← Back to Field Notes

March 29, 2026 · The Bloomfield Team

AI vs. MES: Do You Need Both?

A Manufacturing Execution System tracks what happened on the shop floor. An AI system tells you what to do about it. They solve fundamentally different problems, which is why manufacturers keep asking whether they need one or both.

The answer for most operations with 50 or more employees: you need both, but probably in a different order than you think.

What an MES Actually Does

An MES sits between your ERP and your shop floor. It captures real-time production data: which jobs are running on which machines, how long each operation takes, who is running it, what the scrap rate looks like, and whether the work order is on schedule. Products like Plex, IQMS (now DELMIAworks), and ProShop handle this layer.

The core function is execution tracking. When a job moves from one work center to the next, the MES records it. When an operator logs a quality check, the MES stores it. When a machine goes down, the MES captures the downtime event. The MES creates a digital record of how work actually flows through your facility.

What an MES does not do is learn from that data. It does not identify patterns in your setup times, predict which jobs will run late based on historical performance, or recommend a scheduling sequence that reduces changeovers. An MES is a record-keeping system. A very good one, but record-keeping is the ceiling.

What AI Does With That Data

AI takes the data an MES collects and finds patterns that humans cannot see across thousands of records. A production visibility system powered by AI can analyze three years of job data and identify that jobs with a specific combination of material, tolerance range, and machine assignment consistently run 22% over estimated cycle time. That pattern exists in the data. Nobody found it because nobody had time to cross-reference 4,000 job records.

AI also handles unstructured data that MES systems ignore. Setup notes from operators. Email threads about tooling problems. Engineering change orders buried in shared drives. Quality nonconformance reports written in free text. MES captures structured events. AI reads everything.

The practical difference shows up in daily operations. An MES tells your production manager that Job 4872 is two hours behind schedule on the Mazak. An AI system tells them that jobs with this tolerance profile on this machine run late 68% of the time and recommends moving similar jobs to the Okuma, where the historical on-time rate is 94%.

Where Manufacturers Get the Sequence Wrong

Many manufacturers assume they need a fully deployed MES before they can use AI. That assumption delays AI adoption by 12 to 24 months and often leads to MES projects that stall because the implementation is too large and too disruptive.

Here is what actually works. If you have an ERP with three or more years of job data, you have enough structured data for AI to deliver value today. You do not need real-time machine monitoring to build an AI quoting tool. You do not need operator login tracking to build a knowledge capture system. You do not need OEE dashboards to build a scheduling optimization tool.

AI can work with the data you already have. Your ERP holds more useful data than most manufacturers realize. The historical record of every job you have quoted, won, lost, produced, and shipped contains enough signal for AI to improve quoting accuracy, predict delivery timelines, and identify margin leaks.

When You Need MES First

There are specific scenarios where MES should come first. If your operation has no digital record of production activity beyond what the ERP captures at job close, you are missing the in-process data that makes AI most powerful for scheduling and production optimization. If your floor runs on paper travelers with no digital tracking of job movement between work centers, an MES or a lightweight digital tracking system gives you the data foundation AI needs for real-time production intelligence.

Shops running 200 or more active jobs simultaneously with complex routing, especially in aerospace or medical device manufacturing where traceability is regulatory, benefit from MES infrastructure before layering AI on top.

When AI Should Come First

If your biggest bottleneck is in the front office, AI wins first. Quoting, estimating, customer communication, knowledge capture, and job costing analysis all run on data that already exists in your ERP and your files. No MES required.

For a 50-person job shop where quoting takes four to five days and win rates are under 20%, spending $150,000 on an MES that takes nine months to deploy makes less sense than spending $80,000 on an AI quoting tool that ships in 12 weeks and starts recovering revenue immediately.

The sequence should follow the constraint. Where is the bottleneck? If the answer is front-office throughput, start with AI. If the answer is shop-floor visibility, start with MES or a lightweight tracking system, then add AI.

The Combined System

The long-term architecture for a competitive manufacturing operation includes both. MES captures the real-time production record. AI reads that record alongside everything else, including ERP data, quoting history, quality records, and operator knowledge, to surface patterns and drive decisions.

The manufacturers building this combined system today are creating an operational advantage that accumulates over time. Every job produces more data. Every data point makes the AI more accurate. The gap between shops running on integrated AI and MES and shops running on spreadsheets and memory will widen every quarter for the next decade.

The question is which piece to deploy first. Follow the constraint.