STOP Struggling with AI! Unlock Easy Implementation Secrets Now!
August 5th, 2025
20 min read
This presentation by Drew Rob, AI advisor at ERP Suites, covers the key steps in creating an AI implementation road map for businesses. It focuses on identifying viable AI use cases, managing scope, setting milestones, and ensuring data and security readiness. Drew emphasizes the importance of stakeholder involvement, addressing risks like data quality and security, and fostering change management within organizations. Using a sales order entry example, the session highlights how to effectively plan and execute AI initiatives for successful implementation and measurable results. Ask ChatGPT
Table of Contents
- Introduction
- Key Factors for Building a Roadmap
- Factors to Consider in Implementation
- Sales Order Entry Example
- Conclusion
Transcript
Introduction
So I'm going to go ahead and get started with the presentation. Looking forward to this one today. You know, creating an implementation road map is very important in your business when you're looking to implement AI and start your AI journey. So I gave a presentation yesterday on identifying strong AI use cases in the steps behind that. And just a real quick overview of that. I would, I would love for you guys to go watch the video 'cause it kind of leads into this one. And and it's really the four steps are determining AI versus not AI solutions. Oh, sorry, I wasn't sharing my screen. My apologies. Share it now. There you go. Sorry about that guys. So yes, so again, happy to give this presentation today and really looking forward to it. You know, creating an AI implementation road map is essential in it. And as I said before, it streams off of the presentation I gave yesterday on identifying strong AI use cases and, and real quick, just want to go over those four that we did yesterday.
So really determining AI versus not AI and then identifying pain points in your business processes by outlining the current processes you have and what issues you are currently having, what current bottlenecks you may be experiencing as well in those processes. And then moving into actually brainstorming AI use cases. So taking those current problems, those current processes and issues you're having and actually formulating them into strong AI use cases using AI technology. And then lastly, it's prioritizing three or five use cases that we can go into more detail and build out via an implementation road map. So that's what we're gonna really focus on today. Really taking those three to five use cases, we're gonna use one as an example and start to build out that implementation road map at a very high level.
So real quick introduction of me, I've given I think 5 or 6 sessions now. So gonna make it pretty quick. I'm Drew Robb, I'm an AI advisor at ERP Suites. I have been at ERP Suites for five years now, started in the product space working on our various products, clarity, mobility and scanability. And then I transitioned over into the data analytics, data consulting world where I actually spent a lot of my years as a data analyst working in that space. So that was great. It was an easy transition. So in my role now is the as an AI advisor. So my main role as an AI advisor is to really present user groups, conferences, educate our customers about AI. Whether you're just starting with AI, you just hit square 1, you're just starting with those use cases, you're starting to learn about the technology or you've already thought about AI. You've, you've established some use cases. You just got to figure out if it's viable for your business and then we'll actually produce a return on investment. Or you can just be, you know, even further along and starting to implement some AI in your business, but you want to take it further, or you just have a budget for AI and you want to figure out how to use it. I'm, I'm here to help you as well as the rest of my ERP suites team, who you've probably seen throughout this week at AI week.
So with that, I'm going to move forward here and go over just a quick, quick agenda. Keep it pretty simple. So we'll go over the importance of building an implementation road map. Why is it important? Why should you focus on this? Why do you really need it? And then we'll go over a list of factors. And I will say it's a long list of factors. I will try to make it easier with that third agenda item there by using an example of a sales order entry implementation road map. And again, it's not a full-fledged implementation road map, but it's really high level to really get you a good understanding of what some of these factors may be and what may be involved and may be important when building out your implementation road map.
Key Factors for Building a Road Map
So I want to make this as easy as possible. I'll try to keep my eye on the chat up here again, Nate, unfortunately, I cannot hear him. I was having these audio issues yesterday, but I'll keep an eye on the chat up here on the right here. Just ask questions whenever I'll get to them when I can. And yeah, so looking forward to this presentation. And with that, I'm going to move forward and to the importance of building an, an implementation road map. And I really want to start with why road map? And really the biggest thing is to get everyone on the same page, right? Make sure everybody's hand in hand and they understand the planned road map for actually building out a solution. I think the biggest reason for this is really just no, no implementation gives this plan. There's always adjustment. There's always things that might change. Now we want to make sure that those things stay inside of scope. But the changes do occur. You know, a resource may become sick or a budget budget may need to, the budget may need to be put towards other projects that you're working on that are outside AI solutions or AI projects. So it's important to understand that everyone has knowledge about the, the AI road map and the AI plan from the top down, from C level all the way down from the business users. Everyone's on the same page and understands their role in actually building out the making the use case into a great AI solution. So I just wanted to preference that as why road map?
