Vocabulary
Shared language for structured AI systems.
Hybrid AI ArchitectureA decision system built in layers. Rules, machine learning models, and language model capability, each with a defined role.
Hybrid AI ArchitectureA decision system built in layers. Rules, machine learning models, and language model capability, each with a defined role.
Deterministic LayerRules that behave the same way every time. The foundation for everything that must hold under scrutiny.
Deterministic LayerRules that behave the same way every time. The foundation for everything that must hold under scrutiny.
Probabilistic LayerMachine learning that handles patterns and volume. Powerful within its role, not beyond it.
Probabilistic LayerMachine learning that handles patterns and volume. Powerful within its role, not beyond it.
Generative LayerLanguage model capability for interpretation and contextual understanding. The last layer, not the first.
Generative LayerLanguage model capability for interpretation and contextual understanding. The last layer, not the first.
Decision TraceabilityThe ability to reconstruct a decision. Which layer. Which input. Which outcome.
Decision TraceabilityThe ability to reconstruct a decision. Which layer. Which input. Which outcome.
Auditability by DesignTraceability from the start. Not retrofitted. The difference becomes visible when oversight arrives.
Auditability by DesignTraceability from the start. Not retrofitted. The difference becomes visible when oversight arrives.
Confidence ThresholdThe point where the system escalates to human judgment.
Confidence ThresholdThe point where the system escalates to human judgment.
Regulatory AlignmentBuilt to absorb changing requirements. EU AI Act, AML, sanctions — parameters to design for, not exceptions to manage.
Regulatory AlignmentBuilt to absorb changing requirements. EU AI Act, AML, sanctions — parameters to design for, not exceptions to manage.