AI-Powered EDI: Simplifying Partner Onboarding and Mapping

Dan Magid

Use AI. Why, why is AI helpful in this process? And then the best, the best part I think is going to be Aaron is going to actually give you a demo of what he did with a customer. So in a real life situation where we were able to get a customer up and running, they got a notification from one of their major vendors that their, their AI process or the, their API process was going to change and they had to respond to that in very short order. And we were able on the phone with them to be able to get the new integration up and running in about an hour and a half. So basically everything, all the new mappings, understanding the documents, getting things into the database, testing it, deploying it, all that, we’re able to do that in just over an hour. So anyway, he’s going to show you a little bit about how he did that. So let’s talk a little bit about AI. So we’re going to start out and talk about, well, why is AI particularly well suited for the EDI process? What does an AI first strategy look like? And the idea of AI first is when you start the whole process, you think about, well, what part of this process can I use AI for? How can I use AI to make this faster and better? And then we’re going to get that real live use case demonstration and a little bit about monitoring your AI transactions, monitoring your EDI transactions. And as Maya said, if you have questions during the process, please feel free to, to post those questions. We’ll try to answer them as we go or if we don’t have time during the presentation, we’ll answer them at the end or we’ll send you an email with the answers. Okay, so why do we want to do EDI and adi? What are the things that make EDI hard? And probably the thing we hear most from our customers is that EDI is the non standard standard that they get different versions of the same EDI document from their different business partners or they have to send the EDI document in different ways for their different, their different business partners. So they need to be able to customize that. So how can AI help you with that since it has an understanding of the documents and it can actually read the documents that you’re getting and translate that and figure out where the tweaks are. So how can it help you? EDI is by its very nature it’s kind of a cryptic document. Originally EDI was developed I think 50 years ago or so, or maybe even 60 years ago. So it was in the very early days of computing when you had, when had very, very little space for identifying what things were. And so the labels in an EDI document can be very, very hard to decipher if you don’t know anything about edi. So it requires specialized knowledge in order to work with the EDI documents. And then doing data transformations, there may be a whole variety of things you have to do with the EDI data. So you may get a PO as an EDI document and then you need to turn it into a program call. So you need to take that data and turn it into parameters for a program, or you need to store it in the database, or you need to create a PDF document, or you have to send out a text message with that information. There are a lot of things you might have to do with the data. And so it’s really good to have an engine that understands how to do that and then something that monitors the traffic at very, very high speed and in real time to make sure that you’re not getting defective documents into your system or that you’re not generating defective documents and sending them out.

Aaron Magid

Right. And Dan, actually if I can jump in for a second, one of my favorite examples of the non standard documents, I was doing an integration recently and it became clear pretty quickly that the company we were talking to wasn’t actually, actually using edi, but they were kind of pretending to use edi. Basically what I mean by that is they were sending a996, an EDI996 document, which is a. Just really, just an acknowledgement, right? Just saying, yes, I got your document. That’s what that document actually means. But there’s a segment in the996 spec, it’s called K3, which says this is unstructured data. So I can just put whatever character data I want in this K3 segment. Okay, so what was this trading partner doing? They were saying, I’m going to send you a996, which is an acknowledgment, and then I’m going to use those K3s and I’m going to put my own proprietary data format in there and that’s the actual data I want to send you. So meaning what they were doing basically is they were using EDI as an envelope in a totally non standard way for something a document was never meant to do using, using an acknowledgment to send actual data and basically only using it because it had a hole in the spec that allowed them to do whatever they wanted with that, with that data, just like that. To me was one of the best illustrations of seeing a document that makes absolutely no sense within the context of edi. It’s just violates every standard. But that’s the kind of thing that we run into, right. You know, we see things where these documents get really weird and, and another thing actually is, to Dan’s point about we have to do complicated things with them. Another one that I was working on recently to show another, you know, kind of this broad landscape, it’s a fairly standard document. But what they needed to do is when they received this document, it was a 214. Actually, when they received this document, they needed to go turn around and make an API call. Right. So meaning it wasn’t taking data from A214 and then putting it in their database or calling a program or orchestrating a process. No, they were calling an API. Right. So we’re seeing more and more of these complex processes that are being orchestrated on the basis of edi. And we’re also seeing more and more people break the standard as they get as standards, as their needs evolve, and the standard doesn’t evolve with it. So starting to see this more and more as we do more integrations. I just wanted to, to bring those up because these are, these are very real things that I’m. That I’m dealing with on a, on a daily basis as I work on these kinds of integrations.

