The Evolving AI ERP Landscape: Practical Approaches to Maximize Your AI Investment

[00:00:00] Doug: Welcome to season two of the Acumatica ERP podcast, where we explore practical innovation for growing businesses. I'm David De Rego, VP of Product Marketing here at Acumatica. This season, we're digging into how AI and modern ERP capabilities are reshaping the day-to-day of businesses. In each episode, we'll talk to the people building and using this technology So you get the real wins, the real challenges, and what's coming next.

Today, our focus is squarely on AI and ERP, the features and capabilities growing businesses are using now, the innovations on the horizon, and what organizations need to know and consider as they incorporate AI into their operations. For insights on this, we're joined by two members of Acumatica's product team, Doug Johnson, VP of Solution Architecture, and Omar Ghazi, Director of Product Management, Platform, AI, and Technology.

Doug brings more than 30 years of experience in product management and marketing, helping shape the vision and requirements behind Acumatica's Cloud ERP platform. Omar brings more than 15 years of experience leading enterprise technology products with a strong focus on AI and emerging technologies that turn innovation into practical customer value.

As part of our focus on AI, we'll take a peek at what Acumatica is delivering for customers with its latest product release. Doug, Omar, super happy to have you on this, on the show. Welcome to the podcast. Thanks, David. Glad to be here.

[00:01:32] Omar: Thank you, David.

[00:01:33] Doug: I'd like to start with you, Doug. Can you briefly describe your role and how AI shows up in your work today?

The solution architecture team serves two different purposes. First of all, we help prospects learn about Acumatica and how it can manage their business needs and make their business more efficient. Secondly, we work with existing customers to help them get the most of their deployment. And my use of AI is also twofold.

I'm using AI to help me perform my job better, and then I also get to use AI to show how our prospects and clients can use theirs better and get more out of their, uh, uh, business as well. So it's, uh, kind of a twofer here with me. Yeah. Awesome, good. I'm hoping we can dive into that a little bit more as we get further into the show.

Omar, how about you? Can you describe your role and how AI shows up in your work?

[00:02:21] Omar: Yeah, yeah, David. Um, I, uh, am in platform, and part of our Acumatica platform also extends into all of our AI capabilities that we build. Um, and lately, more of my role has been heavily involved in bringing our AI features, functionalities, and capabilities into the market, working with our partners, our customers, prospects to understand how we can thoughtful designs of AI solutions and bring them to market to help business scenarios, business use cases, business needs.

Specifically, our new AI assistant, our AI automation, our MCP server that's coming up, uh, which I'm sure we'll be able to talk more pretty soon. But, uh, but, but most of my time has been spent working with all of these products. And on a day-to-day basis, um, I used AI all the time. It is my thinking partner.

It helps me understand the world in a different lens, multiple perspectives. So not only do I use it to help, uh, inform how I build my roadmap out But also I use it daily to help with triaging, drafting, tracking work, tracking progress, project management. So AI is in my life, uh, a lot more than it was just even six months ago.

[00:03:38] Doug: That's super cool. Um So, uh, next question is for you, Omar. So, how are growing businesses approaching AI adoption right now, and, and what are you hearing from, from our customers?

[00:03:49] Omar: Yeah. So one of the things that I have noticed is in the past, the question was more on should we use AI, can we use AI? And that, that kind of conversation, those kinds of questions have shifted into where does AI or where can AI save us time?

And then one of the things that I've also noticed with businesses is that some businesses may not have a dedicated AI team, whereas others do have a AI team or teams, and they're implementing AI from a very broad stroke, stroke perspective to understand where AI can fit so they can move fast. On the other hand, if you don't have the dedicated AI team, you need, uh, tools and capabilities that bu- are built into the systems that you're using because that AI is kind of a, the great leveler in those sense.

Uh, it helps those smaller companies and businesses to help scale a lot faster, only if you know how to design it correctly. And not that you'll, you'll get it wrong, but, uh, it helps you iterate a mu- a lot faster when you're trying to design how AI fits into your organization. And so with that, one of the things that I've also noticed is that there's a kind of a split that's emerging between these types of organizations, the small to medium-sized businesses specifically, where we have certain companies, certain organizations that are broadly experimenting with AI, and then others are going very, very deep into one to two very key workflows that are very critical to their, to their process and organization, and then just th- their standard procedure and trying to figure out how AI can help augment their day-to-day work.