These five key factors here of of the importance of building a road map. I feel like we're very important. I've highlighted a few of them already, but it's really provide clear direction for those use cases. You know, the use case is only as good as our discovery and what we figure out from those current processes, but we really want to build onto them and make them into actual working AI solutions. And that's the important thing. And again, just aligning teams, aligning stakeholders down all the way down to the business users. It's important to get everyone involved early.
You know, the stakeholders in the upper level management might not be involved in the actual building out of the solution and we'll get into that a little bit later, but they're still important in the decision making. And the emphasis users are essential for actually building out the AI solution in in the end, because they're the ones that understand the processes through and through. No matter what you're going through to actually building out that AI solution, ensure progress is measured. Again, you know, we're not going to really show it today, but ensuring you have really key milestones, a great timeline with key objectives that are hit and who is involved in those objectives is also very important when building out an implementation road map. And then the 4th one is, is a huge one that we've heard from customers as well, is really identifying those risks. You know, every use case is gonna come up with risks. You're gonna have things that might come up such as security data, you know, the process flow, maybe it needs to be improved, just various things like that. It's, it's crucial that we identify those and figure out ways to mitigate those risks in the implementation road map to help make the process of actually building out the AI solution more seamless. And then lastly, and then again I mentioned identifying risk is a big one. This might be the biggest one. It's definitely promoting change management and adoption inside of your business. It's important that your organization really understands their jobs when implementing an AI solution. And I've talked about this a couple times in my previous sessions as well. It's really about just making employees, making business users, whether it be, you know, upper level management, middle management, lower management, 'cause we can implement AI solutions at all those different levels. Making sure they understand that AI solutions are there to enhance their jobs, right? Not to take their jobs away, get them involved and actually building out the solution and doing the testing of the solution as well. It's very paramount is in, in, in building an implementation road map. So just wanted to highlight these five key important aspects of building out an implementation road map before we actually get into the factors of that.
Factors to Consider in Implementation
So with that, I'm going to transition into the factors of building an implementation road map.
And, and I will tell you it's a long list, but don't be discouraged. I hope my example helps alleviate some stress and, and gives you a little bit of Peace of Mind when starting to build out an implementation road map. And the one thing I'll also say at ERP Suites is we want to be advisors for you. So again, this is the time that I want to pitch us is reaching out, reach out to us to help you along the path of actually building out this road map. Because we, you know, we deal with building out these solutions every day and educating our customers on on these solutions, technologies, different risks, security data. So just remember that and keep that in mind as we go through this. But again, feel free to ask questions. So with that, I listed out factors to consider here and we'll go through each one of these in the example.
Use Case vs Value vs Level of Effort
So first one is use case versus value, use case, value versus level of effort. So it's really looking at those three to five use cases I indicated before and going through the value and level of effort it takes for each one. Now, even if one may take a lot more effort and might be more important in the end, maybe, you know, the finance team wants to get something out earlier than the sales team or they they're ready to implement the sales team. Isn't it just it's gonna depend on resources, but it's still something we like to highlight here. Obviously scope's a huge one. You know, you want to make sure everything's inside scope, even if changes are made, you want to make sure that, you know, scope creep doesn't show up and and you're able to actually implement the AI solution without going off the rails. Phases and milestones are important as well, and I'll indicate those.
Not milestones, getting the AI project done with that first initial installation, but the phases being, you know, where can we take it? You know, with AI solutions, we use the crawl route, walk run method or phase one, phase two, phase three. So if you start with automation, where can you take this further? Where can you get in the predictability? Where can you go to autonomous ERP? You know, we don't, we just want to build a solution and really stop there. We want to make sure we built off of that solution. So I'll show a quick example of that as well.
Data and Security Readiness
And then just really understanding technical requirements, both the technical requirements flow, how the process flows from your data to the end data source or the the AI services you're using as well as just the technical services in general. What all are you using in the package that we're building out? And then data readiness and security readiness really go hand in hand. One thing I want to preference here that that tends to, you know, it tends to scare people when you talk AI, Big, big uplift, big big changes with data security and data readiness. What we like to preference here is we're gonna focus on that one specific use case, that one specific process. So when you talk about data readiness and data cleansing, data validation and security readiness, it's all about focusing on that one process for data and security and makes it a lot easier to start to implement those 'cause you're not focused on the whole data ecosystem, many different data sources.
