Customerland
Customerland is a podcast about …. Customers. How to get more of them. How to keep them. What makes them tick. We talk to the experts, the technologies and occasionally, actual people – you know, customers – to find out what they’re all about.So if you’re a CX pro, a loyalty marketer, a brand owner, an agency planner … if you’re a CRM & personalization geek, if you’re a customer service / CSAT / NPS nerd – you finally have a home.
Customerland
Beyond the Shiny Object: How AI is Transforming Retail Operations
The retail landscape is transforming through artificial intelligence, moving beyond flashy demonstrations to delivering real business value. In this illuminating conversation with Matt Bertucci, Director of Retail Solutions at Lenovo, we unpack how AI in retail has evolved dramatically over just three years.
Matt reveals how retailers have progressed from asking "what is AI?" to implementing solutions that deliver measurable ROI while enhancing both customer and employee experiences. Unlike the fear-based narratives about technology replacing workers, today's retail AI focuses on augmentation—handling routine tasks so staff can engage in more meaningful customer interactions.
We explore fascinating real-world applications, from dynamic digital signage that personalizes as shoppers approach to AI-powered return kiosks that not only streamline operations but create additional sales opportunities. Did you know that over 50% of customers who enter a store to make a return end up shopping during that visit? Matt explains how AI makes this process faster and more satisfying while helping retailers better manage inventory and discover new business insights.
The conversation delves into the technical infrastructure making these advances possible, including Lenovo's "AI factory" concept that provides the necessary computing power for everything from natural language processing to visual recognition. Matt shares how retailers of all sizes can implement these solutions, with approaches ranging from on-premises systems to cloud support or hybrid models.
Looking ahead, Matt predicts AI will become increasingly conversational and natural in its interactions, handling multiple languages seamlessly while maintaining that crucial human element that makes shopping a social experience. For forward-thinking retailers, the message is clear: the AI revolution isn't about replacing the retail experience—it's about making it better for everyone involved.
What AI solutions would improve your shopping experience? Share your thoughts and join the conversation about the future of retail technology.
The intent is never to pull away from jobs, to pull away from what, how stores run and why. It's a social interaction kind of driver for people. You want to be there and have that. But it's about augmenting what's being done for multiple reasons.
Speaker 2:Today on Customer Land, matt Bertucci, who is Solutions and Services Specialist Manager at Lenovo, or much more succinctly how do I say that more succinctly, matt?
Speaker 1:Director of Retail.
Speaker 2:Solutions. There you go, uh. North america in latin america, latin america and, as I understand it, amia. Okay in your title and in your geography you've got the whole thing, basically nailed there, yeah, so well, well, thank you for joining me.
Speaker 2:Um, just to to let people know who have no interest in this whatsoever, but Matt and company and I have been working on setting this up no lie for probably at least two months, probably three months, with missed schedules and miscommunications and, you know, two ships passing in the night multiple times. We finally made it so and I've been looking forward to this for some time. Having said all that, the kind of genesis of du jour for the past three years at NRF has been AI, but there's been an evolution. As to what that has meant, I'd love to hear your take on it.
Speaker 1:Yeah, I mean it's kind of following the pattern of how GPU involvement's been to some degree. I think you see a lot of what our partners are doing and that's driving those discussions. Year one was just general AI high level. What is it? What can it do? Retailers are probably at the forefront of interest and that they were out there looking at it. I'm trying to understand. Some started dabbling in and where can it go? Um, you know some of the work that we've done with with kroger around loss prevention was very much on the forefront. We're now going in year two of that program.
Speaker 1:So development was kind of starting two years or three years ago. The second year was the generative AI age, right, how does that now start to relate into larger language models? Chat, tbts became popular. You could ask questions, get answers. Where does that take us? What are the product lines that retailers can utilize that type of solution for All this, while everything's kind of been developing?
Speaker 1:And then this past year has been what they're calling the launch of the agentic AI, which is taking what you worked on before and really helping fine tune that.