Dan Magid

Great, thanks, Eric. So then, so why would you want to use AI? How can AI help you with these problems? Well, one of the issues again that we run into is that because EDI requires special knowledge, there are just a few people or maybe one person in the organization who actually understands the EDI stuff. There’s the EDI person and they know everything about EDI and nobody else really understand. It’s a black box. So why not get that knowledge into an AI engine so that you don’t, you’re not completely dependent on that one person who knows how the EDI stuff works, but you have something that everybody can use. Anybody who you need to have work with EDI documents can use the engine that understands what’s happening in the EDI documents and how your partners are using the EDI stuff. So get that knowledge somewhere where it can be reused by other people so you’re not completely reliant and that your EDI person can actually take a vacation sometimes. And then the speed of change, that things are changing very fast. And we are seeing that business partners are starting to change their EDI documents. So there are things happening that where You’ve got to respond like the customer I was talking about earlier, where your business partner says, hey, we’ve changed this about the EDI document, by the way, we’re deploying it next week and you better be ready for it. And so you’ve got to be able to respond to changes that are being made or you may need to make changes to how you’re, you’re using the EDI document, things that, the fields that you’re using. So things are happening very, very quickly. You need to be able to respond very, very quickly or you want to onboard your, your partners more quickly. You want to be able to get that going very, very fast so you’re not losing revenue while you’re waiting for the onboarding process and the mapping process to happen. So you want to be able to respond very quickly to changes in the marketplace and changes in opportunity. And then you want to be able to be very agile with your integration. So as people change, as your partners change how they deal with the documents, you want to be able to respond as quickly as possible to what they need and then observability, you want to be able to see what’s happening inside the EDI documents. We’ve heard that as many as 15% of EDI documents are defective. So there’s something wrong, there’s data missing or data in the wrong format. And so you want to be able to see that and respond very, very quickly. Especially for documents you’re sending out, we have customers who say that their partners actually ding them, that they get fined for sending defective documents. So you want to make sure that you’re seeing that stuff as quickly as possible so that you can fix the problems before they really become a big headache. And then you want to be able to be flexible. You want to be able to move from edi documents to APIs, because we’re seeing some companies that are doing that. You need to be able to be but flexible in the kind of communications that you’re doing. And AI can help you with all of those problems. And again, you’ll see how, in, in what Aaron’s going to show you and what’s, what’s great about the, the EDI solution that we’re going to be showing you is it’s all built on the aerodyne Connect framework. And I just bring this up, you know, the purple box here is Erdani Connect, in that you can use Erdani Connect to generate these integrations. It understands edi. So we’ve built into, into our EDI module, we’ve built into an understanding of all the EDI documents we license from the X12 organization, all the specifications, so all that information is in there and it understands all the documents so that you can add documents and add new partners with different documents very, very quickly. But it also has all the built in security of aerodyne Connect, all the data transformation capabilities, the logging, the error handling and all the things that Erdonik Connect does just for any integration.

Dan Magid

And so we’re going to be focusing here, aerodyne Connect has all these different ways of connecting to systems, but we’re going to be focusing here today on these last two boxes, which is the file transfer capabilities, whether it’s FTP, SFTP or AS2, and then the actual EDI processing capabilities that it provides. So we’re going to be focusing on those connectors.

