From sensor data to decisions — the gap nobody draws on the whiteboard

At the VDMA AI summit in Frankfurt today, someone put up a slide showing "sensor data → AI → decision." Clean arrows, confident boxes. I've spent my career on printed circuit boards where those arrows are the hard part — not the boxes.

Here is what the diagram skips. A sensor tells you a voltage. It does not tell you whether that voltage means the bearing is warm because it is failing or because it is Tuesday afternoon and the hall is 32 degrees. Context lives in the machine, in the shift schedule, in the operator's memory. It doesn't live in the signal.

This matters for Maschinen- und Anlagenbau specifically: your machines run for twenty years. The sensor that was spec'd in 2008 produces a data stream that was never designed to be fed into a language model. The column headers are in the head of the engineer who retired. The calibration drift is undocumented. Calling that "data" before you've done the archaeology isn't honest.

Apuna's posture — data foundation first — isn't a consulting cliché. It's what the circuit board teaches you. Before the model sees the signal, someone has to trace the board. Who fitted the sensor, to what spec, and what have they done to it since? When you can answer that, you have data. Before you can, you have noise with a timestamp.

The AI isn't the problem. The archaeology is the work.