[00:05:26] Doug: Yeah. And, and I know that we've, we've come a long way as it relates to AI, and I know you have a, a vision which we're gonna get into i-in a little bit about where, where that's gonna go. So, um, super, super awesome. All right, Doug, here's one for you. What separates the companies that are getting practical value from the ones that are still experimenting or kinda stuck in the pilot mode, or maybe not even thinking about it?

I, I don't even know if you run into customers who aren't even thinking about AI. A lot of time has gone by since I first approached customers about AI, let's say 18 months to set a timeframe. In the beginning, I was explaining what AI was. Let's say I've been on 80 to 100 calls having these same discussions.

If I started a call talking about how, what AI is today, people would be like, "Okay, Doug, move it along. We got to get to the good stuff," which is how we're actually gonna bring this to market and how we're gonna benefit from it and how it's gonna save us money and how it's gonna help us, uh, reduce our time to deliver for our customers.

So, the customers that seem to be doing the best are the ones that are thinking a little bit differently. So, think of AI as a tool that can combine... ERP and AI can work together. ERP has the real strong, I wanna call it algorithmic Pristine data that you can use. AI does really well with unstructured data.

Mm. The people that really benefit are the people that learn how to combine those two things into one category and use the unstructured data along with the data that is pristine and accurate inside Acumatica to generate results. And some examples of how that would be like after I've explained all the things AI can do, somebody will say, "Oh, great.

I can use that to add up the total amount of money I've made this year." And I'm like Good, but that's more algorithmic. Somebody who's a little more structured will say, "Wow, I can build a workflow that senses the sentiment of the support case and then automatically directs it to the appropriate person based on things it's seen in the past."

So those are the, the people that are thinking of both structured and unstructured data are doing a lot better than the people that are just thinking of one or the other. Yeah. Big, big, huge game changer, right? When you think about it from, from that perspective, the structured and unstructured data. Super awesome.

So continually, continuing along the, the adoption theme, Doug, as you've introduced new products, what have you learned from customers? How have those experiences shaped your approach, and how have you responded? So everything's a learning experience. At Acumatica, we launched something called GL Anomaly Detection in, uh...

It was quite a while ago. And we were shocked that no one was flocking to it to use it. And then we started asking our partners, you know, "Why aren't you using it?" And they said, "Well, we tried, but it was a little too confusing," and people didn't understand how to use it within the software, within the ERP software.

So basically, nobody adopted it. So we pivoted, and we learned, and we said, "We need to come up with a couple things that guide our development processes." And we call this basically our, our governance policy for innovation. It's responsible, practical, and valuable. So when the very first part, we've always been responsible.

We've always said, "Your data is your data," and we're not gonna give it to an AI to go, you know, spread it throughout the world because keeping your governance policy and your security policies in play is number one and first and foremost. The second one was the practical one. Practical means it's gotta be easy to use.

So we broke that rule on our first initiation into the, the world of AI delivery. Since then, we've developed things that you can enable with three or four clicks and you're off and running, so it makes it super easy. Our new anomaly detection, I... You can use literally three clicks and the system's off and predicting things.

With two or three more clicks, you can add it to a dashboard so that it's part of your daily routine. So that's a, a important part of was that, of that, was fixing the practical aspect of it. And then the last one, just to finish off how we do it, is valuable. If it's gonna cost, you know, $3 of tokens to perform something that's gonna give you $1 of value, it's obviously not gonna take foot.

But generally, it's not a problem because people are seeing the value in how these structured and unstructured data can come together to really drive value for their business. Yeah, I love that practical, right? I mean, if you really come at it from a practical perspective, then, then people just obviously adopt it, right?

They don't have to think about it or, you know, they just kind of see how it fits in and makes it a little bit less scary, I think. That's right. But, uh, awesome. Yeah. Cool. And the, and the valuable piece is just as important too. Having a- Oh, for sure ... pricing policy that's easy for them to understand and follow, it's gonna really help people adopt it as well.

Yeah, for sure. Yeah. If, if, if you're feeling like you're nickel and diming, you're afraid to experiment, then, then, then that's not good. So Yeah. Good. Cool. All right, Omar, where are you seeing the clearest business value delivered inside ERP today?

