How EPG Is Using AI for Warehouse Document Processing

Blythe Brumleve-Milligan

Jul 2, 2026

How EPG Is Using AI for Warehouse Document Processing
Episode Transcript

In this episode of Everything is Logistics, Blythe talks with Jett Chitanand, President at EPG Americas, about how AI is being used inside warehouses and distribution centers.

EPG is a supply chain execution platform covering warehouse management, transportation management, contract and billing, and yard management. Their AI environment, Aura, sits on top of execution systems to help teams solve problems inside the four walls of a warehouse.

They cover:

  • How EPG is using AI across warehouse and supply chain execution
  • Why intelligent document processing matters during goods receiving
  • How AI can read delivery notes, even when documents are messy or incomplete
  • Where the thirteen minutes saved per delivery comes from
  • How intelligent video analytics can help detect safety issues, congestion, and warehouse delays
  • Why AI should work with existing WMS and TMS systems instead of replacing them
  • Why clean data and a phased rollout matter before scaling AI projects

This conversation is part of the CargoRex AI Use Cases in Logistics guide, featuring real examples of how logistics companies are using AI across freight, warehousing, procurement, visibility, and operations.

Links & resources

Related episodes

SPI Logistics

Presented by SPI Logistics

SPI Logistics has been a Day 1 supporter of this podcast which is why we’re proud to promote them in every episode. During that time, we’ve gotten to know the team and their agents to confidently say they are the best home for freight agents in North America for 40 years and counting. Listen to past episodes to hear why.

AI LogTech Podcasts Warehousing
Show Transcript

See full episode transcriptTranscript is autogenerated by AI

SPEAKER_00 0:42

AI has been overhyped when it's positioned as a replacement of systems. That's not what we're doing. We're not ripping out your WMS or your execution stack and replacing it as AI. So once you start thinking about that, then you're going to be like, okay, well, this is going to replace everything. No, it's not. What it's doing is it's helping you rather make intelligent decisions and leveraging the visibility that you might have actually achieved through deploying these systems and then making you giving you that competitive edge to be able to be ahead of your competition and also not only understand the visibility of the supply chain, but also execute and be future.

SPEAKER_02 1:24

Welcome into another edition of the CargoRec series where we are talking about AI use cases in logistics. And today we're talking with Jet Chitanin. He is the president at APG America's, and we are going to be talking about their tools around the new era of AI with their Aura program and where intelligent document processing is part of more of a broader level of products that they are solving with AI. And so, Jet, welcome into the show.

SPEAKER_00 1:53

Thanks for having me, Bright. Looking forward to this conversation.

SPEAKER_02 1:56

Now give us sort of a high-level view of we mentioned that the broad spectrum of products that EPG is targeting. So give us that high-level view of who EPG is, your ICPs, your target customers. Who are you talking to and why?

SPEAKER_00 2:12

So EPG is a at our core, we are a supply chain execution platform. So everything from warehouse management software, transportation management software, contract and billing, which is very important in the 3PL world, yard management as well. And so this year, earlier this year, we launched our AI environment, which sits on top of our execution layer, on top of our platform, or any platform for that matter. And what it does is it's focused on solving challenges inside the floor walls of distribution centers and warehouses and leveraging AI to be able to make really smart decisions using AgenTic AI specifically. And that's gotten us a lot of success, especially having also won uh the best product award at Logimat, which is a large European um show that can be compared with Modex or ProMat.

SPEAKER_02 3:05

And so as you are developing these new products in sort of this new tech renaissance that we kind of find ourselves in working in in logistics and supply chain, how are you choosing to develop the different products and how did they sort of come to fruition? Was this something your customers were asking for, or were you seeing it in kind of the data that you wanted to optimize for them? How are you approaching your customers with these different product sets?

