment. What are the advantages to the enterprise, right? That you're delivering that to, that they, for things that they couldn't do before.

[00:48:34] James Poulton: I think it's good to be grounded on what was and what is now. Mm-hmm. And I mean, really what we've done within the context of Nucleus is we've taken three or four disparate management systems- Mm-hmm ... and, and, uh, surfaces, and we've combined them all together. So if you think about some of our biggest customers, they don't just buy mobile computers.

[00:48:52] They don't just buy scanners. They don't just buy printers. Yeah. They buy all of those things. Right. Now imagine we just say, "Oh yeah, but just you're gonna manage them all [00:49:00] as these, you know- Separate device ... separate devices." Separate thing, yeah. And it- All

[00:49:02] Ricardo Belmar: these scanners over here.

[00:49:03] James Poulton: Exactly. It doesn't, it's not a great experience, so we really wanted to unify all of that into one pane of glass.

[00:49:09] But realize also that, we do a lot of work with MDMs, a lot of our great partners. Mm-hmm. We have, you know, MDM partners around the world. And there's an important space for them as well because we're an API first... Nucleus is API first base. Mm-hmm. So in, in essence, especially on the mobile computing side, we don't see it not including an MDM.

[00:49:27] The beauty is, is the MDM can stitch together into Nucleus and capture all of the capabilities that are unique to Zebra that the MDMs typically don't deliver.

[00:49:35] As well as those core functionalities for doing kind of a broader set of, of management. So it's really, really win-win. But when you look at our scanners, for example, nobody does scanner management.

[00:49:46] We were- Hmm. We are a best in class, and so if you don't create something like Nucleus, if you don't bring that to the table- Mm-hmm ... then you kind of leave some of these devices orphaned.

[00:49:55] Ricardo Belmar: Yeah.

[00:49:55] James Poulton: Nucleus will unify all of that experience.

[00:49:59] Ricardo Belmar: So if I were [00:50:00] to ask you, particularly with, uh, all of the customers that are coming to the ZONE conference this week, what, what would you say is the one thing you would want everyone to really take away from their time here this week?

[00:50:13] James Poulton: Wow. There's some... I think there's some really important messages.

[00:50:15] I'm... I think understanding what our on-device AI strategy is- Mm-hmm ... and realizing... I mean, we've had some pretty smart people say to me, "Well, on-device? I mean, of course it's on-device. It's on the device." And it's, okay, it's a little bit more than that. Yeah,

[00:50:31] Ricardo Belmar: a little more than that.

[00:50:31] James Poulton: Right? Right. So this idea of being able to process and do all the computation on the device, the models are hosted on the device, the inference comes from the device, so there's no connection to the cloud.

[00:50:41] Mm. And I think when you think about, I mean, just in the news last week, what did we see? We saw Uber announce that they'd spent their entire IT budget on tokens in the first quarter- Yeah ... of the year. Right. Microsoft said, "Hey, we're, we're clawing back some of our spend because we're, we're... You know- Mm-hmm

[00:50:57] token mania is kind of out of-" Yeah. "Our [00:51:00] token maxing is, is, is really hitting us." So our customers, as they experiment with this, are, are worried about it. They don't know how to calculate it.

[00:51:08] So ideally, the ability to step into AI in kind of these closed loop models where you know you're not gonna consume any sort of tokens is really low risk.

[00:51:18] Mm. And I think it's a very unique position for Zebra. We're not seeing anybody else- Yeah ... really play up this part of it. And as I said, as we learn more, as we develop more capabilities, and AI is changing on a weekly basis, you know, we have ways of scaling over and scaling up while still staying tokenless, which I think is a really- Mm

[00:51:37] important message. Yeah. So really, hopefully our customers at the end of this ZONE, we're... I'm gonna ask them to be demanding customers. Come to us and make us show you exactly what we're talking about- Yeah ... and show you how it works. Yeah. So it's, it's really, it's really cool.

[00:51:51] Ricardo Belmar: No, that's great.

[00:51:52] That's great. What, what, where do you see this evolving to in the next, couple of years?

[00:51:58] James Poulton: Wow. If I knew [00:52:00] that. No, I mean, listen, I think there's no doubt that our customers, you know, are suffering from a variety of issues when it comes to labor- Mm ... when it comes to the cost of labor. Mm. So to create productivity, to create efficiency, I mean, it is really kind of- There really are no bounds.