[00:10:35] Omar: Yeah. So, uh, kind of like what Doug mentioned on the, on the practicality side of things, uh, a lot of customers are also kind of saying they want AI to do the work, uh, not just answer questions.

So that's a big shift we are also seeing. At the same time, they wanna be in control of what gets acted on, um, on, on their data. So as a result, trust is a big gating factor in, in addition to the capability itself. So data privacy comes up sometimes, often at the same time as what the features are. And the value is coming to, uh, connecting AI to a process that they already are running.

So with that, some of the ways that we have addressed these business values and concerns that customers have brought up inside of our ERP is with, uh, real out-of-the-box agents. Mm-hmm. They are living in the product. Agents such as our AI-based, uh, case summary agent that will help you understand, uh, kind of like what Doug mentioned earlier, like the sentiment, uh, what the next steps based on the case would be.

Some kind of a suggested reply for the client or the customers. All of this would be AI-generated, but based on your real data, and none of that is necessarily leaving outside of the bounds of your Acumatica environment. Right. So you are able to trust, build that trust with AI as you hand on more and more responsibilities to these AI agents.

Similarly, we have case closure notes, which are once you have, uh, solved a case for a customer, being able to automatically create a closure note so you can reference it later. And then eventually you might be able to build your standard operati- operating procedures based on these case closure notes that these agents, uh, do for you.

All the way to kind of the standard that we have probably all, uh, almost all of us have used AI for is, "Help me write an email." Being able to, uh, draft email responses based on some priority, how the customer's back-and-forth interaction has been. Maybe it's a customer that needs some additional care, so, so treat your email with a little bit more care before you send it to customers.

Being an agent to be able to help you alongside what you're already doing is also building that trust. And these same patterns go across our other side of the products or the, the ERP today, where running, uh, on specific leads or opportunities, where sentiment and pattern recognition, being able to understand good pattern and summarize across chaotic data, like Doug mentioned earlier, is what AI is super, super helpful for.

So these high-volume, judgment-adjacent type work where it's easy to check but hard to maybe sometimes start from scratch are some of the areas that are the clearest kind of step one business value for what, uh, is delivered in ERP today.

[00:13:18] Doug: So for the, the people that are just getting into it, where trust, you know, like I, I know for me, when I first started using AI, you're almost afraid to push the Enter button because You know, am I putting confidential information out there?

The fact of the matter is our customers don't have to worry about that. It's everything is safeguarded Right.

[00:13:35] Omar: That's right. AI is often using foundational models, and these foundational models have been trained on generally available data. It's not necessarily being used to keep training further because there is not much more.

Now you give AI your business logic, right? Uh, sometimes often referred to as a semantic layer, so it knows generally how your business logic should perform, but none of that is being used to train AI. It's just used to augment AI's decision-making or reasoning that will help the user at the very end to go through with your process.

[00:14:08] Doug: Very cool. All right. Doug, can you walk us through a concrete example of how AI is helping a customer or user save time, improve accuracy, or make better decisions? Sure. Well, look, there's a couple things that I think of immediately. The first is we use our own product. So Omar was talking about opportunity and case summary notes and things along those lines.

We use that, so we're our own best case study in that case. We can take a situation where there's lots of inputs, emails back and forth, PDF documents consisting of RFPs, meeting recordings, everything else, and we can boil it down to, what do I gotta do next to help this customer out? And so rather than flipping through a million things, it's basically like having your own personal assistant there to say, "Doug, go do this next."

So that's been one thing, and obviously we have customers that are available to take advantage of that as well. Within the product, I see a lot of people using AP document recognition to take filing cabinets full of paper. Uh, we, we had, at one of our summits, we showed a picture of somebody that literally had a room full of paper that they were able to eliminate by doing AP document recognition, and the system can learn and make it, uh, the, continue to improve on how that works.

The thing that people are getting most excited about when I have all these meetings with them is the AI assistant. We now have the AI assistant that can answer questions about your data, and it can even put together stuff about your data. So if I asked it, you know, one of the things is, uh, I've seen people do is, you know, "Tell me what products are most profitable, and give me a list of the ones that I have in stock so I can go try and sell them," and it'll put all, pull all that together for them.