Speaker 1:And when they talk about agents, it's about having that model in place that can improve on itself through additional questions slight, easy tweaks, but gradually becoming more intelligent with the responses and more accurate which is most important, more accurate with responses. So we're seeing that developmental growth in that space and it's really now at a point where I think OEMs likeenovo can come forward with true solutions that work, that are somewhat plug and play. Everything's going to be customized right, but, um, because that's the the nature of retail you, you add your secret sauce and it becomes yours and that's what makes you successful. But from a baseline, constructive perspective. That's really where we're leaping forward and putting together that base construct now for customers to more easily spin up the testing for that agent as it goes through all of the tests and finalizes. How do you make that kind of a hybrid launch where we're helping support from some degree and service that to you as a customer, developing it into an on-prem item Because you need someone at ROI to be able to do that shift.
Speaker 2:Sure, yeah, and that's a whole other conversation. How do you present that up front? How do you even understand that up front? I think it might be helpful just to set a little context about your role at Lenovo and again, this happened just like seconds before we hit the record button. But what you do, if I understood correctly, sits right in between sales, which tends to be high level conversations, and development, which tends to be high-level conversations and development, which tends to be code level, and you're kind of interpreting between those two groups and the customer.
Speaker 1:You have that vertical space correct where gain interest, bring in some expertise around that to focus it and really make sure we're going in the right direction and what are the outcomes wanted, so that then the right expert can be brought in to really drive it home and make sure that it works and runs and can be implemented properly.
Speaker 1:At the same time, there's a breadth to it Because you're talking about different levels of hardware and I'm supporting how do we bring forward handheld devices, scanners, phones, desktops, pos, edge devices for that close to the store compute need for fast, especially as you start building in some other AI services like loss prevention, things like that that need a fast response but carry a large data set. That's where your edge server all the way to data center. And now the new construct, which is side-by-side to the data center, is the AI factory. Which is side-by-side to the data center, is the AI factored. So I get to talk to all of that as well as how does Lenovo help service that? How can we bring it forward so that, whether you're a small provider with only 100 stores can compete with the large provider of a thousand stores Very different kind of funding sets, but Lenovo can then help service in different ways, whether it's building out on-prem or helping support cloud on our structure or somewhere in between, which is now that hybridized world?
Speaker 2:Gotcha gotcha. I'd like to talk just a little bit more about kind of what you were mentioning earlier, which is the evolution of AI in retail, at least from you know, kind of the NRF standpoint. I mean, you're deep into it. I am not, admittedly, not a participant. I'm an observer here, so I get the the luxury of standing on the sidelines and commenting on the work, the actual work people are doing. But you know, um, I think you nailed it that, uh, three years ago the discussion was AI, but it was kind of like we have AI, we're we're deploying AI, we have an AI-powered box of whatever it is, and it was more about the shiny object. And then year two was we started seeing some really interesting deployments, like, okay, wait, there's utility here, and we all knew there would be. But let's face it, if you look back to what was being demoed three years ago, it really was about the shiny object.
Speaker 2:I mean we've got that. And year two became about listen, we can deploy this in some really intelligent ways, and, at least from my perspective, year three became much more about look, we can demonstrate an ROI here pretty quickly, and the bottom line oriented conversations were much more present than they had been.
Speaker 1:And in combination of all of that, you're now year three. You're you're actually getting those deployments to be outcome specific, right? I think at first it was shine like you said, shiny object, but it was so open to interpretation of what that's going to do there was a lot more fear around it. As you develop year two, people started to see how it was supportive. Now we're seeing it as truly we're not taking away any jobs.
Speaker 1:This AI is not withdrawing from what you've got, but really augmenting that. How do you? Because for our retailers, they have two customers their employees and the end customer who comes in. End customer is about experiential. Where can AI do that? And that's kind of a shift that we've been doing with a large project that is coming to conclusion now with dynamic digital signage. That shift from paper to screen. It was never going to happen until you could bring in the AI aspect of it and make it dynamic. Now it has value Now. So the retailer has a way to return on on those funds but they're going to draw more customers in because the experiential aspect of it you know, add changes as you walk up.
Speaker 1:It's like oh, you know, that's for me Watch it, which is pretty wild.
