, ,

Manufacturers Fund Machines Over ERP as AI Raises the Data Bar

Jessica Nicole
Manufacturers Fund Machines Over ERP as AI Raises the Data Bar

Manufacturing Dive published today on how AI is rewriting the job of manufacturing ERP—and why capital budgets still trail the pitch decks. A recent Caddi survey found 69% of manufacturers plan to invest in physical assets, while only 33% plan spend on operational software such as ERP and MES—about 60% less software intent than a year earlier. Shops are buying what moves output this quarter. AI still needs the books underneath.

Srinivas Chippagiri of Tableau put the trap plainly: manufacturers with clean, integrated ERP data can layer AI and get leverage; those with fragmented aging systems mostly get “confident wrong answers faster.” Cache Merrill of Zibtek said customers are not choosing ERP versus AI—they are wiring agents around the ERP they already paid for, modernizing the painful seams first, then automating the work that normally means hopping five reports for one inventory call.

Modern glass-walled office corridor with black pendant lights and a polished concrete floor. Photo: Unsplash.
Business systems only help when the office and the plant share one memory—not parallel spreadsheets. Photo: Unsplash (rights-safe).

Remi Beges of Bonx called legacy ERP the single biggest brake on AI gains: those platforms were built to record transactions, not to act. In theory an agent could spot a late shipment, draft customer notices, and restage schedules; in practice a record-only core cannot. Microsoft’s Dynamics 365 agentic manufacturing path is pushing agents that execute across demand, supply, production, and cost—but Beges’s guardrail still applies: draft the purchase order for a human buyer, do not fire it blind, and surface exceptions before they become chargebacks.

The operators are already moving. Schreiber Foods and Ascendion announced a multi-year agentic-AI partnership on August 5 to modernize ERP, speed food-safety signals, and support plant-floor decisions across more than 40 locations on five continents. Clorox’s ERP shift sits inside a $500 million digitization plan; Nestlé’s ERP estate already left the data-center era. This week at IMTS 2026 in Chicago, MachineMetrics is showing MaximaOS—machine data, labor, and work orders flowing both ways with ERP so the system of record stops describing a plant that no longer exists. Chippagiri’s open question still hangs over every board deck: when an AI-driven action is wrong, who owns it?

Five professionals working on laptops around a long wooden table in a bright shared workspace. Photo: Unsplash.
Agentic ERP is a people-and-rules problem first: who may act, who must approve, and what the audit trail must prove. Photo: Unsplash (rights-safe).

For N23D, today’s brief is a BUS lesson with an ERP spine: capital can buy spindles all day, but AI without one identity plane and one local data foundation just accelerates bad decisions. N23D’s contract is one native Rust binary across BUS, ERP, and MES—person badges via familiar Entra, device and action stamps on the thread, content-addressed truth on the floor—so agents act on the plant you run, not a rented cloud copy of last night’s export.

Sources/References