The most useful number to come out of today's Celonis–Molex announcement is unexciting but concrete. Molex, the Illinois-based electronics and connectivity manufacturer owned by Koch Industries, took its purchase-order confirmation rate from 30 percent to 90 percent by running Celonis process mining across its procure-to-pay cycle. Touchless invoicing, where the invoice clears without a human in the loop, now runs at 87 percent. Warehouse efficiency, measured through dock-to-stock and picking, improved 10 to 15 percent against the same before-and-after benchmark.
Those are the kind of operational figures process-mining vendors rarely get named customers to put a signature beside. Molex buys roughly 70,000 parts a year from thousands of suppliers, so the percentage gains land on a large absolute base.
Celonis pitched the expansion as proof that its Process Intelligence Graph, the mapping layer that sits between ERP transaction data and AI agents acting on it, is what makes enterprise AI reliable rather than flaky. The argument is familiar from every vendor now positioning process context as a prerequisite for agentic AI, but Molex is one of the more credible customer references to attach to it: a high-volume manufacturer willing to quantify gains in a case study.
"Process Intelligence is the heart of our digital transformation and the engine behind our AI journey," said MJ Patil, Director of Process Excellence at Molex.
Imagine you can bring an MRI machine or an X-ray machine on top of your ERP. Celonis allows you to scan the whole process end-to-end with one click, which is an incredibly powerful capability.
"Molex is a perfect example of what happens when you treat your processes as a competitive advantage," said Alex Rinke, co-CEO and co-founder of Celonis. "By using Celonis to ground their AI and automation in process intelligence, Molex isn't just fixing problems, they are reinventing their operations for a new era of intelligence."
Molex is extending its Celonis deployment from supply chain into order-to-cash, logistics and broader manufacturing processes. The intent, per both companies, is to have global plants follow the same process blueprints before agentic AI is layered on top.