[00:52:17] I mean, one of the things I find really exciting about AI and when I talk to my development teams and- Mm-hmm ... who are thinking about new products is that, you know, and this isn't... I'm not taking credit for this. This is kind of a common thing, but, the AI is limited by your imagination. It's not limited by atoms.

[00:52:34] It's not limit- Mm-hmm ... you know, it's, it's literally what you can't imagine is what you can't build. Mm-hmm. So I'm excited to see all of the great new ways that we're gonna improve our customers, our customers' businesses over the next few years, and I can guarantee you that it will include AI.

[00:52:53] Ricardo Belmar: That's a safe bet, yeah.

[00:52:54] James Poulton: Absolutely. Oh, and RFID, but anyways.

[00:52:56] Ricardo Belmar: And RFID, yeah. Yeah, we could- I'm not talking

[00:52:58] James Poulton: about RFID today.

[00:52:58] Ricardo Belmar: Right, right. We could... That's a [00:53:00] separate podcast, for just, just on RFID. Right. Well, I gotta say thank you so much for, for joining me for this quick conversation. Yeah. Uh, lots of exciting things happening this week, and it's gonna be interesting to see how customers take advantage of it and where they go.

[00:53:15] James Poulton: Yeah. Well, thanks for coming, and I hope you... You're here all week, right? I hope you enjoy it.

[00:53:19] Ricardo Belmar: Yeah. Thank you. Awesome. Appreciate it.

[00:53:20] James Poulton: Okay.

[00:53:20] Ricardo Belmar: Thanks.

[00:53:21] Interview Recap & Big Takeaways

[00:53:27] Casey Golden: And we're back one more time. Okay, Ricardo, "AI is limited by your imagination, not by atoms." James Poulton gets the quote of the episode

[00:53:39] Ricardo Belmar: Yeah, I suppose he earned it because everything before that quote was pretty grounded. The model is simple, right? Start with a small language model running right on the device. If you ever need more compute, then scale over to an edge box server. It might be on-prem, not in the cloud. You still stay tokenless the whole way, saves on cost, still delivers the AI.

[00:53:58] And when he brought up companies [00:54:00] blowing through entire IT budgets on tokens, you understand why they've got customers that are leaning in on this.

[00:54:05] Casey Golden: That part I appreciated most was his answer on frontline workers, that it's a fallacy to say workers don't want tools. Do we need... Do they want crap s- crap tools and crap software?

[00:54:16] Like, that's what they keep getting. And he was like, "Nobody's been making an investment in it." It's like the la- it's like the end of the budget and the last pick, and nobody asks them whether or not they're interested in this tool in the first place.

[00:54:28] Ricardo Belmar: exactly

[00:54:28] Casey Golden: So take away anyone's computer and phone and say, "Do your job," and see how that goes.

[00:54:33] Their studies show access to technology actually makes mobile workers happier!

[00:54:39] Ricardo Belmar: 100%. And agentic AI takes the steps nobody enjoys out of the workflow. How often have we been saying that on the show, right? Package delivery going from a five or six-step process to essentially one action. System quality checking the photos so the worker knows they nailed it every time.

[00:54:55] There's no questions about it. There are no errors. Humans keep the work humans are great at. You talked about the [00:55:00] Nucleus product story, three or four separate management systems collapsed into one. It's API first, so it plays nicely with any mobile device management partners that a, a customer may be working with

[00:55:09] Casey Golden: So big picture time. You sat through two days of this.

[00:55:12] Ricardo Belmar: Yeah

[00:55:13] Casey Golden: Sounds pretty content heavy.

[00:55:16] Ricardo Belmar: Yeah

[00:55:16] Casey Golden: A lot of, a lot of opinions, a lot of coverage here. Give us the big takeaways from Zone 2026.

[00:55:24] Ricardo Belmar: So I'll, I'll, I'll give you four.

[00:55:25] So first, edge, that's where the frontline AI should live. It's tokenless, on-device AI. It's solving the cost, security, we didn't really go too deep into security, but both the interviews talked about it, and latency all at once.

[00:55:39] And they made the case that you can't run frontline workflows on cloud round trips. So you want that immediate response.

[00:55:45] Second takeaway, AI on the frontline, it's about augmentation, not replacement. Where have I heard that argument before?

[00:55:51] Casey Golden: I don't

[00:55:52] Ricardo Belmar: I don't know.

[00:55:52] We've only mentioned it, like, every episode for the last,

[00:55:54] Casey Golden: right? Everything was like...

[00:55:56] Ricardo Belmar: Yeah. Every customer story w- was, was about a fixed labor [00:56:00]pool meeting growing demand. You can't hire 20% more people or pour 20% more concrete to solve a problem, so you have to give the people you have the superpowers to get the work done better, easier, more efficiently, all those good things.

[00:56:12] Third takeaway, consolidation is the new innovation. So think of the, the super app they introduced, the Nucleus management console, that Workcloud IO system with the added orchestration layer on top of the 70 or 80 apps. The breakthrough's not that there's a shiny new capability or feature, honestly.

[00:56:29] It's that you can take this huge mix and match of, of other apps and other stuff that, retailers are trying to give their frontline teams and collapse it down to, one management console and a single experience that people can actually use. It's about simplifying all the complexity.

[00:56:48] And then last takeaway, you know, the proof is in the small numbers. A, a second and a half per delivery, four clicks versus eight to complete a, a task, two-minute training videos, not an hour. Pragmatic AI wins on [00:57:00] the accumulation of all those tiny moments that matter to build out a strong ROI, because you scale that out to the millions of interactions that are happening across the enterprise, it starts to add up.

[00:57:11] Casey Golden: And here's why this episode feels like it belongs in the show. This is the thread we've been pulling all along.

[00:57:17] We've spent so many episodes talking about frontline workers and the real differentiator in retail that your associate experience is your customer's experience.

[00:57:27] Ricardo Belmar: Mm-hmm.

[00:57:28] Casey Golden: Everything you just described in this technology industry finally catching up to that

[00:57:33] Ricardo Belmar: Yeah, ex- exactly. Exactly. I just think back to our conversations about employee experience and retention. We kept saying the associate's the most under-invested asset in retail. Now we're seeing turnover and time to competency treated as a solvable technology problem. Just wait till, next month, when we get into that three-part Data Blade series where we've got the National Restaurant Association coming on, on the show to go even deeper on this point.

[00:57:57] We also talked about how store associates get [00:58:00] buried in tasks that come down from corporate, never get time with customers. We've had so many examples now with this of AI that prioritizes the task list, shows you how to do the task, verifies it's done, just so you can get back to the customer, which is what you're really supposed, want to be doing.

[00:58:17] Even The unified commerce conversations we've had. I talked about that, demo at ZONE. That was the buy online pickup in store demo, and everything went end to end and how it was monitored. And it's this whole connected store experience that we keep talking about on the show finally running end to end.

[00:58:31] Casey Golden: The aspirational part for me is that the frontline worker of the next few years gets the kind of intelligence support that used to be reserved for HQ.

[00:58:43] Jonathan Brill's framing was that human plus AI can give a frontline worker the judgment of a senior consultant.

[00:58:50] That changes who gets to make decisions in retail. And I think that can unlock a lot of decision-making down to[00:59:00]

[00:59:00] Ricardo Belmar: Right, Exactly

[00:59:01] Casey Golden: you... It doesn't have to go up, and you can be faster, and you can respond faster, and that's, that's the octopus organization in action, right?

[00:59:09] And it's once again an example we keep coming back to of augmentation versus replacement.

[00:59:15] Ricardo Belmar: Yeah, absolutely. Absolutely. I think that's probably the perfect note to end this episode on.

[00:59:19] So let's give a huge thank you to Zebra Technologies for the invite and the access, and to Tom Bianculli and James Poulton for two really great conversations.

[00:59:27] Casey Golden: Perfect. Well, Ricardo, this bonus episode is officially a wrap!

[00:59:32] Show Close

[00:59:37] Casey Golden: If you enjoyed this bonus episode, do us a favor, follow the show, leave us a rating, share it with a colleague who cares about the frontline as much as we do, and keep an eye out for Ricardo's companion write-up coming to our Substack newsletter.

[00:59:53] I'm Casey Golden.

[00:59:54] Ricardo Belmar: Don't forget to follow us on LinkedIn, Bluesky, Threads, and Instagram, and subscribe to our [01:00:00] Substack newsletter for more highlights from this event. For transcripts and guest info, visit retailrazor.com.

[01:00:05] I'm Ricardo Belmar.

[01:00:06] Casey Golden: Thanks for joining us on the Retail Razor Show, part of the Retail Razor Podcast Network.

[01:00:12] Ricardo Belmar: Until next time, stay sharp, stay human, and stay ahead.

[01:00:15] This is the Retail Razor Show.

[01:00:17]