Dan Magid

So this kind of what, what the process looks like. We. You give, you give Erdani Connect the EDI documents that you want it to work with. It analyzes those documents. It’s an understanding what’s in them and where they’ve tweaked the standard where things are not implemented according to the standards. And then you can give it some database samples or some samples of the outputs of the target of where you want that data to go. And it will then generate the mapping code so it’ll go through and it’ll figure out what the mapping is so you don’t have to tell it that field XYZ is equal to field ABC in the EDI document. It will figure it out. It’ll look and see at the source and at the target and figure out where the data goes. And it’ll then generate the mapping code for that. And then it’ll also allow you to deploy those EDI integrations and to monitor them. So you can generate that mapping code without anybody who is really an expert in the two sides of the house. You don’t need somebody who is an expert in your database and somebody who’s an expert in the EDI documents in order to create that mapping code. It can figure it out on its own. So you don’t need those experts to do that. And it does it very, very fast. And it’s an interactive iterative process in that it’ll run that process. And if it gets something wrong, if it does the mapping wrong, because maybe you’ve got the same value in a database table field as in the EDI document, but they’re really not the same field. You can just go in there and tell it in English or Actually in any language. You can just tell it in natural language and say, hey, this field’s in the wrong place. It needs to be over here. And it will then regenerate and it will fix the mapping. And so when you get a new partner, instead of having to spend hours or days or weeks figuring out how all the data maps, you can get those things up and running very, very quickly. And again, Aaron’s going to show you that. And then it simplifies the management of the whole operational process of processing EDI transactions. So, as I said, it’s an iterative process. So you provide the samples to the system, it generates the mapping code, you look at the results and say, are the results correct? Then you give the AI your feedback on what’s right and what’s wrong, and then generate the code again, look at the results, and you go through that process a few times until it’s right. And then you test and deploy. That’s the process. Now I’m going to turn it over to Aaron and give him the opportunity to show you some stuff.

Aaron Magid

Yeah, well, thank you, Dan, for that. I appreciate it. Just focusing on that iterative process for a second, that really is the core of what we do, and it’s really the target for what we believe is the more effective strategy for edi. Right. Just to give you guys a real example before I go in and I show you my sample, my sample is based on some real examples, but I want to give you an actual real example first. So I was working on an integration and with one of our customers and what we did with them. This is what, this is how our process works. We had them gather what they wanted to happen in their system as a result of receiving a particular document. They grabbed the implementation guide from their trading partner. Right. Standard information, and they grabbed the sample documents that their, that their trading partner provided them. Right. So what we have there then is we have starting state, we have ending state, and we have a little bit of context. Right? Right. Starting state is, I’m going to get this document for my trading partner. Here’s a sample. Ending state is, this is what I want to happen in my database once I receive that. Right. And then the context is that implementation guide. That’s just additional information. We actually operate frequently without the implementation guide, but if it’s available, more information generally doesn’t hurt. What we did is we got on the phone, I had not seen the documents before. I got an idea, this one. I had not seen their documentation before we got on, we fed the information into the generator. The generator went in and analyzed all of their, all their documentation. And it spit out an initial mapping. We compiled that, hit a couple buttons, we deployed it, and we ran a test document through and we looked at the database output. And on that phone call, right? This is a couple minutes later, and most of the fields are right. I think there were like, you know, it was somewhere around 85, 90% correct or a couple fields that were wrong. So we gave it some feedback. We said no, actually. Okay, I see how you mix that up, because this, this wasn’t really clear. So here’s some additional information. We actually want this one to go here, and I don’t want to go there and, oh, that one’s a composite field. It’s got five different values in the same character string. I need you to break that up based on these rules and put it into these five columns in my database, gave it that information, it regenerated, we deployed it again, ran another test document. We did that process for a grand total of 78 minutes. And at the end of that, we went from having never seen the document at the beginning, absolutely no information, to a working EDI integration in production, sending documents that were being accepted by the trading partner, right? That’s the typical process that we’re going through, right? And that includes analyzing the documentation, doing the mappings, running the test documents, setting up the connectivity with the trading partner, you know, with their van or their sftp, and then making sure that we have the mappings, running tests, sending the documents through, giving it feedback, iterating, making changes, and then setting up all the monitoring and deploying everything to production. That entire process happened while we were on the phone, right? And then we met back a couple of days later, actually, and we did it again with another transaction, right? Same metrics, right? That’s how this process works. And the reason why we’re able to do that is because that iteration speed is so fast, right? I can give this documentation over to my AI assistant, and a minute or two later I’ve got an initial mapping, I can give it feedback, and a couple minutes later the mapping is ready, right? And it’s been deployed, it’s ready for me to use it.

Dan Magid

Right.

