AI_PLUMBER
SYSTEM_INDEX
UPLINK STATUS: OPTIMIZED
ACCESS_LEVEL: ADMIN_ROOT
SESSION_ID: 0x99_PIPE_FLOW
LAST_SYNC: 22.03.2026_04:00_GMT
©2026 AI_PLUMBER_CORP
architecture AI PLUMBER
Home / Whitepapers / AI Plumber Framework
Whitepaper

The AI Plumber Framework

Governance-first architecture for agentic AI systems in regulated environments. By Koen Van Lysebetten.

Why AI Governance Must Come First

In regulated AI, the moat is not the model — it's the governance layer.

Most AI deployments start with model selection and end with a governance retrofit. In regulated environments — banking, healthcare, insurance, public sector — this sequence creates compliance debt that compounds with every agent action.

The EU AI Act, SAMA, GDPR Article 9, and sectoral frameworks now require organizations to prove that AI systems are auditable, attributable, and reversible from day one. Retrofitting governance onto a production system is orders of magnitude more expensive than building it in.

The AI Plumber framework reverses the traditional deployment sequence: governance becomes the first architectural layer, not the last compliance checkbox.

The Six Pipes Framework

Six architectural layers that must be operational before any agent gets write access to production systems.

01

Identity

Constrained agent identities with narrowly scoped service accounts. No agent inherits human privileges. Cryptographic verification at every service boundary.

02

Audit

Every agent decision logged with full input context, reasoning trace, and output action. Creates a 100% reversible decision trail for regulatory compliance.

03

Kill Switch

Continuous telemetry tracks agent behavior against predefined safety thresholds. Anomaly detection triggers automatic suspension, rollback, and human escalation.

04

Human Gate

High-stakes actions require explicit human approval. The workflow mechanically pauses and awaits a human authorization token before proceeding.

05

Legacy Insulation

Isolation layers between AI agents and legacy systems. Prevents cascading failures and ensures backward compatibility without compromising governance.

06

Learning Loop

Continuous improvement cycles driven by production telemetry. Model performance, governance violations, and cost trends feed back into architecture decisions.

Production Examples

The framework has been deployed in regulated environments across insurance, public transport, and restaurant intelligence platforms.

Explore More

Download the complete whitepaper or book an architecture review.