Hosts Jakub Zavrel and Stephan Ballenweg examine what it takes to make AI dependable in factories, from domain-specific chemistry knowledge and fragmented plant data to safety, trust, and change management. Stephan draws on his PhD in chemistry and nearly 20 years at BASF, while Jakub connects industrial use cases to emerging AI systems and deployment challenges.
WTCK? Episode 2: AI meets the factory floor
4 NOVEMBER 2026
Key topics
- Stephan explains how his chemistry background helped him understand researchers’ needs and translate domain knowledge into useful product features, including chemical structure search.
- A chatbot’s failure to distinguish polyamide 6 from polyamide 66 illustrates how seemingly small errors can undermine credibility with specialists.
- Industrial AI must account for physical operating conditions and safety, as well as sensor time series, technical drawings, audio, video, and 3D models.
- Factories rely on fragmented systems including MES, ERP, machine controllers, maintenance records, operating procedures, and CRM; connecting information across them remains a work in progress.
- BASF’s PlantGPT gives operators access to plant-specific knowledge to support better-informed decisions, but each plant requires its own knowledge capture.
- Stephan argues that plant conditions change with raw materials and operating circumstances, so a digital twin cannot fully replace experienced operators and scientific understanding.
- The hosts discuss the Siemens-NVIDIA partnership, digital twins, agentic production planning, and reported potential reductions in engineering effort and time to market.
- Moving from demonstrations to dependable deployment requires software maturity, careful build-or-buy decisions, quality across edge cases, trust, and organizational change management.
- Jakub and Stephan conclude that industrial AI should augment people, while domain expertise becomes even more important when AI affects the physical world.
Links
Dr. Stephan Ballenweg
Founder, BKI Advisory