SPEAKER_00 3:34

So it's all about problem solving, right? That's a great question. It's all about problem solve solving. Uh, you know, we observe there are certain areas in the warehouse and even beyond, you know, because some of our applications go beyond uh what happens in four walls. But uh it's all about you know making sure that we understand what the challenges are and can we leverage AI using this? Rather, can we leverage AI to solve these problems? And that's how some of these came into fruition. I'm happy to go into more details regarding use cases, but we saw, just to kind of give you an overview, we saw uh challenges while receiving product. And that is one area where consistently we got feedback about how cumbersome is the process to unload everything, to get the delivery notes, to get all that information for you to then be able to really start doing your work. And then so on and so on, we uh gradually started to uncover more and more challenges that we could solve using AI. So we transform actually our company uh internally as well as uh you know developing this application and leveraging AI because you know, if you don't adapt, then you're left behind. So that's the approach that we're taking.

SPEAKER_02 4:39

So let's let's talk about some of those different use cases because we got a couple of your your case studies that were sent over, and but we're the the IDP product. So tell us a little bit about that product, and because I have some some data that shows that 13 minutes are saved per delivery across the goods reception process and at a 10 dock facility running 70 trucks a day that adds up to 15 hours is saved daily. Walk me through where those 13 minutes actually come from.

SPEAKER_00 5:08

Sure, absolutely. So to sort of center this uh this conversation, IDP, which is our intelligent document processing, sits inside a broader uh array of use cases, which is called as uh the EPG Aura Observer. So observer, it's like an eye. So you you you can an I can read a document, and I can look at a video, and I can look at a podcast, right? And also hear it, of course. So so that's the idea behind it. We've also, you know, um partnered with Nvidia on on some of the camera technology that we're using in in other use cases. I should say the intelligent video analysis, which is part of our uh observer. Here, the challenge was a lot of organizations were using OCR to detect documents, to read those documents. However, you know, when things are inside the warehouse, they're not always perfect. Uh as you know, you get all these delivery notes, they're crumbled up, there's stuff written on it, so on and so forth. So it becomes extremely hard for people to parse out that information, understand what that is, and put that into um, you know, uh enter that information somehow to capture it and enter it into your WMS. So what we did is um, you know, uh using AI uh technology, we were now we're able to understand and recognize patterns. So it's it's very simple. What you do is with a camera, you take a picture, the AI uses and it contextualizes the picture that you've taken, understands what if there's information missing, if there's information that's blurry, so on and so forth. Uh, and and really recognizes those patterns and builds up that information so that you get everything as it should be, and then you then uh take that information and upload it directly to your WMS. So all that process uh happens pretty instantaneously. Where you see people uh save uh save on time is you you mentioned the goods receipt process. So that can be broken down into five, let's call it major categories check-in and unloading. So this is when your trailers come in, you're checking in, you're unloading the trailer, so on and so forth. Then you perform your initial inspection and identification of what product it is. You've got your goods receipt recording, that's when you record all the information that you've received. Then, of course, you have quality control, and then you put away. So, in all of these steps, the major three steps or sub steps, I should say, where we can save time on are initial inspection and identification, just with that camera capturing all that information instantaneously. Similarly, with goods receipt recording as well, and then uh handling of uh you know different uh loading units and so on and so forth, and then quality control. So we we did extensive time studies also uh at a customer site as well, and then we that's how we came up with that 13-minute number. So then you can extrapolate that, and as you mentioned, that's absolutely right, it can potentially save up to um you know almost two shifts worth of work.

SPEAKER_02 8:04

Now, in a traditional shipment process, there could be all different kinds of modes that are used to complete the journey of getting that source, you know, to porch process completed. And as most of us know in this industry, there are you know all of these different information silos, there's acronyms, there's you know, language barriers. How do you sort of account for all of those different information silos and make sure that once that document is processed, that everybody can kind of understand it from the same lens?

SPEAKER_00 8:35

Excellent question. So I'm gonna go one step back and talk about how our AI platform or AI environment is developed. So at the core is something that we call as cognitive core. That is sort of the brain, uh and rather the heart of the AI system. There we have our own technology as we're using multitudes of LLMs to be able to um analyze all the information and use and leverage AI. And then we have something called a semantic brain. And what that does is it contextualizes information. So semantic brain is where you can upload all of your documents, your workflows, and so on and so forth. So let's say you were customer XYZ, and um you have some very specific processes in your warehouse, whether it's related to shipping, whether it's related to just whatever you're doing inside your facility, and you can actually upload all those into the AI environment. The AI environment uses our LLMs as well as that information to contextualize what it's seeing and looking in real time on that specific document. And that way it's able to understand and recognize this acronym in this context means this activity.

