AI Workflows Revolutionizing Wholesale Distribution

‍

Webinar Transcription

The following transcript is an excerpt from a webinar hosted by The User Group from January 2026 called AI Workflows Revolutionizing Wholesale Distribution. As an auto-generated transcript, it may contain spelling or formatting errors. Watch the webinar to learn more and access the full transcript, and follow The User Group for all things Infor.

‍

Webinar Host: It is my pleasure to introduce you to Aung Latt, Head of sales for Canals AI, as he presents a new era of AI workflows revolutionizing wholesale distribution. Take it away, Aung. 

Aung Latt: Thank you very much there, Ashley. And I'm gonna share my screen, and let's go ahead and get going. So, a new era, as Ashley said, it's an interesting title, a fairly big statement to make, to say “revolutionize”, . But I hope you find it's applicable after today.

So, Canals is an AI first company that uses AI to help you automate things that used to not really be possible to automate before, maybe at all, actually without any significant setup work. And I wanna show you how that applies today.

Let's take a step back. Why AI? Why are we talking about AI? Why you can't avoid the topic.

So what's all the fuss about? Actually, who cares? And so, if you think about who cares, well, there's, here's a sample of headlines. Technologists care. Nation states care. Major business institutions care. And,  you're all here. So obviously, I hope you care about AI, but just to take a step back: his slide here, I started with it, this was way back in 2024. We saw the buzz around AI back then, Canals did, and we started to see all that people were talking about all this stuff. Now, fast forward a little bit to 2025, and then things started getting real for wholesale distribution. Here's a couple pictures of articles on the web about people you should recognize, Amazon Home Depot three, . And they're talking about applications, AI applications they're building, in late 2025.

Further along the lines, if you don't follow a group called Distribution Strategy Group, I think it'd be interesting to go ahead and suggest you follow them. Brian Hopkinsville works there, and in this article he published late last year, he predicted that by next year, which actually means this year, the performance gap will be insurmountable in terms of the competitive gap that AI could provide to you. 

So, you could see a couple statements on this slide. Three to five percent of the total revenue and labor cost savings, the logistics savings. So, AI is a very meaningful thing when it comes to distribution. So now that we've established the “why” of AI, let's take a little further, like, what is AI?

So if you take it back from a clinical or educational standpoint, one of the pioneers of AI, Marvin Minsky, said it's a field of study that brings together those other fields of study to basically mimic artificially through machines what hans can do with their intelligence. So, from a purely educational standpoint, that's what it is. But what does that mean for distributors?

I think the easiest way to do that is to say, you know what? Let's take something, a case study of something that we're all familiar with, and then see what that means for that. So, let's take a case study of order entry and go from there. So today, let's imagine for this case study that your mission today is to automate order entry. Let's imagine that you're an IT team or an IT person at your company, and your CEO says to you, Hey, listen, you need to automate order entry. My reps are complaining, they spent too much time doing manual order entry. So you need to help out. 

Actually, it's not so far-fetched. You're probably somebody who may be in this scenario as it is anyway. So I'm gonna make one little key qualifier to this. I'm gonna say with software, and that's significant. So in this presentation, I'm gonna try to distinguish between software and AI. Software's what you've been using for decades. AI is what you've only started to use, if at all. So let's go again, take a look.

So easy case, we're an IT team, we're gonna automate order entry. I have these customers that send me stuff. So here's the easy case. You get great consistent information from a specific customer. They send you an email, it's always from the same sender, it's got a nice CSV attachment. And column one is the item or part number. And column two is the quantity. Nice clean information. So how do we go about doing that? Well, if we're using software, software is rules - if this, then that. And all the software that you've been using to date is really about that. Hardcoded logic to account for conditions of stuff. If you've used Excel and you've done conditional statements in it, if you're a developer, you've done conditional statements. If this, then that. And then the more rules that you compile on top of each other, the more you handle all the different exceptions, the more complicated scenarios that you can handle. So software can get really powerful and you've seen it to date, so far, and it makes it easier to do this. That's what you wanna do. So let's go ahead and apply that to our easy case.

So we're gonna put some rules in place. So again, same sender. We're gonna automate everything. Same sender. If the email is from a sender, then, oh, sorry, and has an Excel attachment, then create a quote. Then for each line in the row of the spreadsheet,  set the part number to column one, set the quantity to column two. Super doable with software. You can look at the columns and stuff to be able to convert that information. Ta-da, success. We're done. Thanks for attending. I'm gonna give you your hour, your 40 minutes back, and I'm wrong. It's not gonna revolutionize anything about what we're doing. That would be the wrong way to look at it.

