GUEST COLUMN by
There are times when current events trump the relentless progress of innovation, causing us to reassess what software could actually do to help companies and countries respond to global economic shocks.
That’s why I came to Las Vegas last week for Epicor Insights 2025. I have heard from plenty of software platform and cloud infrastructure players who are all-in on artificial intelligence, but I was curious about how AI might have an impact on global supply chains — which encompass pretty much any good the world makes, moves or sells.
Indeed, despite all the recent talk about tariffs somehow bringing manufacturing back, the real problem to solve here isn’t a lack of jobs. “In the U.S. we see about 5 million jobs unfilled in the supply chain, and 25% of plants reported they were underperforming due to a skilled labor shortage,” said Vaibhav Vohra, president of Epicor Software Corp.
On top of that, many manufacturers are facing input shortages of key raw materials, from the rare earth minerals needed to make batteries to the lumber required to address a housing shortage. Finding alternative sources at their roots won’t happen overnight. “Only 21% of companies have visibility beyond their Tier 1 suppliers,” Vohra said.
“You’re doing well adjusting to an exceedingly difficult time,” Epicor Chief Executive Steve Murphy (pictured) told the audience of 4,000 manufacturing and logistics leaders in his opening keynote. “The make, move and sell economy is more mission-critical than ever. With our adoption of AI, we believe in augmentation, not replacement, of front-line workers. It’s about giving them the tools they need to succeed.”
Here’s a few pragmatic approaches to leveraging AI I learned about at the event that might mitigate the perceived risks of working on the bleeding edge of technology:
In talking to some of the customers here, many of whom would be considered midsized manufacturers or distributors (or, in the $500 million to $2 billion range), I learned they are mostly using AI features already, and they find it very productive. Sounds useful, but not as exciting as I expected.
This is because Epicor takes a stance of “AI-forward” development. Instead of developing a big-bang solution in the office that covers universal issues and sending out service teams to make it fit the customer’s environment, the delivery teams seek out a problem area within the customer’s process, and model a very context-specific AI that solves only that problem. AI solution first, product rationalization later.
“The problems that AI can solve are real, but it’s our job to stay tied to the problems, not to technology, said Arturo Buzzalino, Epicor’s chief innovation officer. “The beauty of ERP is that we run the core of these businesses. So we get direct exposure to the challenges that a business faces in the supply chain.”
Interestingly, the field development team doesn’t worry about making one-off tools with AI when on site. “We’ll go really deep,” Buzzalino said. “But once we find a problem that’s successful and is generating value, we will then bring that back into the platform and generalize it.”
The firm already covers many of their customers’ core supply chain processes and supports hundreds of discrete industry use cases with specialized AI inference engines and agents running on its Epicor Prism fabric, one of several AI product infusions announced at the conference.
When I worked at the high-energy supply chain software company i2 around 25 years ago, data engineers would run an ETL (extract, transform, load) batch overnight and load up the beefiest Compaq server with demand and supply signals from enterprise resource planning and manufacturing resource planning to feed “genetic algorithms” that could run in-memory and churn out resulting plans, sometimes in hours.
Since then, Moore’s Law continued to drive higher expectations as the cost of compute and storage power decreased, and bandwidth and data capacity increased year over year. Now the industry needs predictive models with far richer data sources that will show them market shifts so they can react in minutes or seconds, not hours.
There’s a lot more to getting that much data in the pipeline to properly train planning and analytics solutions, much less feed a host of AI agents. Epicor’s 2022 acquisition of Grow and broad integration sets give them an automated way to source and route data using elastic cloud resources or private infrastructure.
“Our LLM Inference Pipeline in Prism is model-agnostic,” said Kerrie Jordan, group vice president of product management at Epicor. “So we’re able to feed training data into OpenAI, Anthropic, Llama, whatever model makes the most sense to answer the query. There are so many variables. It’s incredibly difficult to predict what will happen. However, it’s more achievable to build flexibility into your business, to make fast decisions based on data. You will still need to have modern technology in place, access to rich data, and the tools to process all that information and make that decision faster than your competition.”
Using ERP as a cognitive hub for predictive AI makes sense. You already have current order signals and years of well-tagged order data within ERP, plus connections to millions of records of product and supplier data in catalogs and product information management or PIM systems from Epicor and other software providers.
Machine learning that surveys the data estate of an org and its suppliers offers a unique opportunity to safely train AI models that will only use each customer’s own data for better context.
The Covid pandemic taught leaders a lesson about having a solid digital backbone. Most businesses could no longer interact with customers, suppliers or even their own employees in person. Global supply chains were disrupted, and companies that were unable to adapt their digital processes to accommodate remote work and collaboration went by the wayside.
“If there was any silver lining of Covid, it was a stress test for the supply chain,” said Buzzalino. “A lot of companies that survived had already put in place plan B, and those that thrived had plans C, D and E ready to go. What happened in 2019 means the supply chain is better prepared today.”
Take, for instance, rare earth minerals such as neodymium and yttrium that go into magnets and catalytic converters. China extracts, produces and refines as much as 70% of these raw materials. If the supply is affected by tariffs or lack of a trade deal, manufacturers of robots, smartphones and cars will need to develop alternative sources, processing partners and part suppliers — and it’s not like the minerals are sitting around on a map, ready to collect like resources in a video game.
To answer that, Epicor builds and invests in content, such as the catalog information in their recent KYKLO acquisition, and by implementing AI routines for data movement (leveraging the Grow BI solution) to attach current regulatory and tariff rules to multinational supplier data.
“All of this rich catalog and supplier data just becomes more training for the AI,” said John Carrico, vice president of product management at Epicor. “So if they’re doing data migrations and they have duplicates, we can just say there’s a system of record that we know is accurate, and replace it with this content.”
Doing this right requires much more than just passing costs up the chain. “We can introduce new features into our continuous delivery pipeline to help customers stay flexible and account for the new changes that they’re seeing with tariffs as well,” Carrico said.
Epicor introduced new Harmonized Tariff Schedule or HTS functionality into its Prophet 21 solution for distributors, and Incoterms in its Kinetic package for manufacturers as well as international valuation and compliance features for shippers.
If you own a brand that depends upon a supply chain, you actually need to know everything about the behavior of your suppliers, as they operate within their own local and regional trade regulatory environments.
One example manufacturer was experiencing costly problems handling environmental regulations in sourcing throughout their supply chain, especially for medium-sized suppliers that don’t provide solid information, which delayed onboarding, price quoting and order-to-promise times.
Epicor announced a new carbon capture solution at the event, which treats CO2 emissions as a currency and rolls their costs up into financial systems and regional compliance reporting tools.
It also recently acquired Acadia, a connected platform with an auto-generated supplier intake process, to help qualify and upgrade suppliers based on their level of sustainability and data standardization.
“Tariffs and regulations make it even more important to be as modern as possible, because you have to be very nimble,” said Buzzalino. “AI is going to make everything dynamic. Physical AI is coming on very quickly too, using advanced manufacturing and engineering with IoT and robots sending signals back into our systems from production lines.”
The expo floor was really more of an educational hub for Epicor’s broad portfolio of solutions than a typical tradeshow, with connected devices, demo stations and support engineers standing by to help customers figure out solutions to their unique problems.
Of particular note was a cool demonstration of how a product entry in a catalog could be rendered out as a 3-D model within their CPQ, or configure, price and quote, solution, complete with materials specifications.
While some tech conferences might put on an AI “guru” to get everyone’s mind up in the clouds, I found the down-to-earth nature of this show refreshing, and it was fun to return to my supply chain roots. These attendees aren’t software engineers, they’re people who operate factories and transport and make the goods that power local and global economies.
Interestingly, most of these firms are already finding value using AI to enable agile responsiveness to real-world change, to a much greater degree than I could say for most digital-first outfits that are often just experimenting. The agentic AI of Epicor’s Prism functions sort of like an expert curator of machine-learning-prepared enterprise data, and sort of like a concierge who always knows what is happening in the customer’s environment.
As it turns out, there’s no need for anxiety about adopting the latest AI technology, if the adoption seems like a natural extension of everyday business, empowers workers and helps leaders make better decisions atop a wealth of industry-specific knowledge.
Jason English is principal analyst and chief marketing officer at Intellyx. He wrote this article for SiliconANGLE. At the time of writing, Epicor is an Intellyx customer and covered the analyst’s attendance cost for the event, a standard industry practice. No other companies mentioned are Intellyx customers. ©2025 Intellyx B.V.
THANK YOU
Absorbing supply chain shocks with AI at Epicor Insights 2025
Perplexity reportedly near $500M funding round at $14B valuation
What to expect at Red Hat Summit: Join theCUBE May 19–21
Saudi Arabia launches multibillion-dollar AI venture
IBM Think exclusive coverage: Join theCUBE May 15
Salesforce Agentforce and Data Cloud: A path to the software-only hyperscaler
Absorbing supply chain shocks with AI at Epicor Insights 2025
AI – BY . 1 MIN AGO
Perplexity reportedly near $500M funding round at $14B valuation
AI – BY . 16 MINS AGO
What to expect at Red Hat Summit: Join theCUBE May 19–21
CLOUD – BY . 2 HOURS AGO
Saudi Arabia launches multibillion-dollar AI venture
AI – BY . 2 HOURS AGO
IBM Think exclusive coverage: Join theCUBE May 15
AI – BY . 4 HOURS AGO
Salesforce Agentforce and Data Cloud: A path to the software-only hyperscaler
AI – BY . 5 HOURS AGO
Forgot Password?
Like Free Content? Subscribe to follow.
Leave a Comment