Streamlining AI in JD Edwards with Release 24+, OCI, and Franklin Assistant

This session, led by Frank Jordan and Carl Di Giovanni from ERP Suites, offers a high-level walkthrough of integrating Oracle Cloud Infrastructure (OCI) AI services with JD Edwards EnterpriseOne through orchestrations. The discussion centers on recent enhancements in Tools Release 9.2.8.2 and the capabilities it unlocks, such as using REST APIs to connect JD Edwards to OCI services like Document Understanding. They demonstrate how to set up orchestrator connections, manage authentication, and build orchestrations that send and retrieve files from OCI object storage. Carl showcases a real-world example of extracting data from a PDF to automatically populate address book entries, and highlights use cases for the Oracle Digital Assistant.

Table of Contents 

1. Introduction

2. Overview of Tools Release 9.2.8.2

3. The Value of Low Code and Openness of the Tools
4. Release 24 and New Orchestrator Features 
5. Learning Resources and Documentation
6. Live Demo: Using Document Understanding
7. Franklin the Digital Assistant
8. Q&A

Transcript

Introduction

This is Frank Jordan with ERP Suites. I'm the director of JD Edwards Technology and I've also got Carl Di Giovanni with me as well. He's one of our software engineers.  Again, this is a less than 30 minutes, so this is a high level and go through the agenda. Carl's going to show you some things on the orchestrator side. So we'll introduce ourselves real quick, go through. I'll show a little bit about what came in release 24 last year. We'll have some demos of that. And if we have time, we've got an Oracle digital assistant called Franklin as well. But if you have questions, go ahead and we'll try to keep this interactive if we can.  Like I said, I've been with ERP Suites now since 2012, also legacy Oracle, JD Edwards and so forth. Basically, I have AC and C. So this is technology-based, but most of this again is high level — where to go find things.  Go ahead and introduce yourself, Carl. Carl Di Giovanni, I work in our products group as one of our developers, mostly focusing on orchestrator these days. Yeah, he is one of our orchestrator men, for sure.

Overview of Tools Release 9.2.8.2

All right, so as a quick demo, we're going to try to show things as much as we have.  So also just as a quick aside, this is an Oracle trial edition. What I just used there was what we call our single sign-on solution. This is something with Tools 9.2.5.5 and later, but Carl's going to show you basically a lot of this is the OCI connector that was recently put out in 9.2.8.2, which we'll talk about.
It is allowing us to use AI services or other OCI services, and Carl will kind of show some of this.  From a studio perspective, all this talk is about enabling other elements. The key point here is it is a connection. In this case, we've got a couple. One goes into object storage, the other goes to the API for document understanding, which basically you send in a PDF, it goes through, can mark some elements, return a JSON, and read those types of things. It's scanning through the document based on the models that you're running for a custom model.

The Value of Low Code and Openness of the Tools

I'll allude to it, but Thursday at 12 Eastern, I'll have a little bit more on kind of the steps that Oracle put together. Just as background, Oracle JD Edwards cannot natively add AI services into the product. So what they're doing is giving us tools so that we, as business partners and customers, can add the services — whatever type of REST service we want — and integrate it into EnterpriseOne.  I actually like that a lot because it allows us, just like with orchestrations, to use whatever channel you want. It can be your desktop, mobile, an AI agent,  whatever you want to throw at that orchestration and have it talk to. The sky's the limit in that particular scenario.

Release 24 and New Orchestrator Features

So high level again, what they released last September/October in Tools 9.2.8.2. This is great because it allows us to now have the orchestrator connection be able to pass an API key into Oracle Cloud to verify you. It makes it really easy to consume those. It was like I think 150+ Oracle Cloud APIs. It's not just AI. There's infrastructure, services, management of different things as well.  One of the things that's really interesting for us is, we're continuously learning. I know other partners are as well, and customers are starting to get into it.  I'm part of the Quest Tech SIG. We were having talks about it because a lot of this documentation just came out in January/February.
So it's still relatively new, but there are a number of things to go through.  For example, it's not just orchestrator. You have to have knowledge about orchestrations and Oracle Cloud Services. You also have to have the authority to provide the security and logins needed to get to those APIs. There's also what's called buckets or Oracle storage where you're temporarily putting files in and out that you may have to pass along to various services. But you know, again, if you, if you were to come into this, like to go to learn JDE, which I'll have some references here in a moment. But if you're going into it, you basically, depending on your experience level and your expertise, it might be yourself, it might be a team, but it'd probably take one to three days to go through the documentations and the Oracle by examples and the lessons. They’ve done a really nice job. Kudos to Oracle for really documenting this so you can get through the nuances. This is not no-code. I will say that. It is low-code, but it works pretty well.