So, two and a half inch galvanized rigid conduit. There are actually five ways to say two and a half. Two and a half with a dash there, two space, one half, two with a different thing and a smaller one. Two,  all the different ways to say it. There's five ways to say inch, a double quote, ININ, period, INCH. And,  five more ways to say GRC galvanized rigid conduit with periods with abbreviations. So when you're starting to do that, if you really look at it, there's like over 150 combinations that are, how to just say that one term. And that's before typos and before ignoring like curve balls where you are,  putting in certain, , abbreviations, they're different, or you're writing it differently and saying, Hey, I need 10 feet lengths of two and a half inch, GRC.

So if you think about going back to what we're trying to do is build rules. That's a whole lot of rules that we need to build, and then that's just for the one item. How about for the next couple items in there, right now, you have to do the same sort of thing for the next item, and the next item, and the next item like that. What if the data's not even formatted at all? What if they're sending it to you? And it's not in columns, it's just plain text in an email, like you see on the left hand side where like, Hey, I need the following price and availability on all, all these items. What if it's not at text at all? What if it's a handwritten note? Now, what do you do? We need to zoom out and say what?

That's why AI is significant versus software. We started with a really great case, an easy case, but the plain fact is that rules can't get you there. And now to be fair, there's some tools that can get you partly there. If you think, and this is gonna age me, but if you think about EDI. It was our first attempt to really put constructs around things. We said, Hey, you're gonna talk to a partner and we're gonna follow rules and everybody's gonna follow a certain set of ways to do things, so there are ways that you could parse out that information to make people understand that information. And then people said, all right, maybe you want to do some things where you have to set up some templates or you wanna map the data. And there was another iteration around a software that said, you don't need EDI, you can do it this way, and so there have been attempts at it, but in general, it all dealt with good cases. The good cases that you were doing, it didn't handle the vast majority of the noise of all the exceptions of all the different ways that your customers interact with you. And that's why AI has hype. That's what we're doing. We're enter AI. That's why I say “revolutionizing” is because it's really as, hopefully you'll see, gonna revolutionize it and make it transformational. So what is AI? AI is automation that doesn't really depend on rules. It's just, if you think about it, it's a different approach to tackling the same problem here, just if you put it in columns in terms of what it is.

So if you look at software, it's rules, AI is reasoning. Software is like, Hey, I'm gonna determine things with program logic. AI is training something, training it to act a certain way. It's probable. Software is rigid and finite, and AI is flexible and actually infinite. And the easiest way to look at it is, you might have heard of the term models or large language models when it comes to AI. So what is a model? If you think about it, it's just a large file that you filled with as many training examples of what you want the model to do. So you get the input and output that you want, as opposed to using rules to tell you how to get from the input to the output. The model, given enough training, predicts very accurately as to what the output should be. 

A great example is ChatGPT. Its task is to have a good conversation. And, and it feels like you're doing it. And the reason it feels that way is because it's been trained on so many models. It's also a little, the downside of it, when you look at it, when you've heard of things like hallucinations. Hallucinations come about. Because if it doesn't have enough information in a certain area, well then it might suggest that the output of what it's trying to say is, is what you're trying to do. So that being said, if you really look at why AI is great, it can address all of the cases as opposed to just looking at things, for the clean cut cases of what you're trying to do.

All right, so we'll take a stop there. I just went through some information about it from a mostly theoretical perspective. And before I get into showing you some information, I wanted to stop and see if there's any questions out there.

Actually, we do have a question. There's an anonymous question that says, what are the integrations to Burst?

Oh, a great question. So, Burst would be your BI side of things. If I interpret correctly. And so we can provide data out of Canals to be consumed by it. A lot of times when people ask me about the reporting side of Canals, I really say, when I show you the application, it's helping you do a certain part, but we're not trying to replace the ERP or all that data areas from there, because we're just one way to get the orders in there. So when you look at all the ways you're getting an order entry, which is what I'm gonna show you, or accounts payable or whatever the things may be, you're gonna get 'em from different sources as well. So we don't try to replace Burst or any of the other tools in terms of getting to the data. We can't contribute to it. So some people have asked and said, Hey, can you send some data to my data lake? And we can do that. So it's just a matter of how we would do it. Sure. 

And then there is a second anonymous question, and it's how do you manage security and privacy? 

