Hello and welcome to Modernization Roadmap for an AI ready JD Edwards. Today's webinar is sponsored by Napa State and produced by Actual Tech Media. My name is Scott Becker with Actual Tech, and I'm excited to be your moderator for this webinar. So I've got a lot to cover today. Some demos to show. You can't wait to get started, but first, do you want to go over a few housekeeping details that are going to help you get the most out of this session?

So let's start out in the chat tab on the far right hand side of your screen. A bunch of you have already been saying hi in there. It's great to see you guys where we, see folks from, California, Atlanta, South Carolina, Texas. It's, it's, if your state or country seems underrepresented, you can correct that by giving us a greeting in there.

So the chat is, of course, for hellos like that. You don't have to stop there. You can share your thoughts about the topic today, your reactions to the presentation, insights that you have that you want to share, whatever you're excited about. So, so a good place to go. In the unlikely event that you have any technical issues during the session, browser refresh will fix most of those.

Other thing to try is the light mode option at the bottom of your screen. But if none of those text fixes a problem for you, my colleague Irina and I are going to be standing by to help you out. We'll be monitoring that chat for any issues that you might have. Now, the absolute best way to dive into the information we're going to explore today is by asking questions.

So let's take two hops over from the chat tab there to the Q&A tab. So that's the spot. So chat for all those general thoughts and ideas, connections, Q&A areas where you should go to ask questions directly of our expert presenter. And then obviously team, that's the key that we'll be looking at during the Q&A portion.

And we'll do our best to get to as many questions as we can. But we do have a lot of info to cover today. Depending on how many questions we get may not get to them all during the session. That's the case for your question. Don't worry, we'll make sure to pass along all of these questions to the team from the other side for follow up.

Now, another product console tour today is the docs tab right there between chat and Q&A. Had some great stuff to go along with the webinar. So there's a, there's a nice one page modernization roadmap for an AI ready JD Edwards environment. And there's also, a one on one executive session sign up. So this is a session with navigate about modernizing your Oracle powered environment.

So two great resources there. Click around and explore. Also be aware that we have closed captioning available in 18 languages. If you hover over the lower part of your screen, you'll see that CC option as well as the drop down menu for languages. And finally, you may have noticed a price icon. We are in fact giving away an Amazon gift card worth $250 as a prize drawing.

At the end of the webinar. You must be in attendance for the duration of the live, but to qualify for the prize and you got to meet the actual tech media prize terms and conditions, which you can find in your docs tab as well. All right. Well, with all that housekeeping out of the way, I think we're ready to get into our session.

I'm really excited to introduce you all to our expert speaker. Today. We have Connor Woods, who is, JD, Professional Services lead at native site. We bring Connor on stage here. Connor. Welcome. Thank you. Scott. Hi everyone. Yeah, thanks for being here. Well, I'm looking forward to this. I know everybody else's too, so I'll. I'll, leave the the of the stage to you.

Thank you. Sir. Okay. Hi, everyone. Let me introduce myself first. My name is Connor Woods, and I'm a director of JD Edwards Professional Services at Nava side, part of Accenture. I've been working in JD Edwards for over 20 years now. For some of you JD gurus, I started with JD Edwards, actually, as a developer and as I in 2001, I came out of uni as a graduate.

I then moved into the CNC world, progressed on and took a global JD Edwards team management role for all our tech consultants. I had a brief secondment, into a DevOps team, novice site. And finally what I say came home, to JD again and I moved into the project services team where I am today.

So at the moment I work with JD Edwards customers both existing customers and new customers, to ensure they're getting the most out of their JD Edwards product and hope to to help them deliver some new features and functionalities for users. So at the moment I run a global team of functional and technical JD consultants really focused on all modules, all platforms, all combinations.

So if there's anything at all, even after this call that you guys have got a question on or anything you want help on, you can contact me and obviously the center, and we'll be sure to try and have a chat. Okay. Everything in our webinar today really comes from what we're working on with customers. So I want to just, you know, make the point that we're hoping to share some real experiences here, give you some honest advice.

And as you heard at the top in the intro, if there's any questions or any comments, please put them in the chat. We'd love to get back to those towards the end of the webinar.

In terms of today's agenda, broadly speaking, three areas. For today we'll focus on number one foundation first. So why modern ization really matters? In an AI roadmap. Well then look at acceleration next both in cloud and what we call continuous innovation. And finally outcome focused. Last right. We'll look at some of the concepts of AI, both genitive energetic AI and have an exploration of some demos and guidance.

So we'll go from foundation to acceleration and hopefully to some outcomes. Today.

Before we get going, I just want to put this slide up. I want to introduce what we here at novice say to say to consider the phi maturity levels, which we look at, as a diagnostic for AI, acceptability or readiness. Maybe you can take a moment to identify where you think your organization is now, based on what you know about your 12 or 18 or 24 month roadmap, where you want to be as an organization.

