Agentic AI with open source: a preview of DINAcon 2026

16.11.2026 | News | Luca Albrecht

On Wednesday, the 18th November 2026, DINAcon returns to Bern, and we will be there with a session by our CEO Michael Eichenberger. His talk explores how AI is evolving from answering questions to completing tasks, and what this means for performance, costs and security. We take a practical look at open models and show why sovereign, open-source infrastructure matters. Our message is optimistic: AI is not something we need to fear, but a powerful technology we can shape and control.

The talk begins with a fundamental shift: from chat to completed work. Instead of simply generating answers, AI agents pursue goals, interact with tools, investigate problems and verify results. This raises an important question: are today's open models capable enough for these tasks? We examine their real-world performance, discuss the limitations of public benchmarks and explain why evaluating successful task completion matters more than impressive scores alone.

Agentic AI also changes the economics of inference. A seemingly simple request can trigger numerous model calls, tool interactions and repeated processing steps, increasing computing requirements and costs. We explore how context engineering, efficient caching and selecting the right model for each task can make these systems more economical. We also introduce agent sandboxes as an important building block for secure execution.

Finally, we look at what these developments mean for cloud providers. Agent workloads are less predictable, while requirements for security, transparency and operational control continue to grow. Yet we see opportunities rather than reasons for alarm. With open technologies, sovereign infrastructure and appropriate human oversight, organizations can harness AI without surrendering control. Together with our security partners, we believe agentic systems can help tackle complex technical challenges that have remained unresolved for years.

→ More about our AI on Demand offering.

More News