Diagnostic uncertainty: the real cost driver in service

What diagnostic uncertainty means

A customer reports a fault. What they describe is a symptom — the machine will not run, it makes a noise, an indicator is lit. Between that symptom and the cause sits an uncertainty the service process has to bridge somehow before it can act. That gap is diagnostic uncertainty.

Traditionally it is bridged with the most expensive instrument available: you send a person to go and look. The visit is not the resolution — it is the diagnostic instrument, and that is exactly why it is so expensive.

Where it turns into cost

Four cost types come directly out of not knowing. The second visit, because the part was missing. The wrong skill, because the problem turned out different. Spare-parts stock that is too large because you want to be ready for everything — and still lacks the rare item. And extended downtime at the customer, which in industrial applications often exceeds the repair cost several times over.

None of these appear under a line item called "diagnosis". They spread across logistics, labour, inventory and goodwill — which is why they are rarely seen for what they are: one cause with four invoices.

What AI changes about it

The step from symptom to probable cause was expensive for decades because it required experience — and experience sat in heads you had to pay and send out. Modern models can perform that step at very low marginal cost, provided the foundation is there: structured machine and fault data, maintained documentation, clean parts masters.

That foundation is the actual bottleneck. The jump from document to queryable structure is the precondition for everything else — and it is the step no software purchase replaces. Skip it and you have bought an assistant that guesses on unusable data.