In terms of AI, we see most feel your rates at the early maturity levels. What we often see is there's an actual flurry of activity, some quick wins and levels one and two, and people think they're making good progress. And then as we start to try and what we say punch up a little to levels three, four and five, we get a little bit of resistance.

It requires further adoption and buy in and maybe the fundamentals, maybe the the platform, the framework, the foundations, you know, are not there to enable the AI. So we're going to talk a little bit about that today. But I just wanted to put it up there. And everyone can be at different levels. Again ultimately it's where you as an organization want to be as part of your roadmap, and we can certainly help you get there.

Okay, so just moving on. Then let's first look at establishing the foundation for AI readiness.

So in terms of modernization, let's have a think about why the modern infrastructure really matters okay. We can see many of the common challenges on the left in relation to ERP, specifically to JD Edwards. Okay. We know that versions change quick. We know the support matrix. The Amt is continually evolves more so than ever in AI. And we know that, you know, older OS's, older database versions, legacy WebLogic server versions.

Legacy. Maybe you know WebSphere a lot of those features are holding you back. If you're trying to modernize and adopt you know for AI. So what we would definitely be saying is that we see a lot of the new AI and automation solutions being built upon the leader versions. It's nearly a prerequisite now for a lot of the functionality modernized, and really puts you in the right spot to take advantage.

I want to call out one particular one in the left. We see a lot of uptake around automation and AI in terms of integrations and interfaces. It's a it's a nice sweet spot, to automate, and only then wrap a little bit of AI around that, to help the process. But a lot of those integrations get held back when they're quite cumbersome.

Maybe they're file based or they need a lot of hand-holding. You know, there's a lot of, I guess, incomplete error management. So there's really a nice area there in the ERP challenges, you know, to try and really see something better there, maybe moving towards a more API based solution. We can all envisage and point services orchestration.

So that really gives us a little bit more opportunity, as we start to think about modernization, I would also say as well that I will help your system move quicker and process faster, but your platform must be ready. Must be capable to scale for that. So when we're in certain cloud platforms, we know we can dynamically scale.

We can proactively scale ahead of time. So again, having that modern infrastructure and the capacity to move quick and move as quick as I need to move will really put you in a really good spot. There. And I would say to you, if you've got to take something away from this slide, AI readiness really starts with modernizing the foundation.

And that's before we started that define AI projects. Sometimes people get ahead. They've got the AI projects lined up, they know what they're going to do. And then when they come to the platform, there's limitations or there's complexity in terms of the security of the platform. It maybe can scale to meet some of the peak demand that they're planning to consume for AI.

So you please think about modernization first and what you need to do to get that foundation in place.

So on the theme of modernization, let's have a think damn what that means for JD Edwards itself. We've talked about the platform. We know that we need to have a platform that's modern. It's up to scale, and it's at the front of all the compatibility. Similarly for JD Edwards, we need to think about the modern aspect. For that, I would say think of modernization not as a version but more as a capability.

Right? We know that the modern the toy, the naming convention for JD Edwards now is based around release 24, release 25. We've seen released 26. And this lets you know where you are in terms of versions. But if you're not using the features within, let's say, release 26 to the maximum, you might not be fully modernized. Okay, you can take advantage of the functionality, but you have to enable it.

You must deploy it to be truly modern in this case. So when we moved to release 26 as customers, if you guys are there or you're thinking to move in there, you know, Oscar sales questions like, are we using orchestrations as our preferred solution for development? Are we implementing widgets out of the box? Are we using self-service password reset automation?

Are we deploying Iwan pages? Are we using logic extensions, the new enterprise modeler. So all this functionality is what we're perceiving to be a modern JD Edwards. Platform. I would also say as well, you know, keep in mind that Oracle invest a huge amount of R&D dollars, you know, new JD Edwards features and functionality. And we, you know, as Oracle consumers, as people who have maintenance contracts, we are paying inadvertently or indirectly some of that.

So please make sure that you take advantage of the investments that you guys are making and pass that investment back to your users so that they can take advantage of that functionality. So that would really help ensure that we've got a current modern platform. It's you know, it's all about making sure that not only are we taking the issues, the tools, but also that we're enabling the functionality that's coming with the platform.

The second area in the middle layer is really a key one that we've probably all heard of, and it's the use of automation through orchestrations. Okay. So just to set the scene for anyone who's not aware, but, you know, JD orchestrations, you know, for novice, say to sage, your enemy is key for modernization and deployment of AI because it really provides that bridge between the JD Edwards instance and other enterprise systems.

So it enables real time data and workflow integration. We see orchestrations everywhere now, but orchestrations will really unlock the power of AI. And we'll see a little bit more about that in a demo later on. We would also say as well, right, if you keep your web tier updated because the orchestration layer, really relies on that. And it's, it's something that's fundamental.

And the leader versions of WebLogic, you know the latest OS is, you know good specs on the web layer. But really ensure that you're in the right place.