Aaron Magid

And what that means is I can actually go through the entire iterative process without getting off the phone. One of the big delays that we see in EDI projects is I’ve actually, I’ve dealt with this before, before we had these tools is. And, and with groups that actually were where. Where they’ll refuse to meet. So we Push pretty hard on that is you get on the initial call, they give you their documentation and then you generate your mappings and then you get ready to do the call and then you schedule a follow up where you’re going to do the test and then you find out that their EDI person is on vacation. Okay, so then you wait two weeks and then they come back and then you sk. Try to schedule the meeting again and you have a meeting, but someone else wasn’t able to attend because they’re on vacation. Oh shoot. Okay, so I guess we got to keep going, right? That kind of delay process I see happen over and over and over again. And if you get into the testing meeting, just to drive this home, if you get into the testing meeting and there’s a problem, right, the mapping’s wrong, which is happens most of the time, then you have to go back, you have to get off the call and go make tweaks to it, which means that now you’re going to schedule another call and then someone else is going to be on vacation when you try to have that call, right? That’s been my experience when we have to do that process, which is why we target this one call, get on the phone, get the documentation, generate iterate, test, deployment, right? Refine, deploy, refine, deploy, right? We do that over and over and over again until we actually like the integration and everything is ready to go. So that’s what I’m going to go through with you here.

Dan Magid

Aaron, just real quick, there’s a couple questions here. One is, does it handle mapping and translating the EDI document into a fixed length file? Great. Flash file.

Aaron Magid

Yeah, yeah, really good question. So yes, it’s totally fine. See, here’s the thing. The AI, this is one of the really powerful things about AI. It doesn’t actually care about your database. Meaning what Aerodynamic Connect does as an, as an AI first integration platform. What it does is it says, I’ve got data here in this system. Okay, let me understand how that system works, right? Maybe that’s an SFTP server, that’s got EDI documents on it, right? Let me understand how that works. Okay, now I’ve got an ERP system over here. Let me see how that works. Okay, I gotta put data into the database or I need to call these APIs or both, right? Or I need to submit data to a Kafka topic. Whatever I need to do. Now that I understand those two, I know the data that I need to be sending. I know how, how I need to start and how I end And I know how to talk to them. Let me generate the connectivity between them. Right. That’s what it’s doing. So in this particular case I was talking about, the requirements of the integration was that it needed to go into the database. That was just what this company wanted to do. They wanted to take in the document, put in the records into their IBMI database. That’s a very common thing. Okay. That’s what they were doing. And then from there they were, they had other programs that were going to process that, so that’s why that one was designed that way. But we’ve got other integrations out there where, for example, we take the EDI document and then make an API call, right? No database, no particular platforms involved, no dependency on any particular system. And, and those. Those processes work just as well. Right. So I see another question here from, from John this. Most targets are not a single database, but an ERP or a payable system like SAP. How does that work? It’s the same thing, right? What we’re going to do is we’re going to look at SAP and we’re going to give that ending state. In, in this case, the ending state to SAP would be an API called SAP, right? So instead of database records, we’re going to take an API call spec or API docs, right? And we’re going to feed that in and that’ll tell the system. And then in probably an additional context, we’ll say, just so you know, this is an SAP API, so you’re going to want to call, you know, look up SAP’s documentation, make sure that you know how to call that. Right? That might be some additional context that we put in there, right? And then what it’s going to do is it’s going to say, okay, EDI to SAP. Great, let’s generate it, right? Those, regardless of what the starting state and the ending state is, that process will be able to generate it. Because fundamentally what it’s doing is it’s saying, I’ve got data here, I’ve got data there, I need to map between them. Right? It’s not tied to EDI or APIs or SAP or databases. Right. And we did a webinar a while ago actually about EDI modernization and it follows from the same concept, right? Because it’s not tied to that. I can also say, just as a side point, I’ve got EDI to SAP right now. And then I can say, you know what, I don’t really want to do the EDI thing anymore. I’m going to replace that with a REST API. Now I got a REST API to SAP. Fine. Right. The tool’s not bound to what I’m doing. It’s smart enough to be able to jump between protocols and figure this out. So anyway, that’s is really good questions, but that’s one of the big differences with the AI first strategy. As opposed to pre built drag and drop tools. Right, Drag and drop tools. You can only drag and drop to something that the company that built the drag and drop tool has built a block for. Right. If they have support for databases, you can drag that block and it can do it. If they don’t have support for your database, tough. Right? Nothing you can do about it. The AI first strategy, we explain it and, and it will figure it out. Sometimes you’ll have to iterate, you know, sometimes it gets wrong first time, it needs a little bit of tweaking. You know, we give it some feedback, tell it what the errors are, and it’ll get it right second time or the third time, but. But it’ll get there regardless of what it is. John, great question. I am not actually an SAP expert, so I’d have to look at that and see what that is. But as a general rule, if there’s an interface to talk to it, you probably can do it. I’d want to look at that particular support from SAP and see what that is and how they want you to talk to it.

