Bosch teaches AI to hunt for faults — and it finds what seasoned diagnosticians miss

Bosch described a project in which artificial intelligence is being taught to hunt for faults in machinery — and, by the company's own admission, the results already unsettle experienced diagnosticians. A model trained on large failure datasets spots early signs of trouble in signals humans have learned to dismiss as noise, and does so consistently, without fatigue or tunnel vision.

The professionals' reaction is telling: it stings when a machine finds what you missed. But the economics outweigh the emotion — early diagnosis is always cheaper than repair after failure, and in vehicle fleets and factories the difference is measured in downtime and contracts. The diagnostician doesn't disappear: they stop manually cycling through hypotheses and work from a ready shortlist of suspicions.

Outside the garage the principle transfers directly: every company has its own "engine knocks" — duplicated payments, unusual discounts, stock levels that don't match sales. The auto-checks in 1Cifer are built on the same logic: an AI agent continuously scans operations and flags anomalies while they cost pennies, before they become a quarter-end shortfall. The cheapest problem is the one spotted first.