Coming along to cloud ready resilient infrastructure okay. So having that resilient and scalable infrastructure will really help support you in your AI journey. So if you think about it, when you start to extend JD with AI, you know those systems that you're extending to become even more critical, okay, systems and processes now become interlinked, okay. Therefore, you need to ensure that as you grow that landscape that you're evolving, your doctor position.

Because now what we might find is you might have all these touchpoints, all these systems hooked up, but you've maybe not got all of those, components and your doctor, plan and runbook. Okay. So very important as you scale out and expand your enterprise platform that you stand the doctor at as well, I would also say something quite important as we start to develop some AI solutions in the near future.

Really make sure that your non prod environments are equally well aligned as your production in terms of consistency really around solution, because it's going to become really important. You're not testing, okay, I will be quite powerful, but it can also, you know, extend though quite a lot. So you want to make sure that as you're testing these orchestration and AI enabled solutions, that you've got a really valid test environment for, that you really don't want to be deploying this functionality initially enemy or certainly where avoidable directly into production.

Okay, we all know that we take things up through DVP. EPD typically, so we should think like that as well when we're developing our orchestration and AI solutions. And also think about high availability for orchestration in line with that resilient infrastructure. If you start to push more requests and more load onto the orchestration layer, you get more dependance on that layer.

And why I want to flag this up is because recently, we had, announced those from, someone who wasn't a customer of ours, but they were in a bit of a, a pickle. And really, what happened there was that engineered quite a convoluted integration, layer where they had lots of individual eyes. Ports for those that can understand that as URLs, if you want to think of it.

And they're assigned various URLs to various, applications. And because of that, they'd be known to be in one port would become overloaded or unavailable. That application would go offline, and that pattern would repeat from time to time. Now, there was a fundamental reason for that. At the WebLogic layer, which we we helped address. But I guess the key point is they didn't have any sort of vertical or horizontal load balancer in place.

So they had if I remember about eight as instances. But none of them were joined up to a load balancer. So they reached individual points of failure. So again have a think about that. Just as we as we scale up and we put more, more effort, and more reliance on these certain layers, make sure that you've got resilience in place so that, that layer can cope with any single point of failure.

Okay. Next we'll have a look at enabling AI through cloud and what we call what Oracle calls continuous innovation.

Excuse me. So continuous innovation for anyone who's maybe not very clear is really Oracle's strategy for continual value add updates for JD okay. So Oracle years gone by with release code updates issues issues or even tools periodically okay. Nowadays we're seeing quarterly updates or what they call drop sometimes. And we're also seeing a much higher frequency of tools updates.

And we all know now that we can adopt certain tool web packs and things. So you can do a little bit of mix and matching. So there's definitely now a faster adoption of new features available with Oracle's continuous innovation strategy. That allows us to do that. In fact, I think we all remember moving probably from certain versions from 9.0 to 9.1 to 9.2.

And for those who maybe don't know, 9.2 will be the last release of JD Edwards officially. But the only reason is they've now completely decoupled everything so they can deliver continuous changes now through the tools and the issues. Okay, so that's where the whole released 24 release 25 is coming back to. You are continually getting updates, at the technical and functional level without the need to take a mass upgrade.

In fact, the word upgrade is rarely used now, the Oracle terminology is update. Okay. So I think very much in line with that continuous innovation. So we are getting quicker access to new functionality. Oracle are releasing new functionality that's embedded for AI or will enable new functionality. And the continuous innovation will allow you to adopt that functionality earlier.

It also allows you to stay ahead as a business. Okay. You can respond to demands from your users, your customers, your suppliers much quicker. Now. So when they come to you with new challenges about integrations or interfaces, you protocols, because you're in that continuous innovation loop, you're also picking up on those examples. So we had a recent example right.

Protocol called TLS, where a very large supermarket chain, rolled out a de facto standard that everybody the talk to it through EDI, on a secure channels had to be at a certain TLS level because some of our customers that we support and host where you know, resisting and staying on an older version to meet internal requirements, they then had to sort of pivot quite quick, to, to ensure they could continue to do business, with, with the supermarkets.

So again, security and stability come with that as well. When we, when we keep our continuous innovation in play, we also know now that we see a bigger shift in customer approach, years gone by, I guess we all recall the big buying upgrades. We would maybe not do anything for 3 to 5 years, and then we would have to do the big buying.

Whereas now we see a big uptick now in customers preferring to do small but frequent changes, which is more preferable, I believe, and favorable that a periodic big buying disruptive upgrade. So again, the continuous innovation fits in line with our model, let alone often okay, keep apply and keep updating. You keep innovate and keep changing. So really keep an eye for the the continuous innovation model at Oracle.

You'll see this pushed a lot. And we can define this to be key. It's a it's not something you do once and you fire and forget. It's something that we encourage our customers to, to keep on top of every six months, every year and build out into a multi-year strategy.

So we've had to think about the application itself. We've had to think about, you know, in terms of innovation. And I want to just have a little bit of a chart on cloud. Okay. So again, if we think about cloud, as a platform, I guess 5 to 10 years ago, we, we probably I thought of it certainly just as a public data center.