Dan Magid

Basically, IDOCs are intermediate. They’re intermediate docs. They’re basically a. They’re meant to be an exchange document, but they have their own formats in there, documented by SAP. Yeah.

Aaron Magid

So then my expectation of our systems based on what I’ve seen is we’d take a sample IDOC and we’d take SAP’s documentation on how you submit one and how you actually work with it, feed that into the generator and it would generate one. And then we’d run a test. SAP would come back and say, thumbs up, thumbs down. And if it’s thumbs down, we’d go see what’s wrong and give that back to the. Back to the generator and have it tweak it.

Dan Magid

And they’re meant to be asynchronous. Communications in Eridani Connect is designed to work with asynchronous communication.

Aaron Magid

Yeah, Great question. All right, this has been a lot of theory. Let’s do it. Let’s see what this looks like. And again, if anyone has any other questions, please feel free to throw them in. I am always. I think Daniel Dan will watch the Q and A Box while I’m, while I’m going through this. All right, one second and I will share my screen.

Aaron Magid

Okay, cool.

Dan Magid

All right.

Aaron Magid

All right. How’s that looking, Dan? Can you see my screen?

Dan Magid

Yes, I can see your screen.

Aaron Magid

All right, might as well wait for that. Okay, so let’s generate an integration. So this is actually what I went in and did on that meeting that I mentioned. This is actually how we do this. So what I’m going to do here is I’m going to process a document. Let me pull it up. That looks fine. Maximize for me. Okay, there we go. I’m going to process a document that, that looks like this. So this is a 204 that I’m processing. It’s coming from the trading partner EFGH2. It’s going to my ABCD. And, and I just want to, I just want to point out my address here at the 8 through 3 Mendocino Avenue in Berkeley, California, 94707. That’s, that’s, that’s some data that we’re just going to focus on here. There’s a lot of other data in here and focus on those. So what I want to be able to do is I want to be able to take this document and I want to be able to process it now. I’m going to use that database paradigm, EDI to database for this particular demo. That’s the most common structure that I’ve seen in EDI integration. So that’s why I use it. But again, just keep in mind, we’ve got plenty of integrations in production that are doing radically different things. It’s totally fine. I focus on this one because it’s the most common. So what I’m going to do is I’m going to say, okay, I need an EDI integration. Please give me an EDI integration. All right, so my workbench is going to come up and it’s going to say, okay, let’s get you some scaffolding for this new integration you’re going to build.

Dan Magid

Right.

Aaron Magid

Very basic information. So from here, what I’m going to pull is in my document. Got it up on my other monitor. E, F, G, H. Spell that right. E, F, G, H2 and it’s a 204 and it’s inbound, meaning I am going to be receiving this document. Okay, let’s generate that, Give that a second. That’s going to generate my base scaffolding for the integration. So that’s going to be just some starting some, just the starter integration here. What it’s going to give me is just the sort of very high level resources that I need in my system to be able to accept that document. Basically created a, a, an empty flow for that document. I can look at what it generates if I want. That’s all going to be in code. So I’m going to keep it off for now. We can look at it, look at it later if you’d like. What it’s created here again is an empty flow. So I could actually give it a 204 at this point from the FGH2. It will route it in here and it will do nothing with it because I haven’t told it what to do with it yet. So let’s tell it what to do. It’s asking me for a sample of my EDI document. Remember, starting state, ending state, that’s what we’re looking at. So here’s my starting state. It’s the CDI document. It knows I want to do database. So I’m going to come over here and what I’m going to do here is I’m going to give it samples of what my data looks like. So I’ve got a table here, DMO 204 and I’ve got a data sample. So there’s my first data sample. These are, this is a sample of the record that I want to go into my database. Okay. Add another sample. Because I actually need in this case a second data sync here. I need data to go to two places. I also have a DMO 204D. Of course it’s not gonna let me type 204D. There you go. Helps if I click on places before I try to type in them. All right. And once I’ve done that, I could put additional context and files in here. For example, trading partner implementation guides. I’m not going to do that, but I could. And in real production integrations we typically do. So I’m going to ask it to generate now. So what that’s going to do is a couple of things. It’s going to send out that EDI document that I gave it to our system. First thing it’s going to do is it’s going to compare that against the X12 spec. That’s very first thing. Right, because the very first thing we need to do is we need to understand where is this document standard and where does it depart from the standards. So that’s the first thing. Very, very first thing that it’s going to do. Once it knows how this document fits within the standards, it’s then going to say, okay, let me find the data in those fields and, and find those in my ending state and start generating mappings. So at this point I wasn’t timing that, but usually it’s about 90 seconds. I have a starting integration here and you can see here it’s created what it calls a DMO 204 Mapper and a DMO 204 D Mapper and it’s got some definitions in here that it’s going to use, got validators and things like that that will help me with my integration. Again, Erdani connects integrations are all generated in industry standard code. You do not actually have to look at that code if you don’t want to, but you are welcome to jump into it if you would like. Within AerodynyConnect EDI users, it’s about a 50, 50 split between the users that actually look at the generated mapping code and, and the ones that just let the AI do whatever it’s going to do. It’s really up to you. If you were to pull it up, what you would see is mappings that’ll look like this, where it’ll say, okay, here’s this field coming in here and this one going in there, and here’s how I’m mapping it to your database. And here’s where there are things I couldn’t find.

