Human Governance: Who Designs the Conditions for Success?

Governance has a reputation problem. It is experienced as the brake, the audit, the signature nobody wants to give. Yet the organisations that move fastest with confidence are the ones where people have been given the standing, the competence, and the safety to use their judgement. Human governance is the design of those conditions, and it is what allows human-led organisations to carry strategy at speed.

Two credit officers

Two financial services firms introduce the same kind of model to pre-screen commercial credit applications. Both do the formal work properly: a risk classification, a signed policy, and a named senior credit officer with authority to override the model.

In the first firm, the credit officer notices in the sixth week that the model is scoring one segment on data that no longer reflects the market. She overrides a handful of recommendations and flags the pattern. Nobody asks her to justify herself against the programme's momentum, because using the override is exactly what the role was designed for. The model is recalibrated within a month. The executive team, now confident that problems will surface early, extends the programme to two further business lines ahead of plan.

In the second firm, the credit officer notices the same drift. The CEO has presented the programme to investors as the firm's flagship AI initiative. She raises her concern informally, is told the vendor is looking into it, and does not use her override, because doing so would mean publicly contradicting that commitment on the basis of a pattern she can sense but not yet prove. Months later, a cluster of defaults appears. Accountability lands on her. The programme is frozen.

Same model. Same policy. Same authority on paper. The difference was not the technology, and it was not the person. It was whether anyone had designed the conditions under which she could act.

Same model, policy and authority

The only variable: the conditions

Conditions designed

Override used in week 6

Thanked, not questioned

Model recalibrated in a month

Problem caught while small

Expanded ahead of plan

Governance as driver

Conditions assumed

Concern raised informally

Override never used

Defaults surface months later

Problem caught late

Programme frozen

Accountability lands on her

(C) p-m-z.ch Philip Zollinger

That is what this article means by human governance. And in the first firm, it was not a brake. It was the reason the programme could go faster.

Why governance is cast as the devil

Most people meet governance as something done to them. It arrives as approval chains, compliance checklists, and the sense that someone further up does not trust them with the decision. Leaders experience it as a tax on momentum. It is no surprise that in the AI era, when speed is treated as the decisive advantage, governance is so often positioned as the thing standing in the way.

That reputation is understandable, and it is mostly earned by governance that was designed around distrust: control added after something went wrong, aimed at preventing people from acting rather than enabling them to act well.

But it describes a particular kind of governance, not governance itself. The same word covers the structures that give a junior engineer the right to stop a production line, a first officer the duty to challenge a captain, and a credit officer the confidence to override a model her CEO has championed. Those structures do not restrict human judgement. They are what make it usable.

The question the previous article left open

The previous article argued that strategy will remain human. A strategy is a chain of commitments that binds people who did not make them, and only accountable humans can author those commitments, carry their trade-offs in the open, and generate the willingness that makes them hold.

It ended on a fair objection. If people are the mechanism through which strategy becomes real, people are also where it can fail. Leaders are fallible. They can hear only what suits them. The same leadership states that bring strategy to life can be used badly.

There are two ways to respond to that objection. One is to reduce reliance on people: more automation, more rules, less discretion. That path quietly abandons the argument that strategy is human. The other is to take human leadership seriously enough to design for it: to build the conditions under which people can exercise judgement well, recover when they drift, and be heard when they see something others do not.

This article takes the second path. Human governance is not the answer to the question of how to control people. It is the answer to the question of how to make a human-led organisation work at speed.

Brakes are what let you drive fast

The CCC foundation introduced a principle that runs through this whole series: stability precedes speed. Acceleration amplifies whatever conditions already exist.

There is a simple image for what that means in practice. Nobody would drive a car at motorway speed if it had no brakes. The brakes are not what slows the car down on the motorway; they are the reason anyone is willing to go that fast in the first place. The same holds for architecture. Structural engineering does not limit how high a building can go. It is the only reason a building can go high at all.

Human governance plays that role in an organisation. When people know who can stop something, when they are competent to judge it, and when using that judgement is safe and expected, leaders can say yes sooner and scale further. Confidence travels faster than caution, but only when it is justified by conditions people can rely on.

Conditions are governed by conditions

The CCC framework describes leadership as the design of three conditions under which people can execute reliably: Clarity, Capability, and Climate. Leaders are condition architects.

Applied honestly, the framework turns back on itself. The architect is not standing outside the building. They are inside it, on a floor someone else designed, working under conditions they did not choose. Executives are asked to create psychological safety while working in rooms where dissent can carry career cost. They are asked to build AI fluency while often having the least hands-on exposure to the technology. They are asked to supply clarity while the boundary between board and executive authority on AI remains unsettled.

The architect is also a condition-taker. That is not a flaw in the people. It is a normal property of layered systems, and it points to something generous: the people we ask to design conditions for others deserve well-designed conditions themselves.

