Self-Governance Before AI: Why Human Coherence Has to Come First

Most organizations are trying to govern their AI before they have governed themselves, and it shows. The governance conversation almost always starts in the same place — a policy for the model. Which tools are approved, what data can go in, who signs off, what the audit trail looks like. All of that matters. None of it holds if the people deploying it are not coherent. A fragmented leadership team does not become aligned because it bought a governance framework. It automates its fragmentation faster.

Arnaud Saint-Paul, founder of the Heart Leadership Institute at Tapuat, put a sharp point on this. He took his Heart Leadership Index — a structured read on leadership coherence across seven dimensions — and ran it through several of the leading large language models. The models disagreed. Same instrument, same questions, materially different results, because each model carries the assumptions and weighting of the people who built it. His conclusion is not that the models are broken. It is that bias is baked in everywhere, human and machine, and the organizations that pretend otherwise are the ones most exposed.

At Hutchins Data Strategy Consultants we see the same pattern from the operations side: the enthusiasm for deploying autonomous systems is highest exactly where the underlying organization is least prepared to govern them.

When Everyone Governs the Model and No One Governs Themselves

If the model reflects its makers' assumptions, and your team reflects its own unexamined ones, then putting AI on top of an incoherent organization does not take judgment out of the loop. It encodes whatever judgment was already there and runs it faster. Self-governance is the work of getting the human operating state right before a system starts executing it at scale.

That is the layer the standard governance program skips. Frameworks for the model are necessary, but they sit on top of a question they rarely ask: is the organization itself coherent enough to be trusted with this much acceleration?

What Self-Governance Actually Means

In practice it means treating leadership readiness as a precondition, not an afterthought. Can the team state plainly what it actually optimizes for? Do the people closest to the work trust leadership enough to flag where an AI output is wrong? Is there an honest, shared picture of where the organization is fragmented — between stated values and real incentives, between the dashboard and the decision that gets made anyway?

Those are not soft questions. They decide whether an AI program surfaces better decisions or launders bad ones. This is continuous with the work on leadership trust and just-culture governance that any serious program depends on — the difference is that AI raises the cost of getting it wrong.

The Healthcare Stakes

Healthcare makes the abstraction concrete. A patient now arrives holding an AI-generated workup, and the clinician has to decide what to do with it — often without seeing the prompt that produced it, the data it leaned on, or where it is confidently wrong. The governance question there is not only whether the model is approved. It is whether the clinician has the time, the context, and the institutional backing to override it when their judgment says otherwise.

An AI tool that does not give the clinician time back, or that quietly erodes their authority to say no, fails the coherence test regardless of its benchmark scores. That is why human oversight of clinical AI cannot be bolted on after deployment — it has to be designed into how the organization makes decisions in the first place.

How Hutchins Approaches Self-Governance and AI Readiness

Our work sequences the governance the right way: the humans, then the data, then the model — not the reverse. Before an organization scales, we help it establish a coherence baseline — a clear read on where leadership is aligned and where it is not — so it knows which decisions are safe to accelerate and which need to stay slow and human. Saint-Paul's work, including his book Heart Leadership: Fostering Coherence to Accelerate Impact, is one way to make that baseline measurable; the principle holds regardless of the instrument.

The uncomfortable version is that most AI governance failures are not technology failures. They are leadership failures that the technology made visible. These questions run throughout The Signal Room podcast, where leaders describe what actually carried the weight when they moved from experimentation to scale.

Authoritative sources

• NIST AI Risk Management Framework
• WHO — Ethics & governance of AI for health
• OECD AI Principles