So, one of the things about us, and that I'll show you when we get in, is we’re a web-based application. So we're out on the web, we're actually on AWS, so you don't have to have any sort of issues or concerns about what we're doing. We check that every year. So, in fact, if you are looking at different applications, you should see whether or not they're SOC II compliant because it's a fairly rigorous test to go through and a very good standard to say, this is what your company's actually being, we're showing you that we are, we've taken care of all the different things in terms of security and privacy and things like that. 

Awesome. I know how important that is too. And that's all the questions I have for right now, Aung.

Okay, great. Well then what we'll do is we'll move on and kind of get into what AI is for distributors.

To show you a little bit about Canals. So aside from the order entry mission,  we've really kind of gone over, what I would consider like a more academic treatment of what AI is. I'd kind of like to show you the real world example of what that would look like. So I am going to switch out and go to my application.

So hold on one second, share my screen Al. So you should see what looks like an inbox, and that is Canals. So Canals is a web-based application. You would log into it, like you would log into many of the other things that you do, via the web. And the purpose of it is really to take and provide - we like to call it human in the loop AI - What we're trying to do is we're trying to take in those inbound emails, the different ways that you get communication from your customers and interpret them and try to figure out what it is. So for the purpose of this demo on James or Jim Knight, I'm a customer service rep. I work at a distributor, and I come into Canals and I see a whole bunch of jobs here that are new that have come in, this one's already been submitted, but these are things that I need to address. This is my workflow, this is my queue that I need to work on. So what does Canals do? So I'm gonna click on one just to show you kind of what we do. I’ll click on that one.

And this is our job screen. And what you see here is on the left hand side, the source document for what we want to create on the right hand side. So this is just an unstructured email, kinda like I showed you an example. It's a conversation email that I'm having between me and my customer. So if you look at it, I'm Jim Knight, I'm the rep. And I reached out to my customer, James Anderson, and said, Hey, James, how are things going? Is business picking up? He responds to me, Hey, thanks for asking. Things are great, super busy. Speaking of, I need the following things.

So what does Canals do? Canals first of all says, all right, who is it that we're talking to. So we pull out and extract by asking it like, Hey, who was the person we're, we're talking to - we pull out James Anderson and Anderson Electric. We now compare that to our AI model that we'll have built out of the data we're pulling out of Infor to say, Hey, this belongs to, or we predict it belongs to Anderson Electric at this ship to, and this contact. 

We'll also try to find other pieces of information like a PO number or delivery instructions, or the appropriate, correct shipping address. We're trying to cut down the amount of work that your reps need to do in order to populate a quote or an order to send back to SX, CSD, whatever, application you're using in infor. So that's what we try to do. We try to build out the header.

But the magic of Canals is really the line items. This is where it really saves a lot of time. So when I look at it, right, James Anderson just listed a bunch of items. So the first item here, he says, Hey, I need two boxes of N95 disposable respirators. So Canal says, Hey, what is he asking for N95 disposable respirators? How many does he want? Two boxes. And then we go, all right, well, we believe that these are the top three matches. That would be that product coming out of your database. And again, we'll pull the data out of your database, build an AI model, and now we present it back to your customer service rep. And all they need to do their motion is simply to select the item that they believe is the correct one. So, if you think about that, right, what would it have taken them to enter the order manually?

They would've had to go through, find the customer, put all that information in the header, they would've then have to come in and type in N95 disposable respirator, or know what that means. And then the quantity. In this case, I'm just selecting an item 'cause we're suggesting the information for them, and they would continue to move down to the next item three of the ScotchBrite, eight by one by three to bearing wheels.

We break it out and say, all right, these are the three items, simply select the item and I move on. Now if I didn't know what that item was and I needed to find out more information, I could come in and actually click on the item and say, Hey, let me get more information. Typically, this same information we get here is the data that we're gonna pull out of Infor, and it's gonna be your descriptions, your product information, things like that. If you have additional PIM data somewhere, we're able to take that as well and put those attributes in. If you have imagery, we'll put all that in. We just work with you in integration to say what you want to provide. Now, the PIM data is great, but it's not necessary. And you can use this data alone here and get great matching on what you're doing. We also do things like this where we're able to go in and say, alright, real time, what is the pricing and availability? So I can click on that little magnifying icon. And if you had multiple locations,  it would pop up with that data. This is just a default screen. We'll customize it.

Watch the rest of the webinar (with transcript) on the User Group.