I thought of as somewhere where maybe ten years ago, we could tell customers there was an opportunity to relook at your infrastructure. You could, you know, keep your costs low. They were predictable and you would obviously get, you know, some performance uptick if it was well configured. Fast forward now and, you know, the cloud platforms are much more than just hosting okay.

We know that the cloud platforms like OCI, are offering you a multitude of SKUs, server shapes, and different disk types, different CPUs, all to deliver high performance compute based on what the application requires. We also know that certain AI functions will be natively available. So if you are in Oracle or Azure or Google Cloud, you may find that you get preference to access early adoption of features.

Or you may get some, reductions in terms of the price of credits that you need to access those features. So again, aligning to a cloud that has some of the functionality you're after will open up some more native IOP AI opportunities. Even if we think about the recent Oracle 26 AI database, okay, AI is in the near no.

So we're starting to see again, a lot of the cloud vendors deploy the AI infrastructure into the operating system and into the underlying hypervisors, and even into some of the application services like the database. So really important to envisage cloud, you know, not just as the hosting platform, but also as, you know, the enabler for your innovation. As well as because we.

In terms of scalability, yes, you can scale on private cloud. You know, that's perfectly, possible. Everyone knows that. But certainly in terms of public cloud, we find it a lot easier, for customers, you know, there's more options. And I think one of the options that our customers really like, in terms of having that scalable platform is the ability to, essentially maybe utilize a pay as you go model for the cost.

Okay. So depending on how you structured it, we often like to propose to customers to scale maybe a month and maybe a year. And we've got a couple of peak customer load cycles in terms of when we know their order entry will go up. So again, we can proactively get in front of that and scale up and also skylight, some of the servers, just to absorb a little bit of that load.

And to make sure, you know that everything goes really well. So again, cloud as a modern platform will natively allow you to do that. And last on this section is security. I don't want to overlook security. I've spent a lot of time talking about, the good things. But in terms of security integration, you know, as I joins up with other enterprise systems now as we start to embed that into our environments and cross environments.

And, you know, I a little bit more autonomy in terms of what it does. New attack surfaces a new threat, actors become a risk. Okay. So that's a given. But obviously with cloud comms cloud security and that in itself is leveraging AI to enhance the protection of the environment. So again, cloud is, embracing the AI within itself, not just to provide services, but to also ensure that we've got security and resilience, under control as well.

So.

So I just want to have a bit of a pause, and just sort of cover back a little bit of what we've really uncovered here. Okay. So we know that we've talked about how the Oracle applications keep in the modern we've talked about the platform in terms of modernization, and we've talked about continuous innovation. And really when you put those 2 or 3 things together, you really then supercharge your gainbridge platform, okay.

You ensure it moves away from being a purely a system of record to a platform that's largely AI enabled. Okay. So all those components combined really well, then ensure that you're in the best place as part of your roadmap to start leveraging AI solutions. So I want you to keep that in mind in terms of the foundation, in terms of this continuous adoption, and ultimately then by being AI ready, we also know you can think of it as continuous optimization and innovation.

So as your business brings new requirements, you're at the best place to decide what the best option is. You've got the right tooling, it's on the right versions, and you've got access to a lot of the native AI services. To make that solution a success.

Okay, let's have a little bit of look now move it away from the theory, I guess, and looking a little bit more towards outcome and even to see what's possible in terms of where we can use AI.

Okay. So don't worry too much about the boxes. It looks a bit busy, but I guess the red boxes are the key words. Which I'll talk to this was a use case that we worked with, for an existing customer. And really the challenge they had was they had lots of field engineers. I'd visit and sites, performing some work, and they encouraged their engineers when the right, you know, for their day, work to, scan photo their expenses, whatever that may be, and email them in to a common distribution list.

Okay, so sounds great, but the email inbox was absolutely bulging, as you can imagine, with receipts. People were falling over each other to try and process them. And at peak times there was a backlog where they could just never get through. Engineers were getting a bit frustrated that their payment cycles weren't commentary in time. So they asked us this try and help them see where we could get involved there.

And what we really did was we married up, orchestrations with, a little bit of AI, to help solve that problem. And what we did was we basically got the, AI services to scan the receipts, whatever they were on, and tag the appropriate data, determine what that, document was. We called it a document.

And then ultimately upload data. JD Edwards, you know, through, the orchestrations to start the expenses process. So we really helped out there by doing a lot of the kind of manual work that would need to be involved. And, and really, the use of the orchestration there allowed us to talk directly to JD Edwards. And it also it was able to work, and call the, the AI agents to do the OCR recognition to up to, to move the document to certain folders.

So really something to think about. And a simple example, and quite often asked where do we start with AI. And you know that could be a good example. Okay. There's also some people who may use certain scanning products on the market. Again, based on your use case, and the cost and everything else, there is options now to deploy some of these OCR type functions, to perform similar levels of work.