Or- Nice ... other ones is, you know, "Compare my financial outlook based on quarter one versus quarter two," and it will do that. But not only, it won't just produce a bunch of numbers, it'll go analyze them and say, you know, "I noticed that your, you know, expense reports came in very high in February," or in, you know, maybe you had a conference or something then, or maybe your supply expenses way, went way up in the end of June and you're like, "Well, that's 'cause everyone's trying to get their expenses and pay for stuff before the budget ended," so maybe there's a rational explanation.

But it comes and it surfaces all that stuff. So people are getting a lot of value out of that already today Yeah, huge productivity increase, right? You don't have to look at all these financial statements and kind of determine where that... What it's telling you. It can kind of highlight that and you can just go dig in.

That's right. Now some of our people are already experts at that. They can probably look at 25 spreadsheets at once and determine exactly what to do. But, uh- ... it's democratizing it. Now I can do that too. All right, Omar, so this episode really lines up with the 2026 R2 release. Um, what AI related functionality are you most excited about?

And why does it matter for our customers?

[00:17:00] Omar: Gosh, David, that's a very good question. I'm very excited to talk about it. I actually have been preparing for our upcoming release for 2026, R2, Acumatica's second release of the year. We have so much jam-packed into it. It was hard to kind of pick. It's hard to pick what to talk about.

But I think that the key areas that we have been pushing for the last two releases are areas that we're pushing even further because that's where we are starting to see our customers, our partners are finding a lot of value in. Our AI automation, for instance, it has evolved quite a bit. We used to call parts of AI automation LLM prompts, where you create a prompt and you, you go about putting some basic instructions to an LLM.

We are completely changing how that works. We are calling them agents now because it's, it's evolved and graduated beyond more than just a prompt. Um, you can do very, very intricate functions and chain together functionality. It works with customized data, so if you have anything customized in your product, you can work, uh, with that with your agents.

Agents can pull directly from generic inquiries as a source before it acts, which is one of the key ways Acumatica works and are how our customers and partners build, uh, their reporting, uh, gathering data. In line with that, things like our business events, which is a very robust mechanism in automating some key workflows in Acumatica, importing and exporting data workflows.

You're able to chain together agents to perform all these functions autonomously if you want to with no development work required. It's just plain, plain language. Uh, I was going to say plain English. It also works, uh, with non-English f- uh, uh, based- Right. Right ... uh, instructions, so I can't really use plain English anymore.

So plain language, uh, it would work just like that. So it empowers everybody, and using that same word that Doug used, it dem- democratizes that work f- across everybody. So that's just AI automation. We have a whole another leg, which is our AI assistant. This is the surface where a user interacts with AI.

We are bringing AI assistant into general availability. So, uh, in our past release, it was available as a managed availability or a limited release, but now it's gener- it's going to be generally available. AI assistant, if, if you haven't, uh, had a chance to see it, it cites its sources. So it is- Mm-hmm ... a very tuned AI assistant, so whenever it answers, it has a source, so you can trace back to the source to check and verify whether it-- you can trust that data.

And because it's built on your data, you can follow it back to where the, the, the source. We have incorporate-- We will be incorporating more out-of-the-box generic inquiries that is exposed to Assistant. So Assistant is armed with kno-more knowledge, more, more of this data to, to draw upon to answer a user's question.

And if it doesn't exist out of b-out of the box, you can create any GI, and then you expose it to Assistant, and then your assistant is ready to go a-to answer your question. And then we gave AI Assistant in this release some additional capabilities or superpowers, if you will. It knows about itself now, so it, it has some self-aware context.

So if you ask, "What can you do? What can't you do?" And it will answer, uh, what it can do and what are its boundaries. It is also context-aware, so if you're on a specific screen, you don't have to necessarily say, "You know what? Tell me the total of this invoice number XYZ." Instead, you can just say, "What is-- Can you summarize this invoice for me?"

And you just have to use these keywords, this, and it will understand that you're talking about something specific on your screen. And then in addition to that, we also included the capability, the, the help. We brought help agents so that AI Assistant, if you're asking, "How do I..." or "How do I do this? How do I do that?"

It will be able to draw upon, uh, your current version of Acumatica that you're in, and then look into our help articles, our knowledge-based articles that's available on our community site, any sort of how-to document, and be able to compile that. Again, those are the sources. And then provide you an answer.

