About this talk
This talk, presented by Philip Langer and Stefan Dirix from EclipseSource, explores ten practical learnings from building domain-specific AI solutions. It emphasizes the importance of defining an "AI vision" and grounding use cases in unique capabilities such as proprietary data access, domain-specific actions, and deep system integrations. The speakers discuss the critical need for structured user workflows and objective validation methods, alongside the integration of human-in-the-loop strategies in the development process. They also address the architectural transition from code-orchestrated flows to more agentic systems, highlighting context management challenges, and the complexities of deploying AI in production environments, where prioritizing precision becomes essential for reliability and continuous improvement.