So this was a really good example of generative AI. Okay. And I think we'll run a small demo next. And I'm going to just explain what's going to happen in the demo. It's pretty pretty short. So I'm going to give you a heads up on what the demos and then we'll get the team to run it for you.

But we just wanted to show you something on screen where we take, a random unstructured document. So it's not a document that we're used to seeing. We picked it up off ChatGPT, and we will basically show that in voice on screen. You'll see some of these data on the invoice. We will then attach that document and JD Edwards just to bring it natively into JD Edwards for the demo.

And then we will run an orchestration which will basically, deploy sorry, deploy will actually invoke, some AI services to extract the data from that invoice. And then that same event will then attach that document with some of the data into JD Edwards. So just to prove how easy it is to get an orchestration on an AI service connected, we thought we do this little demo.

And if you guys have any questions on it afterwards, you can let me know that you will see a little chat screen flash up like a little, dialog box. Again, we just wanted to make it clear how quickly and how easily it was able to determine the invoice numbers. The dates, the items. And again, we didn't have to teach the the OCR, the AI service, this document, we literally gave it to it and it can infer it.

From what it knows. And if you give it a very different example, it would most likely get the same hit rate as well. So extremely powerful. And extremely useful, for these kind of key cases. Okay, maybe. Scott, could you guys run the demo for us? Would that be okay?

Okay. Thank you. Scott. So really, they're quite a simple example. And and you see, when the, when the button was clicked, it was actually then connect the night to the internet, security to run the service, to extract the data. And then you seen some of the data in the chat bot. So effectively very easy to call and integrate with AI.

We've got our orchestration layer. We subscribe to some, you know, OCR cloud services, a small bit of networking to ensure that it can be securely connected. And then it's really just plugging in the orchestration, to do some of the heavy lifting. So hopefully that gives you a little bit of an example, to think about.

You know, whenever you look to see what use cases there might be out there, you can start with something quite small, like the one I just described with the payments, sorry, the expenses or even something like this where you're solving a problem that, you know, maybe if I take someone else quite a lot of time or that you don't have the resources, to fulfill.

So before we move off this screen, I just want to pause because we're going to probably get asked this at the end. We see this question a lot. So sometimes people say, well, when would you use an orchestration versus an AI service? Right. So what we would typically say is for those that don't know, orchestrations are quite late.

They're quite good for probably repetitive, time consuming, consistent execution. You know, pretty low complexity cases. Right. And where we see AI coming in is more around where there's a larger scale, automate automation required, where you need to do maybe some predictive analysis, where there's maybe some decision making that has to be done. Maybe you want to give the the service a little bit more autonomy.

And ultimately then that allows you to make those decisions. So I think really they can coexist. And in fact, you know, I really enhances the capability of orchestrations. Okay. Yeah. And that's why the two sets so nicely in place are well together. So certainly start small. Define the process. Well, you know, look to see if orchestration could help automate it.

And then you can wrap AI into that where the complexity requires something that I can natively fix or do in a much more, you know, convenient or efficient way.

Okay, so I want to pause this thought. I'm just going to let people have a look at this slide, very briefly, because we're going to introduce a topic called a genetic AI.

So here at a center, we have built what we're calling the AI refinery, which is our a genetic AI foundation platform. Okay. And the example we looked at before was what we're referring to is generative AI, which is more pattern based. It's something know like we've seen where it's predictable. Where is the agent? AI is the next level of AI solutions, which really will start to focus, a lot more on autonomous task execution, you know, usually performed by automated agents and quite often without clear rules and instructions.

Okay. So that's where some of the fundamental differences are. And here we start to talk a lot more about concepts about reasoning and planning and memory. Requirements. All all introduced. Okay. So a genetic AI really builds a network of AI agents. And those agents all have different purposes and ranks and rules. Okay. Some people have likened the genetic AI to bees and a hive.

Okay. All working separately but together towards a common goal. Okay. And this is like the agent architecture we refer to okay. So it's really important to understand that the Accenture AI refinery, it isn't just one agent, you know, it's a sequence of different agent services all working together to provide solutions. We probably got about 100, you know, agent solutions available.

And we're starting to see now, you know, a huge opportunity to start to sort of embed some of this functionality. And with JD Edwards.

Excuse me. Okay. So in terms of let's just move on then. Sorry. In terms of why I refinery matters and why we'd like you guys to think about it. Okay. Some of the key points really are on screen there. But it's a ready made platform okay. For multi-agent execution okay. We've developed the framework which you guys can adopt as customers based on your business needs.

Okay. So it really makes the AI enterprise ready by default okay. It's a genetic AI. So it's not just chat interfaces or basic commands okay. It's agents work in collaborating together. It works across systems. So it's not specific to just for example JD Edwards. You don't need to replace your systems okay. You can plug the connectivity in so that the AI refinery solutions will actually be able to integrate with JD Edwards and maybe with, you know, an other product, your DSI scanning or any other interfaces.

