Thursday
Room 1
09:00 - 10:00
(UTC+10)
Talk (60 min)
Keynote: Your Agents Are Smart. Your Data Model Isn’t
Most “AI agents” don’t understand your organization—they just autocomplete their way around it. We give them text and tools and prompts, then act surprised when they hallucinate, mis-sequence APIs, or violate basic business rules. The uncomfortable truth: LLMs aren’t failing to reason. We’re failing to tell them what things mean and how the organization actually operates.
This keynote argues that ontologies are the missing semantic contract between LLMs and the real word. Not just for data—but for APIs, workflows, and for the systems we're nowadays asking our agents to build for us. By explicitly modelling how an organization works—what services exist, what they guarantee, how processes flow—we enable AI systems to reason across data and behavior.
This is not ontology theory, and it’s not more prompt engineering. It’s a shift in how we architect AI-enabled systems: from “clever interfaces over chaos” to explicit, machine-reasonable models of reality. If you’re building agents that touch production systems, coordinate APIs, or make decisions with real consequences, this talk will challenge some comfortable assumptions—and offer a more robust path forward.