This gives the article its signature principle:

Conditions are governed by conditions. Human governance is the deliberate design of the conditions under which people who hold responsibility can exercise it with confidence, be heard when they see what others miss, and return to their best judgement when pressure pulls them away from it.

It helps to separate three layers that are often collapsed into one word. Corporate governance defines who holds authority and who is answerable. AI governance defines how algorithmic systems are developed, deployed, monitored, and stopped. Human governance asks whether the people named in those first two layers have the clarity, capability, and climate to do what those layers expect of them.

Corporate governance

Who holds authority and who is answerable

AI governance

How systems are developed, deployed, monitored and stopped

Human governance

Can the people named in the layers above actually act - especially with AI-enabled work?

Clarity

Standing to act

Capability

Competence to judge

Climate

Safety to use it

(C) p-m-z.ch Philip Zollinger

The first two layers are about the structure and the systems. The third is about the people, and it is the layer that turns the other two from documentation into something that works.

Answering the watchmen question

The oldest question in governance comes from the Roman poet Juvenal:

"Quis custodiet ipsos custodes?"  - Who watches the watchmen?

It is usually read as a counsel of distrust, the start of an endless chain of supervisors supervising supervisors. In a human-led organisation there is a better answer. The watchmen are not best served by another watchman above them. They are best served by conditions around them: a clear mandate, the competence to use it, peers who will speak honestly, and structures that make speaking honestly normal.

Watching people makes them cautious. Designing conditions for them makes them capable.

Every state has a range

Bringing strategy to life requires leaders to move between six states: Attuned, Visionary, Directive, Challenging, Coaching, and Clearing. Each is valuable. Each also has an edge beyond which it stops serving its purpose. We explored these in the previous article.

Attunement can narrow into hearing only what confirms the plan. Visionary can stretch into making a cost sound smaller than it is. Directive can harden into control. Challenging can tip into intimidation. Coaching can turn into keeping people dependent. Clearing can remove friction that was quietly protecting the organisation.

None of this means leaders are dangerous. It means they are human, and humans under sustained pressure tend to get locked into one state. The difficulty is that from the inside, an overextended state feels exactly like the healthy one. A leader drifting into control experiences themselves as being decisive.

That is why self-discipline, though necessary, is not enough on its own, and it is where human governance shows its most generous side. Good structure catches drift early and helps leaders return to the resting state of Attunement, where they can sense what the situation actually needs. It protects the leader's range rather than guarding the organisation against the leader. And a leader who knows that drift will be noticed and named, without drama, is free to be decisive, because decisiveness no longer has to be flawless to be safe.

The human in the loop is a design brief

Regulation already points in this direction, even if it is rarely read that way.

Article 14 of the EU AI Act requires high-risk systems to be designed so that they can be effectively overseen by people, and it describes what those people must be able to do: understand the system's capacities and limitations, stay aware of the tendency to over-rely on its output, interpret that output correctly, decide not to use it, override or reverse it, and stop it. Article 26 requires deployers to assign that oversight to people with the necessary competence, training, and authority. For Swiss organisations whose systems or outputs reach the EU market, this is part of the working environment.

This is usually read as a compliance list. It is better read as a design brief for enabling people. Authority that can be used without asking permission is Clarity. The competence to interpret output and resist automation bias is Capability. An organisation in which overriding the system is acceptable and expected is Climate.

EU AI Act asks for……which is a CCC condition

Authority to override and stop

Without asking permission

→

Clarity

Written into law

Competence to interpret output

Aware of automation bias

→

Capability

Written into law

Safe and expected to use it

No statute can require this

→

Climate

Leadership work

(C) p-m-z.ch Philip Zollinger

Two of those conditions can be written into law. The third cannot. No statute can make a credit officer feel able to contradict a model her CEO has championed. That part is leadership work, and it is exactly where human governance adds value that no regulation can supply.

When those conditions are missing, the result has a name. Madeleine Clare Elish called it the moral crumple zone: responsibility for a system's failure collapses onto the nearest human operator, protecting the technology at that person's expense. The crumple zone is not what governance creates. It is what forms in the absence of human governance, when people are placed in the loop without the conditions to act there.

Two industries that learned to trust people more

The most convincing evidence that governance can be a driver comes from two industries that learned, at great cost, to give people more standing rather than less.

The first is aviation. In the late 1970s, a series of accidents, including the 1977 collision at Tenerife, showed that experienced crews could fly into disaster even when someone in the cockpit had sensed something was wrong. The problem was not a lack of skill. It was a hierarchy in which challenging the captain was hard. The industry's response became known as Crew Resource Management. It made speaking up a defined duty, trained crews in how to challenge and how to receive challenge, and turned the cockpit into a team rather than a chain of command. It did not weaken the captain's authority. It made that authority better informed. Crew Resource Management has since spread well beyond aviation, into surgery and other high-stakes fields.