So it's really, really powerful in that respect. And it's fundamentally an accelerator. So it's pre-built agents with standard patterns that we can start to leverage and JD Edwards solutions, and is essentially release new functionality into our environment for these, you can leverage that functionality straight away.

So this is a really good example of the, AI refinery in play. And this this screen here will show us now we've taken one example of it, which is the invoice to payment, okay, a genetic AI system. And you can see in the right it's got a hierarchical structure of all these agents. We talked about, the bees in the hive working together.

Okay. So you'll see this. We've got multiple different solutions. We've got invoice to pay. We've got order to cash. We've got, you know, inventory, we've got stock reconciliation. And these have been designed in such a way that there's a hierarchical structure to them. And at the top you usually have, you will always have, some kind of control in an orchestrator agent.

And then you have layers of other agents that are assigned to specific functions. And ultimately, these agents will work together and interact with JD Edwards through orchestrations or potentially talk to database direct to perform some of the functions that we, we want and our environment. So in this example, you, the AP teams, similar to what we talked with, expenses, lots of high volume invoices, you know, lots of touch points and probably struggling to get through those at peak times.

And what we did was we explored an idea and to leverage the invoice to pay agent, which will really help van by reducing a lot of that manual work. It can improve accuracy. Okay. We can avoid human error. And ultimately then at scale, we can accelerate the finance operations over the team so much quicker. Okay. End to end automation okay.

So we free up people to get on with probably more valuable tasks. Okay. There is still a human in the loop. Okay. We will hear this term a lot. Human in the loop. So there is still the opportunity, for approvals to be set and, you know, certain guardrails if you, if you want, you know, a human to have to still approve or reject, the steps and it's very much proactive.

Okay. So it knows how to, resolve certain problems itself. And it'll go and try and propose what the fix is. And the human can obviously interact it and really then smarter process. And so it can also work out things like priority okay. If certain payments are due sooner, it knows to prioritize, those kind of tasks. And if certain customers have been tagged potentially VIP or urgent pay, then again, it's capable to understand, you know, some of that business logic.

So I think in the next screen, we will start. We will see a demo of that, in play. No, the demo was about seven minutes. But it's pretty comprehensive. And I would like a if Scott would be happy to, to run the demo for us. You can hopefully see the Atlantic. I process for the invoice to pay, and some of the decision making that goes on in this demo.

Welcome to the demonstration of Accenture's agent invoice to Payment Assistant, an AI powered solution designed to streamline and automate the full lifecycle of invoice processing from ingestion to validation. This assistant supports financial operations teams by reducing manual effort, minimizing errors, and accelerating payment cycles. Finance teams today face increasing pressure to do more with less from handling growing volumes of invoices to ensuring compliance and accuracy across vendors and systems.

The financial operations landscape has become more complex. Teams are often bogged down by manual data entry, scattered communication trails, and legacy systems that lack flexibility, real time insights, and require manual field validation across multiple platforms. These inefficiencies lead to costly delays, compliance risks, and reduced visibility across the payment lifecycle. But with Accenture's multi-agent assistant, financial operations are transformed into a streamlined, intelligent workflow.

Our platform is powered by a collaborative network of AI agents that work together to automate tasks, reduce errors, and provide actionable insights. The system is powered by three core super agents. The Invoice Flow Agent handles the full invoice lifecycle using 12 utility agents that automate each manual step. From parsing to validation. The invoice revalidation agent oversees the revalidation of previously processed invoices using seven utility agents to check for updates and revalidate when exceptions arise.

The Exception Manager agent calls to one of the four utility agents to run exception specific workflows. Together, they deliver fast, accurate, and adaptive invoice processing at scale. Let's take a closer look at how the solution works in action. It's 10:00 am and Jordan, an accounts payable manager, logs into the dashboard. Right away. She is greeted by a ranked list of the top invoice cases that need attention prioritized by criticality.

These are flagged based on priority, giving Jordan an instant pulse on what matters most. A new invoice has just come in. She clicks the Add Invoice button, uploads the invoice file, and it instantly appears on the left panel for quick review with a single click on process, the system takes over, launching the agent pipeline to handle the rest in a live production environment.

This entire process would kick off automatically the moment an invoice is received, but for now, let's see the system in action. Behind the scenes, the orchestrator takes Jordan's request and activate the Invoice Flow Super agent to oversee the process. The first Utility Agent invoice digitizer performs optical character recognition, or OCR, to convert the invoice into structured text that output is handed off to two agents running in parallel the field extractor agent, which pulls values from the digitized text, and the field extractor vision agent, which uses image analysis to extract fields directly from the invoice layout.

Their results are then reconciled by the Critical Fields validator, which cross-check key values and resolves any discrepancies automatically. Next, the invoice is classified by type and region using the invoice type classifier and field level classifier. From here, the system activates a series of agents to validate and enhance the data. The exception handler identifies Po or Non-po status and addresses exceptions detected.

