findIQ diagnostic AI

What it does

• It turns your knowledge base, the heatmap of symptoms and causes, into guided troubleshooting: step by step from the observed symptom to the cause, with as few questions as possible.

• It learns from real service cases, so diagnoses become faster and more accurate over time.

How it works

The findIQ diagnostic AI is a model-based approach that combines several self-learning statistical methods. It was developed and validated under real industrial conditions in a research project with several Fraunhofer institutes and industrial partners (2020 to 2022).

It is not a large language model. findIQ uses LLMs only for the language part, see Large language models in findIQ.

Why not an LLM?

• Traceable: The diagnosis is based only on the knowledge your experts recorded. It does not add information or reinterpret expert knowledge, which matters when wrong advice can damage machines or put people at risk.

• Not random: The diagnosis is deterministic, not generated.

• Built for machine knowledge: It works reliably with the incomplete and heterogeneous information typical for machines and plants.

• Efficient: It needs significantly less computing power than generative AI.

Learning under your control

The AI does not change your knowledge base on its own. Improvements are proposed and only take effect once an Editor or Administrator accepts them:

• AI Suggestions are derived from real service cases, for example adjusted probabilities.

• User suggestions are new symptoms or causes submitted by technicians during troubleshooting.

Learning only ever improves your own knowledge base. Your data is not used to train a central model and never comes into contact with data from other customers.

Where it runs

The findIQ diagnostic AI was developed by findIQ GmbH, is owned by findIQ GmbH and runs on findIQ servers in the EU. No third-party provider is involved.