Published research programme

Six papers, five manifestos, one causal arc

A systematic research programme on epistemic governance for regulated industries. Papers co-authored with Arnaud Gelas. Five published on SSRN, one on Zenodo, a seventh in preparation.

The causal spine

Enterprise AI fails because of dynamics blindness (A) → the resolution is architectural (B) → ten independent traditions converge on the same requirements (C) → the practitioner methodology includes epistemic immunity (D) → at sufficient depth, governed initiative emerges (E) → the Knowledge Layer is the reference architecture that makes delegated action admissible at the moment it commits (G) → ontology governance keeps the type system beneath the claim graph governable, versionable, and transferable (F).

Working papers · each under a permanent DOI

Paper A — The diagnosis · Published on SSRN

Dynamics Blindness: When AI Is Locally Correct and Globally Non-Compliant

LLMs process tokens without tracing causal chains through organisational dependencies. Chain-of-thought, RAG, tool use, and multi-agent systems do not add the missing causal infrastructure. The problem is structural, not parametric.

Reichhart, W. & Gelas, A. (2026)  ·  DOI 10.2139/ssrn.6230758  ·  Read on SSRN →  ·  Download PDF ↓  ·  The explainer →

Paper B — The resolution · Published on SSRN

The Predictive Organization: Architecture for Enterprise Intelligence

A tripartite structure - Map, Physics, Player - coupling neural perception with symbolic reasoning, operating on claims-based knowledge with prevalence weighting.

Gelas, A. & Reichhart, W. (2026)  ·  DOI 10.2139/ssrn.6230780  ·  Read on SSRN →  ·  Download PDF ↓

Paper C — The foundations · Published on SSRN

Build the Medium: Why Organizational Intelligence Is Mechanism, Not Metaphor

Ten independent theoretical traditions - from cell biology to social systems theory - converge on the same architectural requirements for organisational intelligence. Introduces the capability/fertility distinction and the autonomy-to-initiative transition.

Reichhart, W. & Gelas, A. (2026)  ·  DOI 10.2139/ssrn.6230858  ·  Read on SSRN →  ·  Download PDF ↓

Paper D — The methodology · Published on SSRN

Governed Intelligence Architecture for Institutional AI

The Governed Intelligence Lifecycle - Ingest, Consolidate, Curate, Expand, Apply - with an epistemic immunity framework protecting against six systemic knowledge failures. Names epistemic operational risk: the risk of institutional harm caused by acting on degraded knowledge.

Gelas, A. & Reichhart, W. (2026)  ·  DOI 10.2139/ssrn.6701941  ·  Read on SSRN →  ·  Download PDF ↓

Paper E — The capstone · Published on SSRN

From Autonomy to Initiative: Enterprise AI's Real Endgame

The AI industry optimises for autonomy when the real prize is initiative - agents that perceive what matters through immersion, not instruction. Three conditions for governed initiative, governance relocation, and the domain graph as the missing middle layer.

Reichhart, W. & Gelas, A. (2026)  ·  DOI 10.2139/ssrn.6702239  ·  Read on SSRN →  ·  Download PDF ↓

Paper G — The reference architecture · Published on Zenodo

The Knowledge Layer: A Reference Architecture for Delegated AI Action in Regulated Institutions

AI capability is no longer the constraint on enterprise deployment. Admissibility is. The claim as the atomic unit, standing on two governed axes of evidential warrant and authority, a consequence-scaled gate, and a replayable action manifest - the record an institution can put in front of a supervisor after the fact.

Reichhart, W. & Gelas, A. (2026)  ·  DOI 10.5281/zenodo.20784662  ·  Read on Zenodo →  ·  The architecture →

Paper F — The engineering · In preparation

Ontology Governance in Claim-Level Architectures: The Bootstrap Problem, Versioning, and Cross-Domain Transfer

The type system beneath the claim graph. Bootstrap recursion, ontology versioning bound into manifest replay, and cross-domain alignment handled as a governed claim class - the ontology governed by the same mechanism as the claims it types.

Gelas, A. & Reichhart, W.  ·  Full draft under review  ·  DOI on publication

The Agentic Governance Stack

Five public manifestos

A five-layer governance framework spanning engineering practice through enterprise transformation. Each layer has a published manifesto with principles and implementation guidance. Layers 1-3 authored by Arnaud Gelas. Layers 4-5 co-authored.

Layer 1
Layer 2
Layer 3
Layer 4

Intelligence Governance Manifesto

Reichhart, W. & Gelas, A.  ·  v1.5, May 2026  ·  CC BY-SA 4.0

Layer 5

Agentic Enterprise Manifesto

Reichhart, W. & Gelas, A.  ·  v0.2, May 2026  ·  CC BY-SA 4.0

Vocabulary

Named contributions

Original concepts introduced across the research programme.

Machine-Readable Intelligence (MRI)

Dynamics Blindness

Governed Intelligence Lifecycle

Epistemic Governance

Epistemic Immunity

Capability / Fertility

Epistemic Operational Risk

Autonomy-to-Initiative

Governance Relocation

Living Medium

Circuit Breaker Principle

Map / Physics / Player

Domain Graph

The Knowledge Layer

Admissibility

Two-Axis Standing

Replayable Action Manifest