So it will graduates to give you that answer, the help capability, and then you can trace back to the source just like anything else that AI Assistant, uh, provides answers on. And this is built upon our new beacon.acumatica.com, which is a new platform that allows you to do, do these help, help searches with help agents manually if you want, or using our AI Assistant.

And then finally, this is a, a really, really cool functionality for AI Assistant is when you're asking this question, kind of like what Doug mentioned, and AI Assistant returns back some kind of a chart, pie chart, bar chart, a KPI, or anything that i-is a visual, uh, summary of the answer. You can now drag and drop it directly into a dashboard without having to rebuild it or anything.

So in the context of a conversation with the AI assistant, you get an answer and you drag it if you want to continue to pay attention to that, that metric into a dashboard directly. So that's AI Assistant. And then one last thing, David, we are also introducing MCP server. We will be re-releasing the first phase of MCP.

It will work with generic inquiries. So what it allows you to do, and MCP is the, is the latest buzzword for the last few weeks, that it allows an external AI, uh, solution like Claude, ChatGPT, or any others that can support MCP to then connect to Acumatica. So now you're able to be outside of Acumatica and query questions about your data that resides in Acumatica, your system of truth.

So you are able to draw upon some patterns and understanding and knowledge of that. So phase one is generic inquiry. It's read-only, so you're able to draw upon that, uh, internal data. And then phase two, we will add some read and write capabilities so that not only are you interact-- you're able to interact with an external solution with your Acumatica data, but also now you can start thinking about building integrations without much development work because- Yeah

all of that is done by AI behind the scenes with you as the, the plain language, uh, instructor in that case. So it, it, it lowers the setup floor. The idea is to lower the setup floor to get going to, to lower that barrier to entry across all of these products that I mentioned, and so I'm super, super excited for the R2 release and what we have for AI capabilities and updates for our customers.

[00:23:29] Doug: I love it. I love it. So for those listeners who might not know, what does MCP stand for, Omar?

[00:23:34] Omar: MCP is, is short for Model Context Protocol. It's a standardized way that was, uh, provided or originally brought to market by Claude, and then very quickly adopted by other market leaders in AI. So it's a way for you to provide a functionality that lets another AI to be able to understand and communicate with a system So you say, "This is how you can communicate with Acumatica.

You send me this request and you get this response back." That's a very simple, uh, layman term explanation, but as you can imagine, it not only lets you have a ch- external chat interface work with your Acumatica data, but now you can build multiple integrations with whatever AI tool you choose. But it lets you kind of, uh, tunnel in and out of different systems that have MCP server capability.

[00:24:28] Doug: And all of that is just what you're excited about. But that means- Yeah ... there's a lot more other stuff in there too. So that's, that's super cool. I love it. Yeah. All right, Doug, two-part question for you. So building on some of the AI innovations Omar just highlighted, can you bring one of those to life with a day in the life example so listeners can picture how it shows up in real work?

And looking beyond a single scenario, uh, how do you see AI reshaping expectations for ERP over time from a system of record to a system that surfaces insights, guides actions, and automates more work? All right, a two-parter. So let's start with things in the day of the life. So one of the things that happens a lot is you have people in your company that build reports and other things, and there's probably, you know, two or three guys, maybe even more in a larger company, that are really kind of the bottleneck for that.

One of the things Omar described was now you can take dashboards that you build using natural language processing, I didn't say English, and you can then take that and populate it to a dashboard without the one or two of the guys that tends to be the bottleneck. So it's really going to make progress happen a lot faster in terms of getting the data out into the hands of other people properly and, and more efficiently.

So people who are real experts at their business process now have the ability to share that expertise and work they do with others inside the company. So it's gonna make things work a lot more smoothly in terms of company collaboration, which is really important because when everybody's using your ERP system, um, you know, if you have unlimited user pricing like we do, everybody uses it.

That allows everybody to collaborate together more effectively. So that's one of the things, uh, that I can think. There's hundreds of others that we've run into, and I get to work closely with Omar and the product team on some of these. So some of the people that originally looked at, uh, some of our automation tools said, "Well, I've created this sales order, but now I just remembered I have to go, you know, change the price on every individual line item."