Learning Resources and Documentation

For example, on Learn JDE there's a link to the AI site. You can start learning how to implement it. They do a nice job showing the actions and how to do that. Another great one talks about augmenting JD Edwards — good one to start with. It gives you a nice overview and task list, kind of like a recipe or cookbook.  Here’s a quick example of how it authenticates using that API key connector: You have your EnterpriseOne users and orchestrations. At certain points, they talk to an OCI service. For example, authenticating and putting a file in storage. That service grabs the file, processes it, and sends back a result, then you delete the file.  E1 security for orchestrations applies just like always, and on top of that, there's OCI security as well. So both sides have to be coordinated.  There are lots of good resources — API references for the 150+ APIs. You can find documentation for Document Understanding, parameters, how to post, examples, etc. The Oracle Learning Library (OLL) is great if you want to dive deeper. There’s also an FAQ section and a learning path for orchestrations and OCI setup.  This presentation will be available and if not, we have email addresses you can contact. Once your OCI setup is in place, it really starts to take off.

Live Demo: Using Document Understanding

Now I'll pass it to Carl.  So part of what we’re doing in our products and AI teams is diving into different areas of AI in OCI. One of those is Document Understanding, where we can take a PDF or image, run it through the service, and extract data.

What I’ll go through today is a quick demo showing how we’re starting to use it:

• How to send a file to OCI

• Kick off Document Understanding

• Get the results

• Do something with them

The kind of the two pro requisites that we'll need. One in the orchestrator side that kind of Frank touched on was needing those connections set up that allow us to communicate with OCI. And for launching document understanding, you're going to need at least 2/1 one service, one to communicate with the OCI object storage AP is and another to communicate with the document understanding. The OCI is, object storage is where we're going to put the file that document understanding is going to process. And it in turn is going to put another file back in object storage that we'll need to pull down in Parse.

Like I said, we're we're kind of folk, but I'm just going to go through a quick example of document understanding as far as the models that are out there there, Oracle does have some pre trained models for things like capturing data off of the expense receipts or supplier invoices or driver's licenses or passports. But you're also able to create your own custom models. And that's part of what some of the members of my team that are part of the same team I'm on are working on are kind of training some of these custom models under document understanding.

The basic steps in orchestrator are:

1. Send a file to OCI object storage

2. Launch Document Understanding and tell it where the file is

3. It places a JSON file back into object storage

4. Retrieve the JSON

5. Delete the JSON file to keep things clean

Once we have the JSON, we can parse and use the data however we want. So probably easier to show it quickly. So I'm an orchestrator here. What I'm going to do is start with just a basic, just kind of a kind of a foundational orchestration here we have that has those five steps needed. One to put the file up into object storage. A second one here to kick off document understanding. And it's going to turn those values I chose that we're going to need to get to the path of where that JSON is. I chose to write it and I a groovy step, but you could also do this with a logic extension. You just need to kind of one of those custom steps to be able to concatenate different values, begin and do a string. This step here is going to go get that file from the the JSON result the Deckman understanding generated and pull it down into orchestrator so that we can have a have some JSON to parse. And then this last step just simply goes out and deletes the that JSON file from object storage.

