On 18 April I spoke at a SWICO event in Zurich to an audience of technology-industry leaders. The topic was AI in knowledge management. My goal was to strip AI of its mystique and give a concrete account of how the technology works and where it applies.
In the Luma Westbau venue, surrounded by floor-to-ceiling bookshelves and art installations, I walked through the mathematical foundations of AI and machine learning: linear regressions, gradients, and vectors, showing that AI rests on statistical methods structurally analogous to biological neural networks.
Using a map visualization, I demonstrated how data selection shapes AI model outputs, specifically in relation to political bias. This underscored the need for critical scrutiny of training data.
I also addressed the constraints facing smaller companies, which are often limited by the compute requirements of large models. I discussed current research on making models more efficient and accessible without sacrificing performance, a prerequisite for better handling data privacy and security concerns.
The event included an open discussion during the apero, briefly interrupted by a fire alarm. Participants raised pointed questions about data privacy and the manipulation potential of AI systems. I emphasized the importance of incremental adoption and non-negotiable ethical standards.
Netzwoche published a > write-up of the event.
The main topics included the mathematical foundations of AI and machine learning, practical applications in knowledge management, challenges for smaller companies, and ethical aspects of AI implementation.
A map was used to demonstrate how data selection influences AI models, particularly regarding political biases. Additionally, approaches for implementing AI in smaller companies were presented.
Central challenges include the high computational requirements of large AI models, data privacy and security concerns, and the need for ethical standards during implementation.
Companies can optimize their knowledge management processes through gradual AI implementation while maintaining ethical standards and data privacy requirements.
.
Copyright 2026 - Joel P. Barmettler