So they could write a simple customization to do that, or they could, you know, tell the assistant, and not a prompt anymore, it's an agent to go ahead and update or, you know, add ten percent to every line on the sales order, and it will accomplish that. That was one of the pieces of feedback that, uh, we got from the initial launch where the practical part of it came in and now we're gonna, gonna launch that one.

So what does all this mean for ERP in general? So where ERP was very good at remembering things, presenting reports, following processes, doing if X then Y, ERP is now moving to a whole another capability, which is what's going to happen? What do you have to look out for in the future? What should you be doing as opposed to what are you doing?

And it's going to ground all those predictions, recommendations, and everything else in that pristine data I talked about before. If you're just using AI out in the wild and you don't have the pristine data, you can ask it questions about what you should do, and it will tell you to do things that may not be relevant to your business.

So it's a, a good tool if I, you know, want to write a for sale sign for, uh, my car because everybody does that the same way. But everybody's business operates a little bit differently, so getting that data and having you write a for sale sign about something you're selling at your company is much more useful if that data is involved in it and it makes it much more valuable, so it can predict what's going to happen.

When you're writing that for sale sign, just to continue that analogy, it might know that there's, uh, you know, a big rainstorm coming up, so maybe you don't want to sell, you know, the suntan lotion that day. Maybe you want to sell-- it'll build the for sale sign and say, "Sell umbrellas instead of suntan lotion."

So it can pull all that together and make predictions, not only based on the pristine data you've got, but on the data that's, uh, out there in the wild as well. So I'm looking forward to ERP, uh, getting together with this, uh, these LLMs to, to, to make all that happen. Nice. Nice. It gets back to what Omar said earlier about his own role, right?

I mean, th-this is really a business partner, right? Leveraging AI as, as a business partner. So that's, that's awesome. So this next question is for, is for both of you. And Doug, we'll start with you. Um, what do you think growing businesses often misunderstand about AI and ERP, and what would you tell a customer who's cautious about adopting these tools?

So being cautious is okay. Um, there's, you know, if you're letting everybody from your company go take your company data, download it, and put it into their personal copy of Claude, that's gonna be a big problem. So I would say get moving sooner than later. Get a corporate version that Omar talked about earlier that has all the guardrails and things in place, so your employees are using that instead of their own, their own stuff at home.

That, that would be number one. But get your governance policy and everything put together first so you know where you're going and what you're trying to protect and what your key, key pieces of data are. And then, of course, if needed, start with a sandbox and some of our, you know, demonstration data that we'll give you, and you can play with that if you, uh, if you have any concerns beyond what I just explained.

Omar, how about you?

[00:30:00] Omar: Yeah. I think, uh, some of the key patterns that I'm noticing, uh, one of the biggest misconception is AI just means chatbot, uh, bolted on top. You know, because it's so pervasive, it's such a, a, a common language that we have come to know, but it's, it's beyond-- It's, it's much, much m- more beyond that.

The real value comes in these agents doing the work. Like, like Doug said, you have these pristine data, you have these automations in place. Now you sprinkle AI magic on top, and then you start to see something super valuable that you didn't expect before, and that's how AI should be working for you. It should be working inside of a process that you're already running.

The other thing is assuming or not necessarily understanding how data is being used, how data is being trained. Again, it goes back to our understanding of these common tools that are out there, which may not be the same as enterprise ERP, right? They're two different ways of functioning, and so it's important to question when you're sitting in front of a, a demo, how is that data used, if it is used, and how will that affect me?

So that, that's another key area to, to understand and think about. It's, is, is it's distinct on our commercial use of something like Claude- Versus our corporate or enterprise level use of Claude are very different. So understanding that is important. What does foundational model mean? What is a model that's being trained on mean, right?

All those things are good to understand. And then one other thing I wanna mention is I think that what I'm noticing, uh, this misconception around the fact that if I need, if I need to incorporate AI, it's a huge, it's a big build. It's a big time commitment, which it, it may be depending on your situation, but for the most part it's not.

So there is this, uh, uh, there-- Most people are underestimating how usable it is and how you can get from zero to one quicker than you would think. So I think experiment, start with one process, not everything at once. Pick something that you do, uh, at a, at a high frequency, but relatively lo- low risk so you can start to build some comfort and trust around that.

