Epicor leads the way on embedded Agentic AI

Late last year, Enterprise Times spoke to Arturo Buzzalino, Chief Innovation Officer at Epicor. Buzzalino started his career as a data scientist and spent over a decade at SAP. He now runs the R&D and design teams worldwide, and is also responsible for manufacturing, cloud, data and product management.

Buzzalino explained how he works closely with customers, noting, “My teams forward deploy on customer sites. We work backwards from real problems that manufacturers are facing in their day-to-day and hopefully solve them or try to streamline some of these processes, using software around the ERP.”

Buzzalino reports to Vaibhav Vohra, Chief Product Officer at Epicor. While his responsibility seems similar to that of a Chief Product Officer, he is more focused on what he calls the “zero to one problems”, or innovation, in which AI plays a massive role.

Predictions

At the start of the year, every vendor considers predictions for the year ahead. I asked Buzzalino to look back at 2025 as well as forward.

For Buzzalino, 2025 was about experimentation and turning those experiments into production-ready technology. One example is the conversational capabilities it launched, which Buzzalino claims was the first in the world. The other was one of the first ERP agents, Prism. This is before the hype around Agentforce, according to Buzzalino, and he is bullish about what is coming.

He concluded, “The pace of innovation has been astounding. It surprised even me how fast we were going to see things in production. For 2026, that’s only accelerating. Everybody’s hitting the gas.”

On the importance of AI

With Prism, Epicor has embedded Agentic AI agents within its flagship ERP solution, Kinetic. Buzzalino believes that today, if you don’t have a clear AI strategy or know how it will be integrated into the fabric of your enterprise software, you can’t win a deal. But are manufacturers who were previously reluctant to migrate to the cloud, now adopting these technologies?

Essentially, while 18 months ago, customers were thinking about AI. Every customer is now thinking about what they can do with AI and how they can drive efficiency. It’s a broad topic because returns are high when there are many repetitive, manual tasks.

Several customers have either bought or are using Prism, and Epicor is getting increasing demand for additional features.

The first trend in Manufacturing

Intelligence at the edge is one of several trends that Buzzalino has written about. I asked him to expand on this trend. This includes technologies such as IoT. For example, sensors that transmit information, such as running time and temperature. These help with predictive maintenance. Buzzalino is looking further into the future and believes that it will have a huge impact on manufacturing.

He commented, “I’m talking around physical AI. We’re in very early innings of that, having physical AI running inside of factories, whether it’s robotics, computer vision, quality control, or other physical AI applications. What I mean by physical AI is, for example, if you roll a ball off this table, you know that on the other side, the ball is going to hit the ground.

“A computer doesn’t know that. When it sees it, the ball is gone, so how is it going to learn that? Well, researchers and billions and billions are being poured into solving that problem. That’s the next frontier.

When I talk about AI at the edge, where we are going to see the revolution is with that physical AI, at the physical level, literally moving real-world objects.

Real use cases for Agentic AI in Manufacturing

Buzzalino sees most Agentic AI being used in the back office. He cited two examples. The first is automating purchase order processes.

The second can also be used on the shop floor. He explained that it might be somebody on the shop floor, with a very specific question, “Where did I ship that item last week?” Historically, they might have raised a ticket with IT because they did not have the time to look for the answer in their ERP system.

Agentic AI understands the data in the ERP. It can quickly provide answers to those types of questions. Saving considerable time for many people within the organisation.

He described another, even more complex use case for Agentic AI, managing suppliers. “I need to buy this last-minute thing, I have my list of suppliers, and I might have to call them. I have to send some emails to figure out the lead time and the cost for this part that I really, really need. Many manufacturers today do that manually. They might have a team of five or six people doing that.”

Agentic AI can provide a solution to this, Buzzalino continued, “If you have a large language model that understands all of your data inside of your ERP, it can carry on a normal conversation.”

If it has that ERP in context, it can conduct those communications with suppliers. Then summarise the information in a single dashboard where a human can review, selecting the supplier and item they want. This is just one of the solutions that Epicor has launched using Agentic AI, according to Buzzalino. There are a further ten solutions that it should be launching within a year, he added.

Digital Assistants or autonomous digital workers?

The key for Buzzalino is that these are digital assistants. He strongly believes that the human should be kept in the loop. He expanded on the Epicor view, adding:

“What we say is we take the repetitive and the mundane so that people can focus on the creative and the strategic. Our focus with Prism is on how we use agents to take the repetitive tasks off the human. But, the strategic decisions, will always have to come back to the human. For us, that’s a non-negotiable part of the experience that we’re building.”

