Introduction
Nell’industria manifatturiera la conoscenza di processo è un vantaggio competitivo decisivo, ma spesso resta nella testa di pochi tecnici esperti che l’hanno costruita in decenni di lavoro sul campo. Con il supporto di MADE Competence Center, AISent ha sperimentato Meti, un assistente basato sull’intelligenza artificiale che trasforma l’esperienza dei senior in modelli causali e la incrocia con i dati delle macchine in tempo reale. Così guida gli operatori junior nella risoluzione dei fermi produttivi. Al progetto hanno collaborato Siemens e Fasthink.
The Problem
Rising staff turnover and a shortage of trained candidates leave companies exposed. If a critical fault happens when the only technician who can fix it isn’t available, production can stop. There are three main issues:
Operational vulnerability: without expert guidance, junior operators struggle to find the root cause of a stoppage and fall back on trial and error, which makes recovery take longer
Loss of know-how: every retirement or staff change risks losing strategic technical knowledge that was never written down.
Data and expertise kept apart: machine data from PLCs and technicians’ experience are still two separate worlds. Only by bringing them together can companies find out why a malfunction happened.
The solution
AISent connected Meti to the factory ecosystem through a software connector developed with Siemens and Fasthink. With the connector, Meti can query the machine’s status directly and compare real-time signals with the digitized knowledge models.
Meti is built on causal models, a form of AI that goes beyond statistics: it reproduces the logic an expert follows when facing a problem, so the diagnosis is guided and transparent. In the test environment, Meti received data from the machine control system through Siemens Industrial Edge and the Fasthink connector. The test checked whether the assistant could suggest a fix right away, based on the current state of the assets.
Meti is built on causal models, a form of AI that goes beyond statistics: it reproduces the logic an expert follows when facing a problem, so the diagnosis is guided and transparent. In the test environment, Meti received data from the machine control system through Siemens Industrial Edge and the Fasthink connector. The test checked whether the assistant could suggest a fix right away, based on the current state of the assets.
Tecnologies
– Meti (AISent): software that combines AI and statistics to map experiential knowledge through causal graphs.
– Siemens Industrial Edge: platform that collects data from the machine control system.
– Fasthink connector: software module that links Meti to real-time machine data.
– Causal models: AI that reproduces an expert’s diagnostic reasoning in a transparent, interpretable way.
Expected impacts
Significantly reduce machine downtime through guided, faster diagnosis.
Capture senior technicians’ expertise as a shared digital asset that is always available.
Help junior operators become independent sooner, even when production varies a lot.
Combine machine data with human experience, so the assistant becomes an active advisor that responds immediately to error signals rather than a passive archive.
Next steps: roll Meti out across multiple sites, add self-learning from newly resolved maintenance events, and add connectors for more industrial protocols and older (legacy) PLC systems.
Benefits for the Company
1. Up to 70% less machine downtime thanks to diagnosis based on causal models.
2. Continuous on-the-job training: junior operators learn while solving real problems, with a “virtual expert” always at their side.
3. More effective diagnosis: raw technical data and human intuition work together, even on the most complex cases.
4. Fast deployment: reaching full production takes about 3 months on average.
5. Less reliance on key individuals: the company keeps its know-how even when staff change.