Dan Magid

Right.

Aaron Magid

It’s going to go through and it’s going to figure all that out for me. Okay? So the first thing I want to do is I’m going to say over here, I’m going to pull up my assistant and the first thing that I’m going to do is I’m going to say, okay, can you check my mapping? Give that a second. What that’s going to do is it’s going to go through, it’s going to do a second pass on the mapping. I like doing that because what I’ve found is, you know, anyone who’s used AI, right? You probably see that it gets things mostly right, right? You know, you work with AI. I didn’t give it very much information to work with. I basically just gave it the documents. I didn’t give it an implementation guide, I didn’t give it anything el. So we might find that there’s something in there that it might be that, you know, it, it may have filled in some information. Actually, it will have filled in a lot of information. There might be some things where it makes a lot of sense. There might be some things where it could have an issue with what it generated. So I’m Having my agent here go through and say, okay, let me, let me make sure that this is all good and that this all makes sense. What came out of that initial AI pass, right? So this is part of that iterative process, right? And this is something by the way, that we recommend. Regardless of what you’re using AI for in your systems, if you have AI systems, it is generally good to take your, to make sure that you’re going through a review and refinement process on, on that AI generated code.

Dan Magid

Right?

Aaron Magid

So that’s what’s happening in here where. And it’s actually one AI reviewing what another AI did. But that, that in, in our experience makes things a lot more effective and our processes do that internally. Whenever you have it generate code, it immediately goes into a review and error fix process to try to resolve any hallucinations that may have come out of the AI. Anything that it might have done improperly, it’s going to go in and it’s going to resolve and make sure that, that your system is functioning and you can actually see it doing here. If, for those of you who are already earned iConnect users, you may recognize some of this. It went in and it actually automatically ran a compilation and a test of the mapper. Right. So the agent is going in and it’s saying, okay, let me make sure that the entire process is working and that everything looks good. And it’s just doing that. Actually it decided it wants to verify it one more time. So I’ll give it a second to do that.

Dan Magid

Right. So Aaron, this is part of the automated process, right? It’s actually going through this multi pass process. So there’s an iterative process where there’s a user involved, where you say, well you didn’t get this mapping right? And you know, here’s what it should be. But there’s also the iterative process that the AI is doing itself of going in, generating the code, checking the code, regenerating the code, checking the code, fixing the errors, you know, going through that process. Which is the difference between doing this using Erdani Assist here like this, versus just going to an LLM and saying here, you know, send this in and generate something for me.

