26th FVI Expert Breakfast

Topic: "The Hammer Seeks the Nail" – Why Predictive Maintenance often fails and how to do it right.

Our guest was Erik (Siemens, expert in AI in manufacturing), who brought a provocative thesis: "We give maintenance staff smartphones so they can hammer nails into the wall."

• The Portfolio Analysis: Erik showed that Siemens does not introduce Predictive Maintenance (PdM) across the board. They first analyze: How often does the machine fail? How expensive is the failure? Insight: Many machines fail daily for 5 minutes. You don't need AI for that, but you need to fix the root cause. PdM is only worthwhile for rare, expensive failures that can be detected in advance (e.g., 2 days). Everything else is a waste of money.
• Trust is good, control is better: Even if the algorithm works perfectly, people often don't trust it. An example: The AI reports an anomaly. The maintenance staff says: "Oh, that's just because we're running a new product." Later, the machine fails. The AI was right, but no one reacted. Solution: You need not only technology but a change process. The technician must understand why the AI is warning ("Vibration at bearing 3 increased"), not just receive a black box message.
• The Copilot as "Fear Killer": Siemens introduced an "Industrial Copilot" that makes manuals searchable. The fear ("Will this take my job away?") quickly disappeared when the technicians realized: "The copilot tells me where to hit, but I still have to hold the hammer myself." AI becomes an assistant, not a replacement.
• Avoiding data graveyards: Karl (participant) reported on a "Technical Controller" who focuses not on costs but on data quality. If the data basis is garbage (e.g., incorrect feedback in SAP), no AI in the world can help.
• User Journey instead of specification sheet: Instead of writing a 100-page specification sheet, you should ask: "Where do you have pain today? What would make your day better?" If you solve the worker's problem (e.g., "I can never find the manual"), acceptance of the new tool is automatic.

Classification: Strategy before technology

This episode confirms our philosophy "Simplify your digital Transformation".

• ADAM is not a "Hammer": We do not sell ADAM as a panacea. We analyze with the customer (like Siemens) where the greatest pain lies. Is it the documentation? Is it the search for spare parts? Is it communication? ADAM is modular: You only use what you need.
• The Copilot as a gateway drug: The "Industrial Copilot" described by Siemens is exactly what ADAM offers (RAG search in documents). It is the perfect entry point to build trust in AI because it immediately helps the worker without threatening them.
• Predictive Maintenance with a sense of proportion: We do not promise miracles. But if (as shown in Ep 16) WAGO reports an anomaly, ADAM ensures that this information does not fizzle out but lands as a concrete work order with the right technician. We bridge the gap between "Algorithm is right" and "Human acts".

Conclusion: AI must solve problems, not create them. Those who introduce AI without understanding the process only digitalize chaos.