The field driver adds inferred fields based on patterns in the data and the anomaly detector flags. Irregularities against predefined business rules. The field validator checks line items against ground truth. The Watchtower agent assesses urgency based on key field metrics, and the report generator creates a full audit trail. Throughout this process. The right panel offers a real time view of agent interactions, while the center panel updates with newly derived and validated fields, giving Jordan full transparency without lifting a finger.

Now, let's say Jordan notices a field that doesn't look quite right with just a click, Jordan navigates to the specific field, opening a pop up editor. Here she can view values extracted by both the field extractor and field extractor vision agent, compare them side by side, and even enter a custom value if needed. Jordan can also leave notes or upload a screenshot to give context to the update.

Once satisfied, they click Save Changes. Scroll down the middle panel and hit revalidate, seamlessly triggering the invoice revalidation process with all the edits included. When revalidation is triggered, the invoice Revalidation agent takes charge. It starts by activating the Revalidation Intent Detector, which analyzes Jordan's update and determines the most appropriate reprocessing path. The exception handler then rejects the invoice for POA or Non-po status and flags any related issues.

From there, the flow mirrors the main pipeline. The field driver regenerates any dependent values. The anomaly detector scans for new inconsistencies, and the field validator ensures the updated data still aligns with ground truth. Finally, the Watchtower Agent Reassesses invoice urgency based on the new inputs and the report generator produces a fresh end to end audit of the Revalidated invoice, all visible in real time through the same informative huddle.

If everything checks out after revalidation, Jordan simply clicks, confirm changes and post finalizing the updates of course, not every invoice flows through perfectly. That's where the exception Manager super agent comes in, overseeing three types of exception workflows the goods received, or your workflow is triggered when there is no record confirming whether goods have been received. The Non-po workflow handles invoices without purchase orders, where critical information, such as general ledger or cost center details is missing.

Lastly, the line item mismatch workflow resolves discrepancies between invoice data and system records by generating an email to the responsible party. For clarification, let's look at one of these in action. Suppose Jordan's.

Okay. Thank you Scott. So really hopefully that gives a little bit of an insight. In terms of. So I just observed the question there. We just cut that short. It was going to go off into another area, showing the email exception report. But what we really wanted to show this group was the combination of agents all working together here to process documents and to make decisions and then come to the next, the next I come, so really there we see the, the, the hierarchy of agents all working together, which is something that Accenture has built into the framework, reusable agents across, you know, other ones.

There was some human oversight. Okay. Where required, we can interject and also that overwrite, some of the the decisions. There was a good visual interaction on the screen. And we've got many of these common genetic offerings available around ordering stock, and finance. So please reach out if you're interested to find out a little bit more about how these can work and interact with your JD Edwards platform and how they interact really is just by one further extension, then, because those agents can Nan Cole orchestrations, to post the invoice for payment or in the case of order entry, they can put an order on hold.

In terms of finance, they can issue a payment. So again, the orchestration will really unlock, you know, the ability for the native AI agents to query and also update JD Edwards. So thinking that a little bit further than, we've also been in exploring some other areas where that could help. I think you're placing orders on hold due to credit checks, if we feel, or maybe reviewing stock levels, which are low and, you know, potentially proactively order in stock when it goes below a certain level.

So no one needs to really go and check that, that's something that could be, of huge interest. But I think the key takeaway is why the orchestration layer is super, super important, you know, to scaling up and scale and all these kinds of processes, because it will largely, unlock, you know, JD Edwards and, and allow that native communication.

Okay. Finally then I just want to touch on some design principles. We've had a good overview of, gender. I Atlantic I where I can be be really useful, you know, some some ideas about use cases where we can maybe think about deploying it. But I wanted to share maybe a few design principles as well. You know, just to give the full picture, so to speak.

So we've got 5 or 6 listed here and I'm sure that others. But, you know, these are the kind of core ones that we stick to in terms of the first one there, you know, is really around the hierarchy of agents that, you know, that's fundamentally how the process works. You'll keep that made a bit like JD curdles, you know, for those that are familiar, we've got network kernel security, kernel skew kernels.

We've also got agents that are designed and developed, and good at specific functions so they can ultimately operate independently. But, you know, there's an overseeing agent and they make use of the word orchestration agent, which is, unrelated to the JD Edwards orchestration. But I guess it's sort of a similar, a similar mindset. Right. Second one down is about continuous learning and performance stewardship.

Right. As as these agents become more autonomous, you know, it's really important to sort of mentioned version control. What changed and why? Okay. Because all the humans on the platform are still accountable for the actions taken. So again, it's really important there that we think about the feedback loop and also the stewardship transparency in accountable decision making okay.

We need to be sure that all the decisions made by agents you are explainable, traceable and can be ultimately traced back and owned. So I want to have a think on that. The setting of boundaries and permissions. Okay. So again standard approach would be least privilege access. You know thinking about what those agents need to do and make sure they're limited only to what they need to do.