Roles and Skills
Next is just roles and skills. Who all will be involved in the process? Who will all be involved in building out this AI implementation road map is also very building out the AI solution is also very important. There's a lot of different people that might get pulled in just having those listed responsibilities are known early who would be involved, who might need to be pooled in it, pooled in at things changes who have a relative title risk management. Again, I mentioned that one earlier key, absolutely key to understand the risk management aspect and and what things may go awry and how to mitigate those. And then again, I'll, I'll go further in the change management adoption on, on one of the final slides. That's very important.
Resource Budgeting and AI Budget
Another two factors that I don't have in my example today, but also very important kind of met your roles and skills with that. Also just resource budgeting, hourly budgeting, allocation of people who would be involved in, in building out the solution would also be involved in a very detailed implementation road map and plan. And the second one's AI budget. Well, what all do you have to spend as a customer on AI this year? What's your yearly budget, quarterly, three-year budget? It just really depends how much are you willing to spend? And we, we sort of start to build that out as well. It is, I didn't have it in here today, but I wanted to highlight that there would be a part of an implementation road map as well and, and key factors.
Sales Order Entry Implementation Example
So with that, I'll move forward. And and this is, remember, this is very high level AI implementation, what it would look like for a sales order entry road map starting first though with the use cases versus effort, this is what it would really look like. And again, this is really highlighting those three to five. I have the sales order entry one bolded here, but it's really listing out what are the different use cases you're looking at. So here I've listed sales order entry, prevent machine downtime and financial digital assistant. And what you're going to see is a quick description of what you're, what you're really trying to accomplish there. And then the ultimate goal is also in that third column. And now I try to categorize these. So a few things which categorizes like a quick win, which we stress a lot about our AI journey is it's a quick win to get executive buy in or upper level management buy in or even lower management buy in just to get AI into your business to produce quick ROI. That's one.
But then as you can see, that third one's a financial digital system. It takes a lot more heavy lifting, a lot more strategic investments, maybe more discovery calls, maybe more data and security things that might need to go on there. It just really depends on, on the use case of how we kind of categorize those. And then lastly, you'll just see a quick little bar of just the value versus level of effort. So we're gonna focus on that first one because obviously document sales order entry from PDF and the JD Edwards is, is a big, big value add as we've heard a lot from customers and the effort's pretty low. If you've seen there's actually a a session we did yesterday on document understanding and if you definitely check that out because if you can see how seamless it is, especially using OCI services out of that that information in. So anyways, we're gonna move forward with that one. Just wanted to give some background on, you know, the sales order entry as that will be the example throughout the whole presentation here.
Phase Approach
So moving forward, just looking at a high level like phase approach, what it would look like. So for example, we start with and really this is going back to what's the future? What would the future look like? So we'll start with that top 1. So that's really phase one, the document extraction. So converting unstructured PDFs in the JD Edwards data, so actually uploading those PDFs in the JD Edwards, that's phase one. That's the automation piece predicting patterns. So document understanding will start to understand product availability, discounts from customers, right? Maybe some price breaks, some different things might change. So it will start to understand the customer's history and what not and start to make insights off of that and provide some predictability there after automation. And then lastly, again, it's just eventually where we want to go. So we want to give you guys an end to end order entry, right? Real time communication with customers such as order confirmation, order status. So having the ability to do that extended out to a customer or even building out a digital assistant. Again, we want to build off of these AI solutions. They only just have one centralized thing. You don't just need to build a document understanding tool. We can start to connect these different things. But again, just really just showing an implementation road map where we want to take it. It's not just a one-time solution here. So just wanted to highlight that for the phase approach.
Milestones
Next we'll go in the milestones. So with the milestones, again, it would be a more in-depth timeline when we build out an implementation road map, but really broke it down to these three things. So really the planning, right? So the ability to train the existing sales order documents, hammering PDF, whatnot, whatever you send in the implementation, still automate the reading extraction of a few. So with a few, then move on to more and then get feedback from the initial end users. So this is the human in the loop. This is when we start to bring end users into the solution and that's when we start permitting that change management as well and then eventually scale. So roll out more document automation to other departments. You saw the milestone phases before we focused distinctly on document on the sales order entry. But with the scaling, we could even start to scale if we want to just stick with automation, the different, you know, upload contracts to a database, expense reports, invoices, stuff like that. So, yeah, so that's, that's the milestones again, will be more detailed in the implementation, but you'll get a good understanding of what it would look like here and what we would provide as well.
Data Readiness
With that, we're moving the data readiness. And before I really get into this one, it's very high level.