SPEAKER_02 9:43

And so as your well, with that, I it almost sounds like maybe like the that's a key part of the onboarding process when you bring a new client on that that's going to be utilizing your services. Because if you're doing, you know, uh if you have a WMS and a TMS, and then you have the these different document imaging processes, then I would imagine that that could create a situation where maybe you're surfacing things that things that are a problem or inefficiencies that were a problem that the customer didn't know about. So, how are how do you approach maybe um the pre-onboarding or the pre-boarding and then the onboarding uh during those customer conversations? What do those conversations look like?

SPEAKER_00 10:24

So it's always uh because applications can be so broad, and this is just one application, you know, the onboarding process can be, depending on what the scope is going to be for the entire project, it can go on, right? So it it's it can be an iterative process. Uh what we've tried to do is make you leveraging AI, and this is what I want you know, sort of the audience to also get away is uh get from this, rather, is you don't have to have your instructions in a specific format. You don't have to have, you can just say, just like give you an example, right? I mean, if you go on um and and any sort of then an AI platform, and then you say, I need a ticket, flight ticket to whatever it is. I live in Raleigh. So Raleigh, uh, and then uh just find me a flight ticket for May 1st, and these are your parameters, do it. So you you don't tell them what to do inside, it's gonna then figure it out uh based on the models, right? We're using a similar approach here where you have the beginning state, end state, that's how you solve problems. However, to your point, that can and will uncover some inefficiencies in this process. Where again, if you if you relate it to a um to a traditional AI that you're using every day, it's gonna say, okay, fine, I finished this action. But now do you want I've seen I noticed that there's something additional. Do you also want me to look at that and and and give you a recommendation? So that's sort of the way that it's gonna work.

SPEAKER_02 11:50

And so it's more of the exception management for uh a lot of these different roles where they don't know what they don't know, especially during the onboarding process, but there could be some opportunities where you know that 13 minutes that we cited earlier could lead into more efficiency saved across an entire, you know, sort of shipment flow. Am I understanding that correct correctly?

SPEAKER_00 12:12

Yes, that's right. And and again, just sort of reiterate that's one use case. We have others as well, but that's the the one where we can absolutely pinpoint and say this is what we've observed in real time. We've done time studies and saved um on cumulative time saved throughout a shift.

SPEAKER_02 12:28

And so uh one of those other use cases that that were brought up is it is IVA on the operational impact. And so the Aura observer, which is one of your products, and then the video partnership uh got a lot of attention, as you mentioned, at Logimat. Without some of those hard numbers, how does IVA, first of all, what does I guess sort of IVAs stand for? And then what is the big problem that it's solving?

SPEAKER_00 12:52

Okay, so IVA stands for intelligent video analytics. So it's as simple as you know having our uh AI system be deployed and and you have any cameras you can have. Like we don't necessarily recommend certain type of camera, but it's basically tied into the camera system. Any camera system that you have. This acts like a supply chain manager for the most part, or help that a supply chain manager can can really use. Imagine a warehouse and uh using cameras, it's constantly detecting or and understanding and contextualizing what's happening. So you're not telling it that this is a person, this is a yellow vest, and so on and so forth. It already knows this because of the LLMs uh that I just talked about and and and other proprietary uh technology. So what it does is let's say that you have um an accident in a specific aisle, or you have a product that is spilled, or uh, or you have obstacles, or if you have a pallet sticking out where it shouldn't be, so on and so forth, it's already going to look at that and it can give you an alert. It can give you an alert to the supply chain manager and say, this is what's happening here. How do you either want to handle it, or if there's a specific action that you need to trigger if I observe this, then go ahead and trigger that action uh by itself. So it's it can be used for uh accident detection, uh safety protocols, whether those are being followed or not, uh observing pallets, like if a pallet is standing there um for uh way too long uh and it needs to be uh put into uh onto a trailer, it will give you an alert and you can set all kinds of alerts using that. Uh overcrowding areas, it can generate heat maps. You know, there's you know, pretty much, I don't want to make it sound like a hyperbole, but it's sky still in it.

