Industry
Your machines generate more data than your sales team. Almost none of it is used.
For small and mid-sized manufacturers and mechanical engineering companies with 5 to 250 employees: machine data into maintenance work orders, completion reports from the machines into the ERP, camera images into quality records. The link between production and ERP, without your own data science team.
Servers in Germany · AI advises, a human decides · Transparency under the EU AI Act
Starting situation
What shapes daily work in production
The controller shows states at the machine. The ERP knows orders. The foreman brings the two together, on paper.
Maintenance by the calendar: some of it too early, some too late. One unplanned downtime can cost more than a year of maintenance.
45% of the companies surveyed name a lack of staff resources as a problem when using AI, 42% insufficient data quality (survey by VDMA, the German Mechanical Engineering Industry Association, published December 2025).
A platform for thousands of sensors does not fit 20 machines. And you do not have time for a two-year project.
Your best people will retire in the coming years. Their knowledge of the machines goes with them.
Source: report on the VDMA survey on AI in mechanical and plant engineering, der-maschinenbau.de, 18 December 2025.
AI in mechanical engineering: what we develop for this industry
Opening up machine data. From the controller, the manufacturer's cloud or retrofitted sensors: what is there gets used; what is missing is added where it is needed. No sensor network if three measuring points are enough.
Predictive maintenance as a closed cycle. Anomaly detected, maintenance work order in the ERP with priority, execution with completion report, a record of whether the failure was avoided. For industrial equipment we offer this as a pilot. → Predictive maintenance & industrial AI
Connecting production and ERP. Order in the ERP, completion report from the machine, stock and material requirements without slips of paper. → AI integration & process automation
Camera as inspector. As a pilot at one station: counting, completeness and label checks on the line; the result goes into the system, the deviation becomes a task. → AI for machines & robotics
Securing knowledge. An AI employee for maintenance that knows the machine history, maintenance plans and the experience of your foremen, and answers the new colleague when the experienced one is no longer there. → AI employees
What matters most in manufacturing and mechanical engineering
No intervention in the controller or safety. We read data and write work orders. We do not control any machine; this boundary is set out in the contract. Our part intervenes neither in the controller nor in safety functions and therefore does not affect the CE conformity of your machine (manufacturer's declaration).
Numbers before technology. We do not start a pilot without a measured baseline. Only then can you say at the end whether it was worthwhile.
Lean instead of a platform. A fixed-scope pilot on 5 to 10 machines: 3–6 months, preceded by 2–4 weeks for the baseline. Afterwards you decide on the basis of your own numbers.
What already runs of the cycle, without customer names
We are currently building the cycle «signal → task → execution → record». Parts of it run today, with scales, printers and scanners as signal sources.
Retrofit instead of a sensor network is the path for the first industrial pilot: individual measuring points, time series, anomaly detection, connected to the ERP you already have.
-
To be honest: a completed industrial pilot is still to come. For the first one we are looking for pilot partners. We discuss the terms in the first call. → Future Lab
The next step for this industry is the closed maintenance cycle, then the camera on the line. → Maintenance cycle: from signal to work order · → Camera and robot on the line
Operation on servers in Germany or as on-site (edge) operation in your plant. Critical data runs offline and locally; we connect external language models only for non-critical data and only as processors under Art. 28 GDPR. Advisory AI without intervention in the controller or safety functions. AI identifies itself as AI. → AI transparency
Frequently asked questions: manufacturing and mechanical engineering
Do you intervene in our machine controller or safety circuits?
No, we do not intervene in the controller or safety circuits: we read data and write work orders. Controller, emergency stop and safety circuits stay with the manufacturer, and this is set out in the contract.
How does a pilot on 5 to 10 machines start, and what do you measure beforehand?
First we measure the baseline for 2–4 weeks: downtime hours, unplanned failures and maintenance costs over the last twelve months. After that the pilot takes 3–6 months: we open up the data, define the logic and close the cycle.
How do we secure our foremen's knowledge before they retire?
With an AI employee for maintenance: machine history, maintenance plans and the rules of your foremen become its memory. The foreman corrects it while they are still there; their successor asks it afterwards.
Do we need a data science team or a platform?
No, you do not need your own data science team or a platform. For 5 to 50 machines, some of them critical, the controller or a few measuring points, your ERP and a fixed-scope pilot are enough. We build and maintain the models.
How the pilot works in detail: → Predictive maintenance & industrial AI
Name the three machines whose downtime costs you the most
And what a day of downtime costs today. In the first call we estimate whether and how that can be changed.
Reply within one working day.