I wanted to reference getting your data AI ready, preparing JD Edwards Master Data for success was a session that was presented yesterday by Manuel Naira and Mo Shujaat. And I encourage you guys to go watch that one because that's really the beginning of this. I don't have enough time to go over it at the highest level. So I want you guys to really 'cause that's kind of your starting point for the data readiness where you're at with implementing a certain process AI solution that we're talking about today. But going back to the slide here, so just quick example, current sales order entry. So 80% would come in from a standard PDF form. In this example, 15% handwritten written 5% just come in, in the e-mail, just detailed e-mail unstructured the current sales order attributes and, and there would be more, but those are the current ones that would be listed. Just give an example, we've listed the attributes that would be involved in this. And then historical sales orders in the P4210 believe data cleansing valid. So this is what you know with the customer. We checked the data. We've, we've done the data cleanse and data validation of the, the sales order entry. And we believe that data validation and cleansing may need to be done 'cause this will enhance future predictability for customers that we talked about in that Phase 2 when we were talking about the phase approaches. So again, just an example of what potentially may come up in that section.
Security Readiness
Next is security readiness. And again, there's too many. We have a whole security track during AI week. I encourage you guys to go back, rewatch those. Those are very important because security is a big question that comes up when we talk about AI week. So I encourage you guys to go back, watch those sessions. But this is a high level just kind of how it works for a sales order entry. You'll have JD Edwards role-based security for them that was already set up. Most users are trained on current manual sales orders. I'll be with the document understanding security. The OCI side need to set up secure data transfer from CSR e-mail folder to OCI storage buckets. That's just where they house. We would house the the sales orders that would come in and then create and then encryption to transfer sales order information from OCI document understanding. So again, and then encryption's important mixing OCI with JD Edwards. So talk about that and, and, and what we need to do is probably that security readiness before implementing a solution.
Technical Requirements
And then next would just be the technical requirements. So this is just high-level technical requirements, right? So you have the OCI services on the left, you have the JD tools on the right. So again, just talking about OCI buckets. Document understanding autonomous data warehouse is there. You might use it. You might not use it though. So that's not that's not one that that's an iffy one that could be thrown in. It's an optional one. And then JD Edwards tools. So JD Edwards orchestrations form extension Judy OS actually built this out. So again, just an example of what we would show there. And then we would we would also have a technical requirements. Well, and what that would kind of look like. So sales order would come in via an e-mail, it would go to an OCI bucket, and then once it goes to that OCI bucket, it would be sent to document under OCI document understanding to be trained. And then once that has been trained, it would then get the upload into JD Edwards of P4210.
Human in the Loop
And now I wanted to bring back up again the importance of the change management adoption, the human in the loop. So you have the CSR validation as a sales order comes in, they're keeping tracks and notified via e-mail of the process, right? And even when it gets to the JD Edwards, as you start to train the model, we need the help from the CSRS, especially, you know, look at this example, the historical data from this example, it was not up to par yet. We still need them to help validate and train the model and make sure that, you know, it's getting better and better and, and we can make this process more automated. So human in the loop is obviously very important. It will, you know, it'd be a part of a lot of our technical flows when they get built out in the implementation road map roles and responsibilities again, just listed high-level project manager, you know, obviously important for building out, you know, important for keeping track and, and making sure that the ad solutions build out correctly, right customer service Rep, they really help the organization with the sales orders that are being trained. They know what's coming in, what's going out, what's really coming in. So they'll be a part of that. They kind of understand the flow there. And then really the business analyst, if there's someone that's more involved with JD Edwards really understand the whole end-to-end process. It's very important when implementing this specific solution. And then also the JD Edwards developer again at ERP Suites who want to partner with us, we do have JD Edwards developers that will build out the orchestrations and really embed the tool inside of JD Edwards in the P4210 and then the OCI developer. So that's the one who really trains the models right, gets the handwritten models, PD, FS, what have you inside of OCI and actually trains them.
And then that's me on the end. Just the AI advisor, just make sure everything runs smoothly. So I'm just there fly on the wall, but also helping everyone, educating when see fit and, and helping our team as well with the process and understanding the business requirements. So just thought it'd be fun to just add me on the corner there.
Risk Management
Risk management, again, huge one that comes up. There is definitely going to be more that come up other than these 4, but the big one, right? Data quality. And again, encourage, encourage, encourage you to watch all the data, all the data and infrastructure sessions that have happened today, all the security ones, 'cause there will be more that, that I have not covered here. But really risk management, right? So data quality and, and it's really just about the kind of on the right there.