Speaker 2:If you've never experienced it, it's kind of cool.
Speaker 1:Yeah, it definitely is, but it really kind of it's meant to do the things that your employees either don't have time for or have some shyness and don't get to it, because it's a lot of add-on. It's how can you advertise something to be additional to what they've got? Or, um, how can you bring forward that avatar that you know is at the customer service desk? When they say I need a toilet, what lyle is it in? It says it's an aisle three. But don't forget, 80 of the customers who buy that also need to buy this piping because it breaks right. And and so now you're, you know, got the customer thinking, oh my gosh, that's right. 80% of the customers who buy that also need to buy this piping because it breaks Right. And so now you're, you know, got the customer thinking, oh my gosh, that's right. Um, I do need to change that out, cause that's gonna break.
Speaker 2:Yeah, yeah, just beyond the cool factor, and I have to confess I'm still I'm still very much, uh like, obsessed with the shiny object nuts of a lot of this stuff. But this past year, if you remember, at some of the entrances to the show floor they had an avatar that would give you directions. Yeah, which was? I think it was there primarily for the cool factor, because I noticed that it stumbled a handful of times, so the tech wasn't quite there.
Speaker 1:That's why ours isn't releasing until a couple months from now. There you go we have. We've done a massive undertaking on this and we've got Lenny, who is our avatar. He launched at the GTC Summit a few weeks ago. We've broken through that gap right and that's really been where a large focus has been from where we were two years ago. Why we didn't launch it at NRF? Because it still needed that tweaking to really be comfortable that we can prove that it can launch properly. So now we're at that point. But, to your point, there's always some growth to it and that's where this agendic AI comes in, because it's learning on itself and it now hears questions in different ways and can give the right answer consistently to each of those methodologies versus, you know, chat GBT great for what? What it was, but how often was it always right?
Speaker 2:yeah right and continues to have some, some significant issues. I don't know if if you want to talk about this, but I'm curious as to what the specific reasons were why lenovo held off and what you were trying to solve for in that interim period, because, you know, the basic technology was there, but in my view and I didn't speak with the vendor behind the technology, so I don't know this but the stumbles that I saw seemed to be more related to its understanding the question. It just really wasn't reading the prompt very well. Yeah, which seems to be very solvable, but it just wasn't doing it.
Speaker 1:Seems to be and is often misconstrued. Sometimes.
Speaker 2:Yes.
Speaker 1:The simplicity of the thought of what it needs is why we've held off, because there's the complexity behind it. And how do you go through the right checks to avoid the rag, that concept of making sure you're not getting the hallucinations, you're not getting the, the quirky outcomes? Um, what are the tweaks needed for asking the question? You know how many times you ask it in different ways. Did you take the time to go through five or did you take the time to go through a hundred?
Speaker 1:right the is going to give you a much better readout of where you're off and to make those adjustments. Uh, additionally, we're now at a point where a year ago, was this development was taking place, we weren't that, which is being able to self-check and kind of loop that internally to the system before it goes out, and having the right internal structure and server set. And that's why we now have what's called the AI factory, because those are kind of GPU towers that work next to your CPU towers and what these do is all of that language, model, visual models, being able to combine the two, and as you build it up and build it out, it becomes better and smarter and it's much easier to integrate in, because you can take five racks, build out another five. All you have to do is have the right connection. Now you have a 10 talking to each other.
Speaker 2:Right right.
Speaker 1:And it's about the communication, and so that's a structure that we weren't at. It was in development but we weren't at before. Lenovo's now got that we're building out libraries as well. Lenovo's now got that we're building out libraries as well. Um, so that's one of the the factors that's helping you know us really kind of speed our process here and working with companies to be able to say, hey, give us what your outcome you want within these guidelines of what we can do. You know, and it's pretty broad, sure, but if you want to do inventory control and you want to utilize, you know, your four cameras that you've got and have them lined on the shelves, knowing what your, your data plan is, your planogram, you can now track your, your inventory, uh, and that's in a way that we couldn't do before, but we can build that up and get it spun up and have that agent built and running minimum 90 days approximately.