Aaron Magid

Right, exactly. The reason why we set up this sophisticated assistant is because we first time we did this we actually tried using you know, just sort of the off the shelf LLMs and you know, didn’t work so well. Right. What Erdani Assist is doing fundamentally is it actually understands edi. It’s got all those additional resources built in so that it actually understands how to do this effectively. So at this point, I have a mapping. Okay, my eight, my, my AI says that it’s done it. Let’s see what it does. So I’m going to reset my system, make sure that I actually get connected. All right, everything’s good. Start it up. And I’m going to jump over here and remember I said earlier, I’ve got my document right here queued up, ready to go. So I’m go over here and hit send. Send that out. Give it a second. Came back, says it worked. All right, let’s. That means that it completed successfully. Let’s go see what it did. I’m going to go look at my database. Got this DMO 204 and I’m going to say show me what’s in that database table. I’m just going to zoom in on that because that’s kind of tiny. But if you look at this, you’ll see. Remember I, I mentioned earlier, focus, pay attention to that. 8 through 3 Mendocino Avenue, Berkeley, CA 94707. Right. This data came from that EDI document. So these records are in here. It’s got my DMO 204 and it’s got also my DMO 204D. It’s got this record here for my. Also generated from my EDI document. Right. So this is a functioning integration. At this point, what I would do is I would actually go and deploy this, hit the button to actually ship it up, and then we’d run a real test document through it and go look at the outputs. Right. At this point, I have that first draft. Now, going back to that example that I was mentioning earlier, we did this, okay, we’re on the phone, we check the results. And it’s mostly right, there’s something missing though. And, and in this particular case, I want to call something in, in particular out, give you an example of something that, you know, kind of thing that we see. Let’s say that I made a mistake and you know, actually I don’t really need this street address. What I actually need is a set of geocoordinates, right? That’s, that’s the real thing that I need. Okay, so shoot. Silly me, I forgot what my requirements were. So. All right, that’s, you know, but, but we have something that’s working. So let’s go and let’s have it, tweak it. So what I’m going to do here is I’m going to say I’m going to get it working on it while we. And then I’ll explain what I’m doing. So I just gave my assistant a prompt, and what that says is the DMAD one, which is this guy. DMAD2DMCTY DM ST for state and DMZIP fields in the DMO 204 table currently receive the address for the stop. Instead of storing the street address in DMAD1, geocode the address and store the lat long in DMAD1 in the format latitude. Sorry, hold on. I just moved Latitude, Longitude. Right. Use the geocode.xyz API to get the latitude and longitude for the address, because I have an account with them and here’s my API key for that call. So that’s what I gave it.

Dan Magid

Right?

Aaron Magid

I gave it this prompt that says, I need you to update this so that it’s going to do this geocoding. Now what my assistant is doing is it’s going through and it’s figuring out, okay, how do I actually make this work within the context of your mapper? Right. So it’s aware of the data sample, it’s aware of the database, it’s aware of all the work that has been done previously on this mapper. And you can see it actually going in, in real time, making the changes to my mappings.

Dan Magid

Right.

Aaron Magid

That’s what I’m doing here. I’m having it actually make the changes to my mappings in real time. And again, this tool understands how to build an EDI integration. It’s been specifically configured to understand how to do edi, and it’s working based on the scaffolding that we created for it. That’s what allows it to actually be effective at doing these integr. So we’ll give it a minute. Right. And again to map this back onto that example I was mentioning. Right. This is what we did. We’re on the phone, we ran a test. One of the fields is wrong. Okay. We gave it some feedback to what we actually needed, then we chatted for a couple of minutes while it worked, and then once it was done, we actually ran it again. So it’s going through now. It’s updated the code. It’s checking it now, and we’ll see how well it goes. Okay. It’s compiling, it’s checking, it’s formatting, cleaning up after its work. Second,

Aaron Magid

I think actually at one point in that call, we both stepped out, just, you know, let it work for a little bit and did other things.

Dan Magid

Yeah, Aaron, that was like the time that I, I saw you, you had two Computers up and running, and you were on the phone on, on a call with a customer, and the, the, the AIs were just sitting there generating code.

Aaron Magid

Yep. Yeah, that’s something that I’ve started doing is all. One of the craziest things these AIs allow you to do is work on more than one project at once. So I’ll have multiple windows open, multiple projects, and I’ve got the assistant coding on it. What happens is I give one assistant instructions. I say, this field is wrong. Then it goes and it works on it for a couple of minutes. And while it’s working, I go and I talk to the other one because it just finished its task. And I say, okay, let’s run another test. Oh, that’s wrong. Got to fix that. I give it some feedback. It goes and works, and then the first one’s done. So I go and give it some feedback. I’m just bouncing back and forth between these projects, working at a speed that I wouldn’t have been able to imagine, you know, just a year ago, you know, without this technology.

Dan Magid

All right, Aaron, John just asked a question. How do you load a customer’s EDI guideline? PDF document, Word document, SEF file.