SPEAKER_02 14:42

And so as the I would imagine that manufacturers, shippers are the the target market that that EPG has has been working with and and going after for you know these different types of use cases. Um I I am curious about the uh sort of formal onboarding process. Do these do these additional products work best with your existing customers, or are new customers coming to you and saying, we have this problem, we don't know how to solve it, and we think you might be able to solve it? Tell me a little bit about those different demographics.

SPEAKER_00 15:13

Yes, absolutely. So um it is it can be both. And and the reason why we have seen so much success early on, and we continue to see a lot of interest and a lot of practical um solving practical applications is because we've kept sort of the end goal in mind is to solve these problems, and we've tried to make it so that it's independent of our own software, which is WMS or TMS, so on and so forth. Of course, we'd prefer it to be on our own platform, makes things easier, especially at Modex. We had a lot of uh interest there uh from companies, you know, wanting to learn more and really come coming to us with uh with problems and challenges, uh, and some of which we didn't think about, uh, and but we could certainly use those and apply those, apply our AI environment to those, to solving those use cases. So I would say it's a it's a bit of both, and we've intentionally kept it agnostic so that you can pretty much use it with any existing software, and you're not beholden to using our platform, uh, but certainly you'll you'll get more benefits from it.

SPEAKER_02 16:13

Oh, that that's interesting. So it it's integrating into what maybe a user is already have already invested in, and so it's an additional intelligence layer on on top of that.

SPEAKER_00 16:23

Yep, absolutely.

SPEAKER_02 16:25

So you've done the the pre-qualification, you you you figured out, oh, this is gonna be, and I'm talking from a customer point of view, this is gonna be a good fit for us. What should I do on my end of things to make sure that I'm prepared from a data lens, from a tech lens to make sure that I'm gonna be able to hit the ground running if I choose to engage with you?

SPEAKER_00 16:46

Yeah, so we have a very detailed and specific process if we go through and walk through the customer. What I want folks to understand is uh I remember somebody mentioning that adding in a new ERP uh is like getting a root canal. It's it's painful, and but you but you know that you have to get it done. Right? That's not the case here. Uh, just like with, you know, it may not be as simple as subscribing to JATGPT and you start using it, but it is closer to that than it is to uh deploying a new ERP or WMS. So uh we have a detailed list of documents and checklists through what through what's required, what's needed, and it's it's very use case specific. Uh that will walk you through, and it's a very uh it's a detailed and thorough process of uh us looking at the environment and then uh assessing where we can add value if we can, and then moving forward with uh with that. So it's it's a very uh detailed specification that we walk through with customers and uh understanding. Because if you don't understand, if you can't see it through their lens, then we can't solve their problem.

SPEAKER_02 17:47

It's as you know, more and more teams are uh adopting these tools, there is you know a segment of the population that is incredibly fearful about it using these tools. And oh, is you know, if I start to use this tool, is it going to replace my workload? Based on your experience and the teams that you've worked with, how are you handling some of the pushback that that comes with that fear?

SPEAKER_00 18:12

So that is, I would say that's the um, you know, boiling the ocean uh category, right? Like that's not what we're doing. We're not boiling the ocean, we're not trying to figure everything out. And and that also comes from some of the hype from AI, right? It's like people saying that AI is gonna solve all problems. So the bottom line is AI has been overhyped when it's positioned as a replacement of systems. That's not what we're doing. You still what would need it's not we're not ripping out your WMS or your execution stack and replacing it as AI. So once you start thinking about that, then you're gonna be like, okay, well, this is gonna replace everything. No, it's not. What it's doing, it's it's making you, it's making helping you rather make intelligent decisions and leveraging the visibility that you might have actually achieved through deploying these systems, which in itself can be a big feat, and then uh making you giving you that competitive edge to be able to be ahead of your competition and also not only understand the visibility of your supply chain, but also execute and be future-proof.

