Navis Mater | Flagship white paper
The Governed Agentic Organization
An Operating Model and Control Architecture for Organizations Where Humans and AI Agents Work Together Under Verifiable Governance
Artificial intelligence is moving from assistance to action. The management challenge is shifting from governing AI applications to governing an organizational system in which humans and machine actors both perform consequential work.
This paper proposes an operating model and control architecture for organizations where humans and AI agents work together, grounded in Navis Mater's reference implementation. It distinguishes demonstrated practice from design objectives and questions that remain to be verified.
Compared with the preceding edition, Edition 2.0 adds evidence from two governed offices led by the group's Chief Technology Officer and its Group Controller, practical cases of how people and agents divide the work, and a measured account of how long-running agent sessions stay coherent. Every new factual claim about the reference implementation is recorded in an evidence register and confirmed by the owner of the facts before it enters the text.
For senior technology, security and business leaders responsible for scaling agentic AI in consequential organizational environments.
Want to discuss GAO for your organization? Book a discovery workshop →
Navis Mater is defining what governed means.
Category axiom
Every Governed Agentic Organization is agentic. Not every Agentic Organization is governed.
Ten principles
The architecture can be reduced to ten principles that are stable across vendors, models and implementation styles.
Purpose before automation
Every material agent capability must serve an explicit organizational outcome.
Identity before authority
No consequential authority without attributable identity.
Authority is not reachability
Technical capability does not create organizational permission.
Every persistent agent has a mandate
Purpose, boundaries, escalation and accountability must be explicit.
Policy without enforcement is not control
Controls must prevent or pause prohibited actions, detect violations, or verify compliance with important boundaries.
Autonomy follows evidence
Authority expands only as verification maturity supports it.
Humans retain accountability
Delegating execution does not silently transfer responsibility.
Independent verification scales trust
The builder cannot remain the sole judge of its own controls, and consequential findings must have an independent path to accountable human authority.
Governance survives technology replacement
Models and runtimes are replaceable infrastructure.
Solve for N
Every meaningful failure should improve the organizational system.
Version history
- Edition 2.0September 2026Current edition.
Follow the development of GAO
Get new editions and significant updates of the paper by email. A few emails a year, no sales emails.
Questions and answers
What is the white paper about?
It sets out an operating model and control architecture for the Governed Agentic Organization (GAO): organizations where humans and AI agents work together under verifiable governance.
Who is it for?
Senior technology, security and business leaders responsible for scaling agentic AI in consequential organizational environments.
Which edition is this, and when was it published?
Edition 2.0, September 2026, by Emanuel Lipschütz. It builds on the preceding edition and adds evidence from two governed offices, practical cases of how people and agents divide the work, and a measured account of how long-running agent sessions stay coherent.
What formats is it available in?
Online, as a PDF of 47 pages, as Markdown and as a single HTML file, all in English.
Do I need to register to download it?
No. The white paper is free to download, with no registration.
What is Verified Autonomy?
The governing doctrine of the model: autonomy is a contextual grant that expands only when the organization can verify the authority, boundaries, execution and evidence around it. Every increase in autonomy should have an explicit evidence requirement and a reversal path.
Sources in the flagship white paper
- NIST AI Risk Management Framework (AI RMF 1.0) NIST AI 100-1, 2023A framework to govern, map, measure and manage AI risk.
- ISO/IEC 42001:2023 ISO and IEC, 2023Management system standard for AI. No conformity is claimed.
- Zero Trust Architecture NIST SP 800-207, 2020The principle that no trust is granted on network location alone.
- Standards for Internal Control in the Federal Government (Green Book) U.S. GAO, GAO-25-107721, 2025Internal control and segregation of duties. The U.S. agency GAO is unrelated to our framework.
- EU AI Act, Article 14 Regulation (EU) 2024/1689, amended by (EU) 2026/1744Human oversight of high-risk AI systems. No compliance is claimed.
- Investigation of agents’ behavior in the OpenAI / Hugging Face incident METR with Redwood Research, 2026Independent review of how cooperating agents behaved in a real incident.