Just to run a quick example of that, I'm just going to come out here and I have a some pre some PDFS that I have just we're just going to take this information and add it to the address book. So if I take this example, drop it in there just to show you what the output that we get back, it's going to run through those five steps and this orchestration is going to return that Jason output that it pulls down from OCI, that document understanding generated. You'll see there's a lot of stuff in here. It's not all going to be relevant for what we want to do here. The one area that we do kind of look at because we're running, we're basically taking that and do some key value pair extractions. And this is all based on how that models trained, but we're able to get, you know, the name of the field that's defined in the model, that the document understanding model, and then the value that document understanding got for this particular example. So now that we have that, I kind of have a second orchestration that is going to utilize that first orchestration. So you'll see that it just has those stops and now we're going to take it a step forward and do something with that. The values that returned. Again, I chose to do just read a Groovy script for doing this. You could do this in logic extensions if you're inclined to do it that way. But we're basically going to take that, JSON, we're going to parse it, take the values that we get out of it then and then just mapping it. We're going to use a form request in this example to just basically create an address book entry for it. And then once that's successful, actually attached the PDF that we submit to it to that address book entry. Again, you could, once you parse it, I mean, you could do whatever other things you want to do with it, maybe send it in a message or notification. Maybe you need to send it somewhere else using another Rust API or other kind of database connection. But it's up to you what you want to do with it. But if I run this other orchestration that's actually going to combine all that stuff together, grab that same file. Completes so now I have created that our address book record and you'll see that it's in there. Now I have the attachment of that PDF that we used to generate it and we just basically mapped a name address information. So that's out there too. You could then, because it's just now, you're right, we're just creating orchestrations to make that call. We just I've also kind of added a form extension here. If I wanted to do this right from here, this form extension just adds a button that calls that same orchestration. I could run and drop another file in there and then once it returns, I have it return the name up here so I can just use that to search for what it just created. Oops, I miss forgot to clean up my old entry there, but I do have. The same same idea, just just running it from a different spot, passing the address for the information.

So that's that's basically all I wanted to go through there just to show a quick example of how what you can do with document understanding how you can, you know, use orchestrator to make those calls. And and as Frank pointed out, there are all those other AP is out there for all the different other document understands just one piece of the AI services that are available in OCI. There are all those other APIs that you can imagine that you could use kind of in a similar fashion once you dive into those, but that gives a quick example, an idea of what you can do with those APIs and and communicating with AI.

Franklin the Digital Assistant

Back to you, Frank.  Awesome. Another area we’ve been using is having the Digital Assistant call our orchestrations. So in this particular case, like we, we've set up a digital assistant. This is just on AE1 page and you know, I can just say good morning for those that are on the other other side in another time zone. But you know, I could I could click on this. So you know, this could be mobile here. We've decided to put it on a desktop, but the digital assistant, basically chatbot, however you would like to call it, you know, there's various names. It's a little bit more intelligent. This is a development systems and a live demo. So if if something misbehaves, I apologize in advance. Slight slightly different system here, but you're basically in this case this was doing some of the the manufacturing and you know, I might say where it comes back here, it should prompt me back in a moment and allow me to kind of see some of my work order routings. This demo was meant to be shown by someone else, so it’s not perfect — apologies. Drew Rob actually demoed this earlier, and that recording is available now if you want to check it out.

Q&A

We’ll leave a few minutes for questions. Please use the Q&A panel or chat. We’ll also put up our contact info. The other neat thing about it is let let's say if you were on an earlier tools release, maybe you're in release 23. You know, again, one of the things that you can do is consider this is a NOCI. You could set up a trial edition. So the trial editions get released usually twice a year. So like when they just announced the Release 25 update 2 for the Tools 9293, which just came out a couple weeks ago, you can install that into an image. The the cost is usually not too bad, somewhere in the range of maybe depending on the processors and what you ask it for, maybe somewhere in the 30 to $100 a month. So it's complete E1 system.  One question: Do we need an active OCI tenancy to use the APIs via orchestrations? Yes, You need a tenancy and an Oracle account to set up the API key and grant appropriate access.  You may not need compartments depending on what you’re doing, but the access setup is required. What we demoed today is using both a trial edition and ERP Suites’ production environments. When you go into production, you’ll likely want dedicated users and restricted access to only certain buckets or APIs. Thanks, everyone. We appreciate your time. Hope this gave you a high-level look at what’s possible today and what’s coming next. Have a great day!