And something like, you know, financial or higher stake risk, go slow, crawl, walk, run. And then you don't want to necessarily wait too long to get there, but you wanna iterate fast and, and fail fast in, in short, short experiments as you build out a bit bigger, better, stronger process, uh, using some automation and using some AI to help you with what you need.

[00:32:26] Doug: Great. Great advice from both of you. Thank you. So Amar, we've been talking about what's coming in the 26R2 release. Looking a bit further out, where do you see Acumatica taking AI over the next few years?

[00:32:39] Omar: Yeah, good question, David. Um, so m- uh, earlier I mentioned that crawl, walk, run approach. Acumatica, we are thinking of these, these, this agentic AI platform where eventually there is going to be capability that work is done autonomously to a degree to-- with the guardrails that you put in place for the AI for you.

And so the way we think about it is across that spectrum of advice, an AI that can assist, an AI that can automate, and then an AI that can orchestrate. So these four kind of buckets or areas of the spectrum, so to speak. So you might have an AI functionality that just provides you basic advice on this is what pattern or summarization was gathered based on this data.

You might have assist, where now AI is providing you with assistance, suggesting actions, and maybe the person maybe is carrying it out, the user is carrying out the function. Then you have levels of automation, like I was super excited about talking earlier, where you can start chaining together existing automation, uh, into a certain, uh, capability or from one to, uh, one to 10, follow steps one, two, three, four, and et cetera.

And then you have the level of AI that can orchestrate functions. Mm-hmm. So it is in the moment when a user asks a- AI for a certain task or something to be done, AI now will be able to provide, uh, or collect together these different components and orchestrate, just like in an orchestra or a conductor in an orchestra.

They're able to orchestrate the tasks that's needed at that time to fulfill the user's need. So i- it's a more of a spectrum versus an end goal. Uh- Mm-hmm ... and different AI capabilities should fall into different spectrum. And you don't need one AI to, to rule them all, so to speak, but you want to have carefully thought-through AI capabilities across the spectrum.

So that's how we are thinking of our AI strategy, and it will expand. You know, today we have, uh, y- we have the traditional machine learning products, uh, such as our anomaly detection, document recognition, cross-sales. So we have had that in the traditional m- machine learning. Now we're in the generative AI phase where we are using AI assistant, AI automation, and MCP.

And then eventually we will evolve more beyond that as the, as, as we learn and understand where the market is going. So I think this, uh, framework that we have put in place today is going to help us achieve these, these next few phases and then we will as we learn more about direction where the world is going with, with AI.

Yeah.

[00:35:07] Doug: Super, super awesome vision and hugely impactful on productivity and powerful for, for our customers. I love that. All right. So if you could leave customers with one takeaway about AI and Acumatica right now What would it be? Omar, you go first.

[00:35:21] Omar: Sure. I think it's important to know that we are thinking, that Acumatica is thinking about AI not as a standalone product or something that you're kind of configuring to fit inside of the ERP, but more on building on top of foundations that Acumatica made.

What Acumatica is today, we have an open platform, we have a very strong configurable platform, and we want to build AI thoughtfully that can leverage so that the sum is, is much larger the c- of the individual parts, that you have a very compounding effect, almost like an exponential effect when you put these pieces together and when you make these connections.

So a business event triggering an action, and that action updating a record which is triggering another business event. Or something in a workflow that you chain together across different modules to fulfill the need that you may not have been able to maybe think about or be able to do very quickly before, but now it's becoming much more fast.

Uh, so the value that you get out of Acumatica and the AI isn't necessarily additive, but it's, it, it compounds. So the more of it you turn on, the more it's worth to you. All

[00:36:35] Doug: right. Cool. Doug? Well, Omar was lucky 'cause he got to go first and say what I was gonna say. Right. So I'll just add a couple, uh, points to that.

We're a platform company, and that's kinda what Omar was saying. We have a single code base, unlike a lot of other companies, with a universal security model, so that's really going to help people in the long term utilize Acumatica to interface with these things. The other thing that's really important is our, the strength of our APIs and web services, um, that we built from the ground up, uh, 'cause we were originally designed for partners to extend our core solution.

So all that's gonna work right into the hands of integrating with these new and exciting platforms that are powered by LLMs and artificial intelligence. Nice. Awesome. All right, so you both ready for our lightning round?

[00:37:24] Omar: Go for it.