Speaker 2:No kidding.
Speaker 1:Because we built that out. So we've got our data sets built out. That's really a big change that we've gone through.
Speaker 2:I think you touched on this a little earlier, but one of the other big themes from this year's change that we've gone through I think you touched on this a little earlier, but you know, one of the other big themes from this year's NRF was how AI is being deployed to solve for returns, to solve for shrinkage and just a huge number of supply chain issues. But I wanted to talk with you a little bit about yeah, you can't see, but we're both chuckling at what that really means. Lenovo is deeply involved in that whole space. How much time do I?
Speaker 1:have to talk.
Speaker 2:You just keep going. There is a lot to talk about there, and I was actually afraid of opening us up because you know we could dogleg and never return. But you know, if we can, why don't we just start with how you and Lenovo kind of see that space and the big problems and how you're approaching it? And we'll start there that space and the big problems and how you're approaching it. And we'll start there.
Speaker 1:So I think from a return, one of the things that we're seeing is this is where the term omni-channel which when the inception of it during COVID was just online buying, it's really becoming an omni-channel, right, multiple channels coming together, because you're finding more people, even if they buy something in the store, get home, decide they want to do a return. More specifically, it's with when you're buying online. It comes home, even if you're doing an in-store purchase through online. Let's say it's a cold purchase, right, that gets shipped to your house. It's the wrong size. You go online, you say I want to return it. What we're finding is the cost for all of that is kind of outweighing the benefits of some of the online returning. So the options that are being given is to do it free. You go into a brick and mortar, right, and that's bringing you back into the store.
Speaker 1:Now where AI is coming involved is you can now start adding the questions behind it Was it size, was it a color? A little bit more, what's the reasoning? And it's a quick answer, but it can preempt what that qr code is going to look like. So when you get into the store to scan your qr code, it can now prompt and you can do it now at a kiosk. We have those from a hardware perspective the ability to bring now a returns kiosk in which automates that return system. So the prompts and questions that you built into this now carries some ai to it. But it's going to ask all of the upsell keep it in the store questions that a person might not have. They might just go through the minutiae and say, all right, have a good day, and you walk out as you go.
Speaker 1:Whereas this is okay, you said it was for for size. You needed, uh, you know, a size 10 and a half, not a 10, in that shoe. Is that correct? Yes, by the time you've done your return, it now says go see the associate in the shoe department because they're going to have the shoe ready for you. So they knew what store you were going to go to the return. You kind of went through what size you needed. So it preempted the store to make sure they had that in stock, which can be done fairly quickly, and now, as you're returning, it's been there waiting for you, right. So now you're not losing sales, you're more often. You know, with some of the Amazon returns that go into the Whole Foods, they find that over 50%, closer to 54% of people that go in to do the return then shop that's massive, it's a monster. Yeah, and in part, return than shop, that's massive.
Speaker 2:It's a monster, yeah.
Speaker 1:And in part it's because it's fast. Now they've got time, they can actually go through, and you know for roofs. So if you're going back to a Kohl's and doing a return, why would you not want it to be fast, so that now you also have the benefit of sending them somewhere where they've got the product waiting for them?
Speaker 2:And they can start adding and it can say oh, by the way, here's a 20% off for these jogging shorts that go great with those sneakers All of which takes an enormous amount of compute power. I mean, and that's the AI factory side, yeah, factory side, yeah, there's just so much that has to go into making those decisions real time, because you're pulling data from all kinds of different sources.
Speaker 1:But you can see, from a benefit of the retailer having that is going to provide ROI. It's going to provide dollars back into the store kept in the company that otherwise would have gone out. I mean, even even amazon admits that most of their sales come from people not finding something in the store and going online and just ordering it right, right um how do you, how do you counter that?
Speaker 1:as a brick and mortar retailer, returns is a great way to do that. At the same time, you're now able to understand where did that order come from. Was it actually from a store? Was it online? What type of product is it? And it can create your returns barcode to get pre sorted to the right direction, because there's different sorting areas based on what the product is. Can it be something that can be resold, repackaged, resold, or is it something that has to go and get waste right and written off altogether or through secondary sales market um?
