Databases must evolve into AI-native systems that securely federate with LLMs, support real-time access, granular permissions, and tools for safe natural-language-to-SQL integration.
QConAI NY 2025 - Designing AI Platforms for Reliability: Tools for Certainty, Agents for Discovery
Reliable agentic AI combines probabilistic model components with deterministic boundaries and integrates models as layers over operational systems rather than replacements.