The manufacturing terms behind data-first AI, in plain English.
CNC, OEE, MES, OT/IT — the vocabulary of the plant floor decides where AI actually pays back. Here's what each term means, and why it matters when the data comes first.
CNC
The automated control of machine tools — mills, lathes, routers — by pre-programmed instructions rather than manual operation. The CNC controller executes a toolpath to cut a part to spec.
Why it matters for data-first AI — CNC machines emit rich signals — loads, speeds, alarms, cycle times — most plants never capture. Data-first AI starts by turning that stream into a usable record.
CAM
The software that converts a part's CAD model into the toolpaths and G-code a CNC machine runs. CAM is where a programmer's expertise — tool choice, cutting strategy — is encoded.
Why it matters for data-first AI — CAM is the scarce-skill bottleneck: senior programmers tied up per part. AI inside the CAM tool scales that expertise instead of replacing the programmer.
G-code
The low-level instruction language a CNC controller reads — coordinates, feeds, speeds and tool changes that move the machine. CAM software generates it from the toolpath.
Why it matters for data-first AI — G-code and the machine's response are ground truth for what actually happened on the floor — invaluable, once captured and contextualised.
Toolpath
The route the cutting tool follows to machine a feature — the sequence of moves, engagement angles and depths that determine cycle time, tool wear and surface finish.
Why it matters for data-first AI — Optimising toolpaths is one of the clearest AI wins — shorter cycles, less tool wear, and best practice standardised across every programmer.
Cycle time
The time to complete one machining operation or produce one part — a primary driver of throughput and cost per part.
Why it matters for data-first AI — Cycle time is where machining AI shows up on the P&L; small per-part gains compound across every spindle.
OT / IT
Operational Technology runs the plant floor — machines, PLCs, SCADA; Information Technology runs the business — ERP, databases, cloud. The OT/IT divide is the main reason plant data stays siloed.
Why it matters for data-first AI — Production AI lives on the OT/IT boundary — you cannot compute a cross-domain KPI until the two worlds share one data model.
PLC
The ruggedised industrial computer that controls machinery and processes in real time — a primary source of live plant signals.
Why it matters for data-first AI — PLC data is high-value and high-frequency; getting it off the controller and into the fabric is often the first integration step.
SCADA
The system that monitors and controls industrial processes across a plant, aggregating signals from PLCs and sensors into a supervisory view.
Why it matters for data-first AI — SCADA is a rich source layer — but its data has to be unified with MES, quality and maintenance to be useful for AI.
MES
The shop-floor system that dispatches, tracks and records production against the order — sitting between enterprise planning (ERP) and the equipment.
Why it matters for data-first AI — MES is where production reality lives; connecting it lets AI surface cross-domain KPIs no single siloed system can compute.
Process historian
The time-series database that captures equipment and process signals — temperatures, pressures, speeds, alarms — at high frequency.
Why it matters for data-first AI — The raw material for predictive maintenance and OEE; historian data is where equipment-failure signatures hide.
OEE
The standard measure of manufacturing productivity — Availability × Performance × Quality, expressed as a single percentage. The universal benchmark for how well an asset is really running.
Why it matters for data-first AI — OEE is the KPI most pilots aim at — but you can only compute it honestly once machine, quality and downtime data share one fabric.
Predictive maintenance
Using equipment and process data to predict failures before they happen — so maintenance is scheduled on condition, not on a fixed calendar or after a breakdown.
Why it matters for data-first AI — The recommended first pilot on well-instrumented assets: fused telemetry and maintenance history flag failures early, cutting unplanned downtime.
Condition-based monitoring
Continuously tracking an asset's health indicators — vibration, temperature, load — to detect degradation. The data foundation predictive maintenance is built on.
Why it matters for data-first AI — Without condition data there is nothing to predict on; instrumenting the right assets is step one.
MTBF / MTTR
MTBF is how long an asset runs before failing; MTTR is how long it stays down. Together they quantify reliability — and are the levers predictive maintenance moves.
Why it matters for data-first AI — These are the numbers a maintenance business case is written in; a data fabric makes them measurable per machine.
Digital twin
A live, data-driven virtual model of a physical asset, line or process — kept in sync with real signals so it can be analysed, simulated and optimised.
Why it matters for data-first AI — A twin is only as trustworthy as the data feeding it; a unified fabric is what makes one worth acting on.
Edge computing
Running compute — and AI — close to where data is generated, on the plant floor, rather than in a distant cloud. Enables low-latency, on-premises, sovereign processing.
Why it matters for data-first AI — Edge plus on-prem is how production AI stays fast, resilient and inside the plant's firewall — no round-trip to the cloud.
These are plain-English summaries. The relevant industry standards include ISA-95 / IEC 62264 (enterprise-control system integration, the MES/ERP model), OEE as defined in Total Productive Maintenance (Nakajima), and IEC 62443 for OT/IT security. Terminology varies by machine builder and MES vendor — always confirm against your own systems.
See these in a working stack.
How the plant floor, the data fabric and the AI modules fit together — on the use-cases and capabilities pages.