Speaker 2:not to mention the, the improvement in customer satisfaction through that whole, that whole kind of series of experiences.
Speaker 1:But it speaks now into your inventory right, your inventory controls and your distribution centers of what's getting shipped where and really understanding that it's a massive amount of data. But the actual analysis that you get from it drives improvement, specifically, more and more, and it spins up greater and greater improvement and it potentially opens doors for ways to tweak your business accordingly, because maybe you're finding something you didn't know before. I mean, this is Kroger. With the LP, I think it was a 10% change, you know, in margin over. You know which is a one and a half percent of their margin, but when you equate it out to their sales it was billions of dollars. Yeah, same, and I mean what sounds like little numbers was massive. But they've also found that it opened up windows of insight into how business can be altered a little to help help alleviate some of that and new opportunities arose as a result.
Speaker 1:Exactly.
Speaker 2:Yeah, a great example. I'm just sitting here wondering whether we want to go down the supply chain route, because that's I'd rather not, if we could.
Speaker 1:Okay, fine.
Speaker 2:That's not as much my expertise.
Speaker 1:Uh, I'm more within the brick and mortar store supply chain starts becoming your manufacturing side. Um, how you know how robotics gets involved in that, which is totally different it's somebody else's deal totally different article for from what?
Speaker 2:okay, good, you just saved us.
Speaker 2:You saved our afternoon here, I think. But I do want to talk a little bit about because I think you're in a really unique position because of who Lenovo is, because of what you do at Lenovo, which I think is a really, really interesting like position within the Lenovo or kind of customer ecosystem, like you're right there at this this kind of meeting point of opinions, needs, wants and possibilities and and and a really really fascinating and important space. But but, having said that, you know I would love to get your kind of high level view of what you think the future of AI is in returns, like, right now, we're on the precipice of doing some of this. Well, people are doing what you're talking about, which is futuristic, but you know, let's look, let's look two years down the road, if we can, and and what do you think that's going to look like? Because I think I I think that the people listening to this podcast and who read the customer land uh content are already looking down the road to that, to those kinds of next chapters.
Speaker 1:Yeah, I mean, I think it's. It's a growth factor where we're getting you know further into the land of possibilities, into the land of actualities, right, it's kind of that that if you build it he will come type of thought process. As much as we've been on this vertical ride of compute change and how systems support the stores, it's almost leaning even more vertically now when it comes to what can AI do to assist in all of this. We're really seeing that ai is is becoming smarter, it's becoming taking on that ability to to check itself, become a check and balance on itself to some degree, but it's still not taking away from the human factor. It still needs to be experiential and I think that's where we're seeing and it's going to become more improved, with what that omni-channel experience looks like, with returns being able to take it away from the employee side, a little bit sort of like what we're doing with some of the checkout systems. It takes it away from the employee side. It gives you as a customer a little bit more feel of control over what's getting done. But now the employees which are hard to come by for multiple economic reasons the retention on that is lower. You're giving them now projects that are more enjoyable and it becomes a post transaction or pre-transaction interaction.
Speaker 1:So your employee gets to come after and say I've got this for you. This is what you were asking for. Can I help you create a full outfit? Can I help you get the other parts to this? You know this was an expired item. I apologize. We'll help you out and fix that for you. But, by the way, did you get all of these other items for your grocery list? Yeah, you know. So it's really driving in that manner and making what the employees do from a company perspective more valuable.
Speaker 2:Right.
Speaker 1:They get to focus on more value added and your retention is going to hold up better for them because their satisfaction is done. The minutia of what transactions are is off their plate. Price changes can be done elsewhere through technology, which we can provide. And for me it's great because I talk about all these areas and it's about opening the eyes of the retailer to what they could do and how can they get there. What should be a first step, second step, third step, based on where the ROI comes in and what can then help support funding for the next level. But when you're looking five years out I mean you've been at NREF for the last three years when we were three years ago and today is so astronomically different. In some degree because of the clarity we've got, I mean I couldn't tell you where it's gonna be in five years, but what I can tell you is over the next two years, as these projects start getting implemented, you're gonna see perfection made on them. Those that are earlier doctors are gonna find more benefits long term because they can more easily stack on.
