Richmond Alake distinguishes an agent control planeAn AI control plane observes AI activity and applies centralized policy without owning the application's primary task logic. from an intelligence plane that brings together knowledge, memoryAI agent memory is stored information that an agent can retrieve and use across steps, sessions, or changing contexts., data permissionsGovernance defines responsibilities, decision rights and controls for how a system or organization is operated. and learning. He argues that the same underlying models can produce different outcomes when systems retain and use relevant contextContext engineering designs how relevant instructions, evidence, state and tool information are assembled for a model's task..
Richmond Alake demonstrates a governed analyst and a comparison agent, discussing how semantic definitions and ontologies affect their answers and reported token use. These results are examples from his demonstration rather than independently verified performance claims.
The presentation covers memory traces, identity-based data accessAn AI agent identity is a persistent, verifiable identifier used to attribute messages, actions, permissions, and history to an agent., private inference and routing different memories. Richmond Alake describes fine-tuning smaller models for routing and using captured traces to support ongoing improvement.
Watch on YouTube