Buzzalino does not see Prism delivering completely autonomous agents; they are doing low-value, high-cost tasks that are repeated a lot within the business. He then added, “The efficiency gains that we’ll see from Agentic AI are going to turn into innovation acceleration for all of these companies, because they’re going to have more time, and I think people are going to do more of what they’re good at, not less.”

The cultural shift from experimental to instinctive

Another trend Buzzalino talks about is the cultural shift from experimental to instinctive. He explained experimentation is relatively easy.

He noted, “It’s one thing when you’re running POCs with AI, it’s really easy to build a prototype and show it a really cool video, it’s taking action inside of a system in an automated way. But there’s a big difference between doing that in a POC environment and that agent seeing scale, inside of a live ERP at a manufacturer.”

To scale up, manufacturers need to understand the metrics they should use, how to evaluate value, and, if they don’t, why not. When rolling out the first AI Agent, you will learn things about the technology and your organisation.

As more agents are rolled out, the learning curve is shorter, until the process becomes instinctive, and innovation becomes instinctive for the manufacturer as they become familiar with a pace of change that is unfamiliar today.

Who is leading the way?

I asked Buzzalino which manufacturers are leading the way with agentic AI deployments. Buzzalino is not seeing a specific manufacturing sector adopt AI faster. He believes it has more to do with the organisational culture. It is also those customers who are willing to co-innovate.

He added, “The customers that are adopting the most are the ones that are: a) manage their data very well, and they have their operation in order. b) They also have a culture of risk taking within the organisation, relatively speaking.”

Buzzalino shared that many manufacturers are risk-averse. One of the first manufacturers that they co-innovated with was a Formula One team. However, things are changing, and Epicor is now innovating with a cabinet manufacturer.

Buzzalino explained, “The pattern is actually the mentality, the ability to say, we’re going to experiment, and we’re going to fail, and that’s okay, because we’re going to learn and then we’re going to apply it to the next one. I think it’s more of that culture than an industry (sector).”

Does Geography make a difference?

What about geographic differences? Epicor has seen the biggest growth of Agentic AI in the US and the Americas. He explained that Epicor is going into Europe in early 2026 as there have been some hurdles. Regulation exists in Europe, with the EU AI Act, GDPR and other measures; in the US, it’s a bit less regulated.

However, Buzzalino added, “I think EMEA will probably have the biggest jump in AI usage, probably in 2026. I see it as really accelerating in Europe.”

On Data

The industry has been talking about data and data quality for years. Buzzalino mentioned that the leaders’ data is in a better state. What do companies need to do to improve the state of their data?

Buzzalino answered, “The healthiest thing an organisation can do, regardless of what you’re going to do with AI, is go through an ERP implementation. When you go through one, you have to understand almost all of your processes and essentially create a digital form of your internal operations. Doing that forces you to cleanse your processes and data.”

“The second is that, historically, enterprise software in general takes longer to implement. Now, it could be 40 days or 18 months. With ERP, for example, it can take years depending on how complex the implementation is. What we’re seeing and likely to see in the next few years is that the time to value or time to implement is essentially going to collapse. Now, is it going to collapse to zero? We don’t know, but it will reduce.

“So those 18 months will turn into 12 months, which will turn into 6 months, which will turn into weeks, and eventually, maybe days. Once that happens, it will empower people to implement new software very quickly. It puts pressure on companies like us. So it’s essentially bringing a consumer-grade experience to the enterprise, and I think we’re going to see that accelerate.

“The third piece of this is to start specific and small. You might have a huge operation, or your data might be messy, and you know it. That’s why we stress the importance of solving a very specific use case where, yes, maybe I can’t get all my data in order, but for this one workflow, we will get that data in the right format so we can automate this process.

“That’s our philosophy, and we’ve seen it be very successful with our customers; a lot of them have gone to production. We were the first in the world with many of these innovations in ERP.”

Technologies to watch out for

In closing, I asked Buzzalino which two technologies will rise in importance for manufacturers in 2026, and which two will lag?

He answered, “Because of the improved time to value migrating to the cloud, we’re going to see an acceleration of cloud adoption in the supply chain, even more than we have in the past 10 years. The other one is agentic adoption in the supply chain.”

Buzzalino believes that physical AI, in the form of robots and humanoid robots, will be slower to reach the mainstream. There will be experimentation, and it has the potential to drive significant differences, but he believes it’s at least two years away.

The post Epicor leads the way on embedded Agentic AI appeared first on Enterprise Times.