[00:37:25] Doug: Okay, here we go. Omar, we're gonna start with you. Okay. What's one productivity hack or tool you swear by to stay on top of your day?

[00:37:32] Omar: There are lots of productivity hacks, but I think, uh, like I mentioned earlier, for me, I'm a Claude guy, at least for now, until I change my mind, uh, which can happen at any time. But I am, uh, deep in Claude code every day. It-- Claude works with my notes, it works with my meetings, it works with my tasks, and if I mention something in one meeting and mention something similar in another meeting, synthesizing that, giving me a daily summary, giving m- me a month- weekly, monthly, and even y- close to a yearly summary now that I've been using it for almost a year.

Uh, so it's very important for that. And, um, a, a lot of things are being captured automatically for me. And so while I do that, I am in a, in a very fortunate position. I feel myself very lucky to be able to use AI. And while I use it, it often lets me kinda think about when I'm designing roadmaps or, or functionality or its little nuances, uh, to, to think about those nuances that I face that likely others, like our customers or users, are also facing.

So it always just, uh, makes me-- Uh, it triggers in my head on how can I fix this solution here? And then maybe this can map into something like a, a user's experience or capability or feature functionality or enhancement on our AI side of things on our actual product. So it's a very interesting relationship that, that's built on what I do on a day-to-day basis and as well as, uh, um, uh, what we can extend to our customers.

I know that was, uh, that was a little long for a lightning round, but I wanted to make sure I-

[00:39:01] Doug: Oh, good. Good, good information. Yeah. All right. Doug, we're gonna turn to you for some career advice. Uh, one piece of advice would you give someone just starting out in product management? Learn, learn, and learn. AI is a great new tool, so learn about it.

Don't give up learning. Keep learning as AI evolves. It, it's constantly evolving. And the reason you're learning is because AI will make smart people smarter- Mm ... and do I dare say, it'll make dumb people dumber if you learn to trust everything it says without having learned something first. Right. Right.

Awesome. Good. All right, Doug, what's an app or piece of technology that you love that has nothing to do with work? I'm gonna go with, uh, old school stuff. So I got these, uh, this model train set I put together and some Arduino chips, and I struggle trying to get the Arduino chips to do stuff on my train layout.

So I'm gonna move AI to help me program these Arduino chips, so I have a high-tech train display running my grandmother's trains from 1918. So maybe we'll see if I can make that happen- Wow ... in the next, uh, in the next year or so. Wow. That's super cool. Very cool. Omar, how about you? Or super geeky. Uh, you, David, you be careful.

Well, I don't know. I'm impressed. Don't, don't slip, don't slip down with me here. I'm feeling inferior right now, Doug. Omar?

[00:40:18] Omar: I think for me, I don't, I don't trust my, uh, memory or my brain, uh, very much. Uh, and I'm a tinkerer all the time, so what- whatever Doug had to say was very interesting. I would love to know more.

But for me, I think, uh, having, having three small children and just a busy life, as much as I can automate in my personal life, sometimes to a detriment, uh, if you ask my wife, I automate everything, whether it's, uh, setting up the laundry and using a little NFC chip to tap my phone and then immediately set a timer so I don't have to remember to change the laundry over, to turning the lights off when the sun goes down, and, uh, uh- Wow

anything and everything that I can do at home to save time. I really, I say it to myself to save time, but it actually ends up taking more time to build it out. But it when, when it works, it works like magic.

[00:41:06] Doug: Kinda cool, right? Kinda cool. Yeah. Oh, man. Awesome. Well, that brings us to the end of our episode.

Doug and Omar, thanks so much for joining us today. Super fun being with you. Appreciate you sharing all the knowledge and sharing practical insights and real-world perspectives that helps our audience take this on their AI journeys, uh, into the future. So thanks so much. Appreciate you being here.

Pleasure to be here. Thank you, David.

[00:41:28] Omar: Thanks for listening to the Acumatica ERP podcast.

[00:41:31] Doug: If you found this episode useful, please subscribe to the show on YouTube, Spotify, Apple Podcasts, or wherever you listen so you can catch every new episode. And if you've got a minute, leave us a rating or a review. It really helps other listeners find the show.

You can find every episode and more information at acumatica.com/erp-podcast. Thanks again for tuning in. We'll see you next time