White Paper

Mapping the AI Landscape

From Statistical Automation to Constitutional Reasoning Systems

Over the last decade, "AI" has collapsed into a single category — obscuring the fact that it is actually a collection of fundamentally different systems, built on different assumptions and optimized for different problem classes. In regulated and high-stakes environments, this conflation has produced persistent confusion, misaligned deployments, and material risk. This paper maps seven distinct AI families, explains where each fails structurally, and defines a new class of systems required to own decision authority.

Key Insights

Seven AI families compared side-by-side — predictive ML, generative, agentic, decision intelligence, world models, ARC-style reasoning, and constitutional — with the structural limit that defines where each must not be deployed

The category-error thesis: most AI failures stem not from model weakness but from applying the wrong class of system to the wrong kind of decision — a mistake no amount of additional data or tuning can correct

A formal definition of Constitutional Reasoning Systems (CRS) and the nine core services they perform — from constitutional resolution and rule extraction through conflict resolution, breach detection, audit trace, and formal refusal

Why stacking LLMs, agents, dashboards, and rules engines cannot replicate a CRS — refusal and constitutional authority have no natural home in any other AI family, and no component in the stack "owns" the constitution

Written for CCOs, CROs, General Counsel, Chief Audit Executives, board-level stakeholders, and technical evaluators across banking, insurance, pharma, venture capital, and the regulated public sector. 28 pages. Framework-level analysis, not vendor-specific architecture. Includes real-life decision narratives in venture capital, insurance underwriting, and pharma portfolio governance.

Download the White Paper

Required fields

A secure download link will be sent to your email. Valid for 1 hour.