The AI Operator: How every shift reaches the level of your best operator

Two shifts, the same machine, the same material. Yet at the end of the day, output and scrap are far apart. The difference is the person at the HMI: the experienced operator can hear that the process is drifting. They know where to look when a fault occurs and which parameter to adjust. None of this is written down in a manual.

This is exactly where the AI Operator comes in. It makes the know-how of the best operators usable on every shift. This article shows how it works and where you as a machine and plant builder can start.

A machine is only as good as its operation

Machine builders deliver highly sophisticated equipment. Their customers increasingly have to run it with less experienced staff. According to ManpowerGroup’s 2026 Talent Shortage Survey, 83 percent of employers in Germany report difficulties filling open positions. Worldwide, the figure is 72 percent. More than 39,000 employers in 41 countries were surveyed.

For the US, Deloitte and the Manufacturing Institute expect manufacturing to need up to 3.8 million additional workers by 2033. Up to 1.9 million of these jobs could remain unfilled.

Then there is the loss of knowledge. When experienced operators retire or change jobs, their knowledge leaves with them. What remains are error messages, manuals and a call to your service team.

Where your customer does the math: TCO and OEE

Your customer evaluates a machine by two measures. Total cost of ownership (TCO) is the sum of investment and operation. The investment is largely determined by materials, wages and location factors. Operation covers the costs over the lifecycle: operators, damage, energy, maintenance.

Overall equipment effectiveness (OEE) measures what the machine delivers in operation. It is the product of three factors – availability × performance × quality:

1. Availability – setup times, planned and unplanned downtime.
2. Performance – optimal parameterization, cycle time deviation, micro-stops.
3. Quality – scrap, rework, first-pass yield.

All three factors depend heavily on who operates the machine. You can hardly set yourself apart on price alone anymore. On operation, you can.

You don’t just deliver the machine. You deliver your best operator along with it.

Five dimensions between the best and the weakest shift

Where exactly the difference arises can be described in five dimensions. Each one has levels ranging from where you are today to optimal operation.

Process optimization

The experienced operator notices when something drifts. Closed-loop AI sees it in the data, recommends the correction and, if you want, writes it back to the PLC. The path leads from process data recording via extended input data and feedback to closed-loop AI.

Operating concept

How confidently someone operates your machine currently depends on their experience. A task-oriented HMI with wizards and visually guided setup turns this into a reproducible procedure.

Knowledge management

Experience builds up over years. Documentation and digital training are the starting point. Best practices, recipe logs and recipe comparisons as well as interactive instructions make this knowledge usable in day-to-day operation.

Fault handling

When a fault occurs, the experienced operator knows where to look. Everyone else calls service. The levels range from the error message via self-diagnosis and guided troubleshooting to autonomous error recovery.

Connectivity and data use

No data, no AI. Many machines today run standalone or with a local data logger. The path leads via remote maintenance and an IoT solution to an autonomous system.

Connect, Store, Learn, Act

Technically, the AI Operator works in four steps:

1. Connect – data from the control system, sensors and other measured variables that influence the process. Connected via OPC UA, S7, Modbus or MQTT, in brownfield and greenfield alike, with preprocessing at the edge.
2. Store – an Industrial Data Platform based on TimescaleDB and PostgreSQL, with an MQTT broker, an asset model and permission management via Keycloak. If a platform already exists, we connect to it.
3. Learn – the operator model. It learns from the data and from the knowledge of the best operators, from instructions and SOPs. Depending on the task, we use regression, classification, reinforcement learning, RAG or agentic AI.
4. Act – recommendations to the operator and parameterization, up to semi-autonomous process and recipe optimization.

The human decides

The AI Operator does not replace any operator. The model suggests, the human decides. Every decision is traceable: the operator reviews a recommendation, approves it or overrides it.

An override is valuable too. It flows back into the model as feedback, together with the results of quality control. This way, the model gets better with every shift. We only close the control loop to the PLC on request and with the operator’s approval.

Where does your machine stand?

The path to the AI Operator is not a leap. It leads through five maturity levels, each of which is useful in its own right:

1. Connect – the machine is connected.
2. Visualize – data comes together in an IoT platform.
3. Collect – targeted data extracts for analyses.
4. Accelerate – a first digital service for your customers.
5. AI Operator – closed loop, semi-autonomous, with human-in-the-loop.

Many machine builders don’t start from zero. Machines are often already connected and data is available. Then it is about taking the next step properly.

In practice

For RAMPF, a closed-loop control based on our Industrial Data Platform adjusts the dispensing parameters on its own. The process stays stable despite raw material fluctuations and changing ambient conditions.

How to get started

1. Introductory call – we talk about your machine, your customers and your goal.
2. Suitability check – in 90 minutes, we identify the use cases with the greatest leverage.
3. Pilot – one use case, real data, a fixed timeframe. After that, we implement further use cases and AI models iteratively.
4. Rollout – integration into series production and scaling across your fleet.

You can find more on the building blocks and technology on the AI Operator page. If you want to know where the biggest AI levers in your machine are, get in touch.