SPEAKER_02 20:10

And so as you're you're, I guess, building these out or building these systems out with different teams, uh, I'm curious as the, you know, maybe some of the executives have bought in, but it's down to you know, that sort of in the trenches employee or that department that has to manage the actual change management and the adoption and making sure that that people are using these tools. And your experience, is it really sort of one person like the AI ops lead, or you know, someone internally that's you know put in charge of like being the champion of you know the different adoption tools of AI? Is there more of a success rate that you've seen with different roles or you know, departments, or evolutions of how you know the modern manufacturers are moving into the in the modern age using AI?

SPEAKER_00 21:01

So I wouldn't say that it there's necessarily a group of people that that we've seen more interest in. It really depends on the company's DNA and whether you know they're a little more forward-looking as opposed to whether they're conservative specifically. Uh, and and the reason why I say that is because our use cases are so varied. So now I talked about the four walls, uh, solving problems in the four walls. That's primarily what we focus on. But we also have something called as an orchestrator, uh, which can understand if there's a delivery coming in that's late, then it can manage all the workflows inside the four walls based on the late arrival of that delivery to reprioritize all the work so that you can get the rest of it out as soon as possible and as efficiently as possible. So I would say operation supply chain and IT all have to sort of align uh to be able to make the decision. But I don't necessarily see a specific category or specific subfunction within a company that that has been more excited or or more um willing to move forward with it as opposed to how the company is structured and and whether they're more of a forward-looking company or or a conservative company.

SPEAKER_02 22:06

Yeah, because there's definitely a little uh, you know, with some conservative companies, there's a you know, a hesitancy almost to adopt these tools and how they, you know, they don't want to, what I've done for you know 10 years has worked for me. I don't want to change, you know, that that kind of mindset, which maybe has to be massaged a little bit. And once they start seeing, you know, sort of the eye-opening moments that these tools can provide, then it leads to to greater adoption. And I and I'm curious if you've seen that maybe with with your own customer base, how they, you know, maybe they'll get started with document processing and then now they want to add on another layer. Are you seeing more of a an all-in-one approach where someone wants to jump all in and rework everything, or maybe like a phased approach?

SPEAKER_00 22:50

Uh phased approach is is what I've seen primarily. And it makes sense. Crawl, walk, run. That approach makes the most sense. And that's what we've seen um more and more. But yes, we have we've had some conversations with customers where we worked with them for a while and and you know, they know uh our capabilities and and they have trust in what we can bring to the table. Those are uh are proceeding with having, even though even though it's phase, they want to go go sort of let's call it a big bang approach, more of a big bang approach. And and uh we've seen some success there as well. What I will say is yes, there's going to be uh a lot of uh trepidation. There's going to be a lot of customers thinking that, you know. What should I like? Should I really do this? But the bottom line is you gotta start somewhere. Start small because think about it, Blythe, just like twenty, thirty years ago, right? We didn't have cell thirty years ago, we didn't have cell phones or we just started getting them. Now we're able to text. Now we're able to have Wi-Fi. Now we're video, now it's AI. So it's constant change, right? So if you don't change at the appropriate rate, then you're gonna be left behind and your competitors are gonna go ahead. It doesn't mean that you have to, you know, go in and and flip everything upside down, but you gotta start somewhere.

SPEAKER_02 24:03

Absolutely. And maybe uh out outside of the use cases we we've already mentioned, are there any other maybe moments that you can pitch to a customer that can get them to start you know crawling before they walk or before they run?

SPEAKER_00 24:17

Uh yeah, no, absolutely. I mean the the the key is to um understand how clean is your data and and and really understand that um because you know if you have naturally if you have garbage data, garbage is gonna be coming out, right? So that is where I would say the customers or prospects or whoever is exploring AI, whether it's our solution or somebody else's, which I don't think specifically I I haven't encountered anyone who's solving problems in this realm the way that we are, but um regardless, you know, they should make sure that their data is uh is clean, the data is accurate, and and ultimately that is gonna then help them to start with a use case, start with saying where is where which outcome is going to yield me the best possible result inside the facility? Uh what's gonna because ultimately it all boils down to ROI. And and if that's the case, then you start there and if you see it working, then you expand, and so on and so forth. But you gotta you gotta start somewhere.