And that they maintain their access and keep within the boundaries permitted. Okay. Second but last value land. Okay. So really focused in on ROI. You know, these agents will consume resources. We know that. And, you know, those resources may have a cost. So you're keeping your cost, but also keeping your risk and autonomy, you know, carefully aligned to outcomes as well.

So as you come across opportunities for this, some of it may not be suitable, you know, based on, you know, the the cost involved or indeed the risk. So, you know, assess each one, based on what it gives the business and in terms of pros and cons, finally then human trust by design. Okay, I think certainly. And actually, you know, keep it human in the flow, for, for any exception or guardrails.

Okay. I think you really just want to avoid any major issues or exposures there. But as your confidence grows and your experience grows, you know, with your AI journey, then you may be losing some of that. But certainly at the start, I guess small steps, would be the key advice. There. Okay. And I think we are at the end of the webinar.

I hope you'll join in. And then, I'll ask you one more name and pick up Scott. Yeah. So. Great. Yeah. We've got time for, just a couple of questions here. The first one, they're asking, what what types of JD use cases are best suited for a generic. I first? Okay. Yeah. Good question. I guess there's no right answer, but let me give a few ideas.

Right. One example could be around interface and integration failure. Okay. I think there's good opportunity there. For those of us of a technical nature or those of us who are responsible for, you know, checking interfaces or vendor interfaces, feel at the weekends. I think there's some opportunities there, to go and, discover the reasons for it, maybe learn the reasons, give a little bit of autonomy in the AI side to maybe, you know, fix something and replay it.

So again, I think that's one area. We seen one on the on the demo, which was pretty powerful around invoicing. So invoice matching okay. If we think about Po receipt and process identifying mismatches, in that process and potentially taking some actions to correct, inventory inventory, replan. Okay. So, you know, those of us that are on the hook for making sure that the the right level of, inventory, and goods, for manufacturing, are we monitoring those levels?

Do we see demand signals drop? Can we identify shortages ahead of time? And and can we create, you know, the planned orders and also expedite the Po that's needed? So there's so much you can do with that. You'll really so have a look into processes that are critical to your business, but also and, you know, involve lots of touchpoints, lots of people.

And that could be a good starting point. I would say, for some use case, thoughts? Okay. Excellent. Thank you. Time for one more here. If we want to integrate our ERP, like you did with the use case was shared, how would we get started? Okay. Well, I think right at the top of the call, we talked about the platform, so there's, like, a check list.

And I think we might be able to release something like that, but it's getting ready for AI first. So there's a platform. There's an application I think also then identifying those areas, and understanding the process. Okay. I think, there's a common thought process that, you know, if you could have, a weak, process, you know, that that's, you're not well defined, then you're exposed not to.

I will just magnify your problem so I can I would I would definitely suggest to get in there and look at that and then find a couple of chapters if you're interested in our refinery, we've got enterprise ready, you know, pre canned agents that just need plumbed in to your ERP platform. We focus today and JD Edwards, but you know, we go across all the major RFPs and on the, the other applications as well.

So, you know, we'd love to talk to people about that. Super. Well, Connor. Great presentation. Great. Great Q&A here. Appreciate it. Any any kind of closing thoughts? I know you've got, I think a resource slide here.

Yeah. So I think we've got, I maybe mentioned, I guess, to give the game away, we've got an AI readiness checklist, that we'll certainly share, with people, you know, to have a look at it. You know, some of this can be a little bit overpowering. It feels like AI is everywhere. It feels like if you're not doing it, you know, you're behind.

That's not the case. Come and have a chat with us. You know, again, take small steps. Get get yourself in a good position, you know, get a roadmap to find and then gradually build it up. And, you know, once you get up to a certain level, you get buy in from the business, you know, it'll be a great opportunity, to help your business accelerate.

Okay. All right. Well, Connor, thank you so much. Really appreciate your time today. Thanks for all your insights. Thank you. Have a good day, everyone.

All right. Well, we do have just a couple more, pieces of business here before we wrap up. We do want to ask, arena is opening up a short survey that we'd really appreciate if you could all fill out. So this is for those of us on the actual tech media team, so want to know how we're doing, which I like and don't like about our sessions.

We want to make sure we're meeting your needs, and the results of the survey are going to affect how we deliver these events for you in the future. So we'll give you all just a minute to fill that out. While you're doing that. I do just want to call your attention one more time to the doc section. Again in there, there were, the opportunity to sign up for a one on one executive session, one that was signed about modernizing your Oracle powered environment.

And there's that one page modernization roadmap for for an AI ready JD Edwards environment. And we do have one more piece of business before we wrap up. It's the prize giveaway. Today's winner of a $250 Amazon gift card is Ted Ross from Wisconsin. So congratulations to Ted Ross. As always, we'll be in touch about claiming your prize after the event.

Well, that and we have the actual tech media team I want to thank now the site for makers of a possible giant. Thank you to Carl Woods and, and the website team for a really interesting presentation. Thank you to everyone in the audience for joining us. And, we hope to see you again next time. Have a great rest of your day.