Speaker 1:It's kind of like we use one of our first AI users. They're already looking at three others now and that's what's happening. You find your first use case. You get funding from it because it feeds back in. And now you start opening your eyes and, oh, I can do this or I can do that, and those then start building up and we can support the customer through all of these different levels.
Speaker 1:And that's one unique aspect of Lenovo as an ODM is that we have that handheld device to data center, to AI center and everything in between to really be a true architect and advisor in all of these aspects. And now, with AI coming in, where do the tweaks happen and how can we add on to improve? And a lot of what I do is also talk about making sure you're preempting what's coming, so that if you're looking at an edge device, where could we use it? Next, let's make sure we have space for it. And you're now cost savings in the long run because you're not doing stacking as much as you've bought something that can do expansion right, interesting, actually beyond interesting, fascinating and um.
Speaker 2:Like I've said a couple times already in this conversation, I wish we had all afternoon to talk about this because it's it's just, it's huge, I think you know, just to put a fine point on this um again, going back to nrf and how the evolution of ai, the ai narrative has changed um ahead of of the ai reality. You know I could do um, but when you think about what agentic ai could become for us, which I've only seen a couple of real world, you know genuinely useful examples of it. So far, most of it's kind of just test cases, and look what I can make it do. Most of it's kind of just test cases and look what I can make it do. But agentic AI could be the giant inflection point where this starts to become. You know itvo's view or matt bertucci's view on agentic ai and what it's going to be doing for us in one year ah, I mean it's.
Speaker 1:It's hard because it's it's all about our partnerships too. You know who we work with the, the intel's of the world out there that are allowing us to conceptualize and build a hardware based on all this, because at the end of the day, it is it, it has to run on something. Um, but it's a year from now. It's really having these ISVs that are very good at what they do, being able to open up and talk to the system that can then transition the visuals and the verbal coming in into a true response, right, less can't almost more natural. That's where I see it becoming, uh, it's more communicative, um, being able to simplify the carrying of multi-language skills within that one structure as well. Instead of now you kind of have to hit the english button or spanish button or french button, instead you come up and you just ask it in your, the natural language do you speak this language? Easy to say yes or no to that right. Once it says yes, go ahead. Pregame those are the things I see really happening and that's going to really, you know, make a shift in qr, qsr space, um, you know, with those types of things, uh, the, the help desk space, even the returns, being able to come in and just have it respond. Naturally that's going to make a big difference and that's where it's going to take us in the next year.
Speaker 1:I see, um, we have the great hardware, we've got the development in place for the, the structures for to run on.
Speaker 1:They look cool, they're, you know, becoming a little more sleek, but at the end of the day it's about what's running it and how it connects the room.
Speaker 1:And that's where retailers now are really kind of taking a step back from where they were two years ago or three, where everything was pushed to the cloud. It's going let me pull this piece back because it needs to be on frown right, or else I'm not going to get that to. It's going to get hiccups, it's going to jitter a little bit or have a longer pause than what is natural. You know, half a second is okay, but when he's getting beyond that, then it's becoming a computer, yeah, and and that's where I think it's, if the shift is going to pull away from jobs and pull away from how stores run and why, it's a social interaction kind of driver for people, you want to be there and have that. But it's about augmenting what's being done for multiple reasons and being able to, as a store, run efficiently and as at lower cost, because your margins are always going to be razor thin brilliant.
Speaker 2:Well, um, I'm going to do us both a favor, even though I don't want to um and just say thanks for all this. Uh, the the conversation could, and I think should, continue. I'm going to try hard to get on our schedule.
Speaker 1:You know where to find me. It's about getting the time to nail it down. It's a challenge.
Speaker 2:We worked hard at this one, but it was worth it. So, Matt, thanks a million. I really appreciate this.
Speaker 1:Oh, my pleasure. Thank you for the opportunity.