What you see is, is what we will aim to strive for and it's building robust preprocessing and human in the loop checks throughout the training for different sales order formats and structures. And then you'll see compliance and regulatory as well, right, Continuously monitoring the sales order entry, making sure they're correct, making sure there's role-based security. That's a big one before even implementing the solution, making sure only certain people have access to certain tables, especially when you start to build it out, it's uploading expense reports or other areas of the business that you can think of invoices. It's just very important that securities in line there and, and the compliance there and then the big one, right, security risk and then encryption. You know, we're working with OCI tools going to sorry, OCI tools going straight into JD Edwards, right? OCI has end-to-end encryption. The JD Edwards has its own security. How do those migrate together? How can we? And again, it's, it's something that's why we use OCI services. It's something that happens very seamlessly. But just making sure that's all good is, is another risk management factor.
And then again, continuous training, right? As new data changes, as models update, continuously adding inputs, you know, there might be different forms that come in different structured data.
You know, as we start to implement a solution, we might use a few sales order documents at the beginning. And as we start to make them and we start to add more and they might be different formats, how can we build this out and make it more streamlined and still having that human in the loop as well to help with the training. So there's just a few. Again, encourage you to watch the data and security sessions as well. Those will provide more information.
Change Management and Adoption
And then this is the big one. This is the one I really wanted to get to is, is really change management and adoption because you're going to get the most push back and you might not just depends where your business is with implementing AI. I talked about earlier, you know, you could be, you know, on the cusp of implementing AI, You could just be getting started. Just really depends where you are on your specific AI journey. But change management and adoption, it's really all about just kind of these five phases, right? So the first one's planning and assessment. So really understanding where your users are that change is coming and really telling them you know, how to really plan for it. Who needs to know when they need to know it. Just really telling them how this process will change and how this AI solution, if, especially if they're immediately involved in their day-to-day operations, how this will enhance the jobs and make them better. And then design and prototype and really have the end users and the ones involved in this be a part of the testing and the prototyping and the designing of this solution. Again, I mentioned before, especially the business analysts, a part of the sales order process. They know it in and out. They're going to understand how the process works. So definitely need them to test it, especially if you're working with us to understand the full process of how you how your process works, because not every business is the same. And then throw it again is deployment and testing. Get it in their hands, get in the end users hands, start having them test to be the Super users that we like to call it, right? And communicate, you know, the road map, the timeline again, keep iterating the benefit of solutions, also important. And then definitely the deployment, right? Real-time support. Again, our team would help with the support, obviously will be ongoing support for you guys promoting usability, right? Even though the solution's in there right away, we've seen it before. Not everyone's going to be accustomed to it. Not everybody's going to be 100% into it. Really try to promote usability, address resistance in in in patterns and how to do that is very important. And then post go live support. So that really references, you know, start to reference the ROI. How's it really helping them start to create their presentations, have those discussions right and continuously check on end users on the solution, send out surveys. Are they really liking the solution? How's it going? And yeah, are they still involved with the process? Are they still do they still feel valued? Do they still feel like the solution was made for them and they can work alongside of it is also very important.
Conclusion
So with that, I meant to give you 5 minutes for questions, but I have too. I hope this session was very impactful. Again, go back to and watch the steps for building a strong AI use case. It leads into this one. But I hope that you guys gained a lot of knowledge out of this again, if you want more information about building an AI implementation road map. If you want more detailed information, feel free to reach out to me or my team and let me know yeah. Let me know if you have any more questions. I hope and I know AI week's coming to an end, but I hope you enjoy the rest of it. Yeah, hope it's been good. Just real quick 'cause Nate will kill me if I don't show this. So Blueprint 4D coming up June 9th to the 13th. We will be there. I will be there as well, so we can talk there. User groups, Atlanta, Houston Plug. I'll probably be at one of the Houston ones, but we should be attending all of those as well. And then Nate's actually doing a video project podcast called Not Your Grandpa's JDE. It's it's on YouTube as well as Spotify, It's at ERP suites. I've been on there a few times, you know, talking about AI risk and remedies and document understanding as well. Look, look forward to that. It's not just AI, it's, it's, you know, it's, it's application managed services orchestration Configurator. So very good podcast there. If you guys, if you guys want to get some more information as well with that questions 1:30. No questions. I hope that's a sign that you know you it made sense that you guys gained a lot out of this session. I was excited to give it awesome. So no questions. Well, with that I will, I will go ahead and end it. Hope you guys have a good rest of your day and a good rest of AI week. Look forward to talking to you in the future. And yeah, just have a good rest of your day.
Topics: