Enterprise resource planning systems, which help integrate and manage distinct business operations in one piece of software, started becoming popular in the late 1990s. Today, their role is still evolving as artificial intelligence gains traction.
A recent whitepaper from manufacturing data platform Caddi found that 69% of surveyed organizations planned to invest in physical assets. Only 33% planned investments in operational software, such as ERPs and manufacturing execution systems MES and ERP, which was 60% less than last year.
Manufacturers currently seem to be prioritizing investments that have an immediate, measurable effect on factory output. This creates a significant problem when it comes to integration with AI, experts said.
“AI raises the bar on the data foundation rather than replacing the ERP completely,” said Srinivas Chippagiri, senior member of technical staff at Tableau, a Salesforce company. “Manufacturers with clean, well-integrated ERP data can layer AI on top and get real leverage, while those with fragmented data across aging systems mostly get confident wrong answers faster.”
Cache Merrill, founder and chief technology officer of Zibtek, agreed. Zibtek works with manufacturing companies on ERP modernization, custom software and AI initiatives. He said the company’s customers aren’t choosing between ERP and AI. Instead, they are starting to use AI around the ERP systems they already depend on.
“There’s a lot of investment in older systems and data, so it can be difficult to justify replacing everything just to get access to newer AI features,” he said. “In many cases, it makes more sense to modernize the parts of the ERP that are creating problems, improve the way data moves between systems, and then introduce AI or automation where there is a clear business case.”
The best use case, according to Merrill, is when AI agents can handle the work that normally requires someone to move between several systems or reports.
For example, to accomplish a simple task like inventory analysis, someone would have to look at current stock, recent orders, purchasing information, supplier data and expected demand before deciding what needs attention. Merrill said an agent could pull that information together, identify patterns and prepare the next step instead of someone doing all of that manually.
“The manufacturers that get the most from this will be the ones that work on both sides of the problem: modernizing the ERP and data where it needs attention, while using AI and automation to remove unnecessary manual work around it,” Merrill said.
But all of this can be difficult to implement, as the technology that these manufacturing companies are working with is simply not designed for such advanced processes.
“I believe legacy ERP is the single biggest thing holding manufacturers back from significant efficiency gains with AI,” Remi Beges, CTO and co-founder of Bonx, an AI-native manufacturing ERP company, said via email. “Most manufacturing companies today are still using a legacy ERP that is not built at its core for AI.”
Beges said these systems were designed around one goal: to record transactions reliably. In other words, they were designed as systems of record, not systems that could take action.
But Beges said that today, in theory, AI could not only identify which systems are affected by a late shipment, but also draft and send emails to customers. For example, it could tell them their order will be delayed and plan accordingly, notifying workers of updated processes and schedules.
However, he said that this is nearly impossible on a system designed only to record transactions.
Specialized AI ERP programs and potential pitfalls
Some companies like Microsoft are launching specialized programs like Dynamics 365, which is an agentic ERP system for manufacturing, connecting demand, supply, production, fulfillment and cost data so AI agents can execute decisions rather than merely answer questions and generate recommendations.
Noting this development, Beges cautioned against handing over tasks to AI completely, as it could take undesirable actions without a human in the loop.
“A system taking actions autonomously is not dangerous in and of itself, and there might be situations where this is perfectly appropriate,” Beges said.
For example, if the ERP sees that a material will be out of stock for an upcoming run based on planned usage, the company might want the system to determine which supplier would be best to fill the order and draft a purchase order for the buyer to review. However, it might not want that PO to be sent without any human checks.
“The key is defining those guardrails upfront (agreeing on how you do and don’t want the system to act) as well as ensuring that the system is smart enough to highlight any exceptions or outliers proactively,” Beges said.
Despite the overall industry’s cautious approach amidst the risks and implementation challenges, some manufacturers have made major strides in this sector.
Clorox announced a transition to a new ERP system in 2025, going from manual spreadsheets to a “fairly automated” process. The shift is part of their $500 million digitization plan, which began in 2021.
Manufacturing giants like Nestlé have also announced the implementation of an AI-focused ERP systems. The company shifted its entire ERP process from data centers to cloud in 2022.
Schreiber Foods recently announced a partnership with Ascendion to modernize its ERP systems by deploying AI agents. The changes will be introduced across more than 40 locations on five continents.
Still, experts caution these shifts should be done carefully.
“The biggest pitfall is trusting probabilistic output in a system people treat as authoritative,” Chippagiri said. “It carries risks like getting wrong answers built on stale or inconsistent data, which are worse than no answer because people act on them.”
It also creates accountability gaps, he said. “When an AI-driven action turns out wrong, who owns it?”