Aaron Magid

Oh, great. Any and all in that, in that generate step, I can upload files and I can also put text. So if you have a file, you know, text in my experience is the best. If you can extract it as text, the AIs understand that. The best. You know, sometimes PDFs, there are things like images embedded in there that, that kind of confuse them a little bit. They’ll still get it most of the time, but in my experience, text is the best format for them. They’re. They generally just do the best with it.

Aaron Magid

All right, let’s see how it did. Now, again, remember, this is an iterative process. I just generated this for the first time. So we’ll see how it does run the call. Okay. It’s taking longer. I think that’s a good sign, actually. Probably calling out to that endpoint. Okay, got a response from the API. Let’s see. Okay, it says it got a success. Let’s see what our database has now. Actually go to the right file here. Let’s run the query again. And at this point, I’ve got in my database now, you notice in my DMA D1, I got another record, and this time it’s the geocoordinates for that address. Right. That’s what it looks like. Right, That’s. That’s what it looks like to make this Kind of tweak to the integration, right? So again, when I did this or when I do this. But the specific example I was talking about, the way that that process works is it’s just this, right? I go in, I give it a tweak, I say, oh, shoot, we forgot a business rule. Happens all the time, right? Somebody didn’t say something, the documentation didn’t have it, right? There’s something that we need. And we go in, we give it the tweak, it changes it, we deploy it, and then we go from there. And I just want to point out a couple of things here as we wrap up, because we have three minutes left. One is all of this is industry standard code that it generated, right? So I have a background as a typescript developer. I look at this code, this is home for me, right? So I might get in here and I might tweak it, and I can, if I want, I can make manual changes, but I don’t have to write. At no point in this process, in what we’ve been talking about in this demo, did I actually get into the code and actually change anything. And I think it’s kind of funny that in the meeting I was mentioning, I did actually go in and I made a tweak directly to the code. Once I actually went in and I changed something without the AI and I broke it. And the one time that I actually went in, I’ve been working with these programming languages for a long time. The one time that I actually thought, you know what, I’m just going to change it myself, I went in and I made the change and it broke, and then I had the AI fix it for me, right? Meaning, frankly, these tools are really smart. And what I’m finding, even though I’ve been using these programming languages for a long time, I’m not really writing it anymore. You know, I don’t even really. I mean, I look at the code typically, but I don’t actually write it. And the reason I bring that up is to say, if this is not a technology that’s familiar with, that you’re familiar with, don’t worry about it. It’s nota something that you actually need to be familiar with. It’s. The AI will understand it. You can ask it questions about what it’s doing if you want. It’ll explain what it’s doing and it can take you through it. It’s honestly a lot like having an EDI specialist on your team who you can just say, hey, can you set this up for me? And it’ll do it. And you’re talking at the business level and it’s working with, you know, with all the details of how to actually make that work.

Aaron Magid

So that’s, that’s it really. I mean, that’s, that’s what the process looks like. And oh, there was one other thing, but if I really have time. Just going to mention Dan mentioned monitoring earlier. These applications are, are modern. They are containerized, they are, they fit with all of the modern standards for applications. So these are things you can deploy to the cloud. You can deploy them to your own servers. If you, whatever platform you’re on, they’ll run there. And they also are monitorable. Right. So this application will tell me exactly what’s going on with it. I can see its log data, its errors, its successes, its warnings. I can see its metrics about how much it’s processing, how long it’s taking, those kinds of things. I can set alert conditions. Right. This is all built in to this process. And again, it’s there to say, I don’t want to have to worry about edi. I want to worry about my business process and let the tools handle everything that I need from a technical perspective. So I hope that was useful. With that, I will kick it back to Maya and Dan to close us out again. I hope that was useful.

Dan Magid

Great. Thanks, Aaron. That was terrific. And yeah, Maya, I’ll leave it to you to close it up.

Maia Samboy

I was just gonna say thanks everyone for coming. If you get an email from me, you will get an email from me by tomorrow with the recording. And if you have any other questions or would like to book a follow up with Aaron and Dan, feel free to reply to that email and I can set that up for y’.

Dan Magid

All.

Maia Samboy

So thank you so much for coming and hopefully we’ll see you next time.

Maia Samboy

Thanks, everybody.

Maia Samboy

Have a good day, y’.

Aaron Magid

All.