SPEAKER_02 25:20

How do you know if your data is dirty or not?

SPEAKER_00 25:23

Um that's that's a good question. I mean, uh you will if you have, you know, it's basically if you have a WMS system or or a system that's been that's a legacy system that you've been using, uh a lot of the times, depending on the size of the organization, if it's too small of an organization, then they know in general they don't have a good system of managing all of their data. Uh but from a mid for a mid-sized organization, you know, it also through their um through their operations, it becomes somewhat evident that they don't have they have disparate systems, disconnected systems. If they can't talk to one another, if you don't have end-to-end visibility, that is also a sign of you not having the right type of either data visibility uh or uh or data cleanliness, if you want to call it that, to be able to execute something like this.

SPEAKER_02 26:14

And so, you know, f a final couple questions here. Uh anything that you feel is important to mention that we haven't already talked about?

SPEAKER_00 26:23

Yeah, no, I mean, uh as I said, you know, visibility used to be the big thing uh for supply chains, right end-to-end visibility. We need to understand where, why, how, so on and so forth. And that's more and more uh is becoming more and more table stakes to be able to get to the next phase as as opposed to the end goal, which it should be, uh which it was perceived to be a number of years ago. So uh what I will say is uh again, going back to the message of you know, you you you gotta start somewhere in terms of innovation, even if it's small, even if it's utilizing a specific use case, so on and so forth. See if it works, but you have to be able to, you know, take that leap uh and start small and and don't try to again, as I said, boil the ocean and measure that result. And if it leads to a positive ROI, then you sort of have the indication of where you need to go.

SPEAKER_02 27:10

All right, perfect. Well, well, well, Jet, I think that's a a great place to end the conversation. Where can I folks or where can I send folks to connect with you, connect with EPG?

SPEAKER_00 27:21

Sure, yeah. So uh, you know, of course, uh visit our website, www.epg.com, uh, echo papagolfepg.com. And you can find me on LinkedIn. It's JetJ-E-T-T, C-H-I-T-A-N-A-N-D. And um I I checked the other the other day, and there's only one result that you can find. So it's uh that you know that's uh perks of having an unusual last name, I guess, but that's where you can find me on LinkedIn. So please feel free to connect and you know, happy to engage in conversations and answer questions, uh, whether it's related to EPG or just around AI or uh or anything for that matter that that pertains to the industry.

SPEAKER_02 27:59

Perfect. Well, well, thank you so much. And I'll I will find that that singular link for for LinkedIn and make sure I put it in the show notes just to make it that much easier for folks. Uh, but but this was really interesting conversation, so thank you, Jet.

SPEAKER_00 28:11

Thanks for having me, Fly. Thank you. Absolutely.

SPEAKER_01 28:17

Thanks for tuning in to another episode of Everything Is Logistics where we talk all things supply chain for the thinkers in freight. If you like this episode, there's plenty more where that came from. Be sure to follow or subscribe on your favorite podcast app so you never miss a conversation. The show is also available in video format over on YouTube just by searching Everything Is Logistics. And if you're working in freight logistics or supply chain marketing, check out my company Digital Dispatch. We help you build smarter websites and marketing systems that actually drive results, not just vanity metrics. Additionally, if you're trying to find the right freight tech tools or partners without getting buried in buzzwords, head on over to Caggorex.io where we're building the largest database of logistics services and solutions. All the links you need are in the show notes. I'll catch you in the next episode and go dive.

About the Author

Blythe Brumleve-Milligan
Blythe Brumleve-Milligan
Creative entrepreneur in logistics. Founder of Digital Dispatch, CargoRex, and host of Everything is Logistics. Co-Founder at Jax Podcasters Unite. Previously: Board member of Transportation Marketing and Sales Association, Freightwaves on-air personality. Annoying Jaguars fan.

To read more about Blythe, check out her full bio here.