The same industry also shows what happens when those conditions erode. In its 2020 investigation into the two Boeing 737 MAX crashes, the US House Committee on Transportation and Infrastructure described production pressure, oversight delegated to people inside the company being overseen, and pilots who had not been told about the automated system they would ultimately have to counter. The lesson is not that aviation's approach failed. It is that enabling conditions have to be maintained, and that they erode quietly when commercial pressure is allowed to outrank them.

The second is manufacturing. Toyota's production system gave every worker on the line the ability to stop production when they saw a problem, through the andon cord. To many managers at the time this looked like an invitation to chaos. In practice it became one of the most studied sources of quality and learning in industrial history. Stopping was cheap, expected, and followed by help rather than blame. Problems surfaced when they were small, so the line could run with confidence.

Both cases share the same design. Authority to act was pushed to the people closest to the signal. Using it was made normal. And the organisation became faster and safer at the same time.

That is the model for human governance in the age of AI.

Where enabling conditions matter most

These conditions matter at every level, but they matter most where power and pace are greatest, because that is where the natural signals are weakest.

The current evidence suggests the executive layer has room to grow. Research by The Conference Board on S&P 500 disclosures found that the share of companies naming AI as a risk rose from 12 to 83 per cent between 2023 and 2025, while disclosed AI expertise among directors rose only from 1.5 to 2.7 per cent over a longer period. Fewer than a quarter of the executives surveyed considered their board highly fluent in AI.

Read positively, this is not an indictment. It is an opportunity. Boards and executive teams that build their own capability to evaluate AI, rather than simply admire it, will be able to approve with confidence where others hesitate.

Clarity at the top brings the same opportunity. In many organisations the question of who can stop a material AI system has a distributed answer, which in practice means nobody holds it. Initiatives drift, because no one person can say no, and therefore no one can say yes with full conviction either. Swiss company law already offers an anchor. The Code of Obligations gives the board of directors duties it cannot delegate, including the ultimate direction of the company and the supervision of those entrusted with management. Supervision that includes the competence to understand what management is deploying is what turns that duty from a formality into a source of confidence for the whole organisation.

Climate at the top is the least measured and perhaps the most powerful lever. Organisations survey engagement and psychological safety across their workforce, yet rarely inside the executive team or the board. In the AI era there is a specific pressure worth naming: the fear of appearing behind. An executive team whose climate allows someone to say "we are not ready yet" can adopt faster when it is ready, because it trusts its own assessment.

For owner-managed companies without a formal board layer, the opportunity is even more direct. The owner is often the architect, the approver, and the only person with the standing to overrule either. A small number of deliberate conditions, such as a trusted sparring partner with a real mandate to challenge or an agreed rule for when a decision is paused, can give that person the kind of support that larger organisations get from their structures.

Where the conditions come from

If the architect is inside the building, their conditions have to come from somewhere. In practice there are four sources.

The first is the layer above. Boards shape conditions for executives, and executives for their teams. This works well when the higher layer has the competence to shape meaningful conditions.

The second is the peer group. An executive team is itself a team, with the same needs as any other: shared understanding, trust, and constructive conflict. Treating it as a working system rather than a decision forum is one of the highest-leverage moves available to a CEO.

The third is structure: mechanisms that do not depend on the mood of the day. Defined stop authority, separated roles, protected dissent, independent review. Structure is what allows relationships to stay honest under pressure, because honesty does not have to be improvised each time.

The fourth is the self: attention, energy, and the willingness to be contradicted. Essential, and most effective when it is supported rather than left to carry everything alone.

At the top of an organisation, the first two sources naturally thin out, which is why structure should carry more weight the higher you go. That sounds restrictive. In practice it is freeing. Senior leaders who can rely on well-designed structure no longer have to be heroic to be safe.

Human governance as enabling mechanism

Human governance becomes real when it is expressed as mechanisms, each designed around what it makes possible:

• Name the person who can stop, so others can go. For every material AI system, one individual holds stop authority and knows it. When risk has a clear owner, everyone else can move faster.

• Make the override cheap. Using it should need no upward approval and carry no personal cost. Agree in advance that a good-faith override will be thanked, not questioned. This is the andon cord for AI.

• Invest in oversight competence. Define what the overseer needs to be able to do and help them build it, refreshing it whenever the system changes. Capability is what turns authority into confidence.

• Separate advocacy from assurance. Let one person champion adoption and another assess readiness. Each role is then free to do its job fully, without having to argue against itself.

• Care for climate at the top. Use the same instruments on the executive team and the board that are used across the workforce, and treat the results as a resource for keeping leadership sustainable.

• Give dissent a route that leads somewhere. A credible concern about an AI deployment should trigger a defined review by someone outside the sponsor's reporting line. This is how the weak signals people sense but cannot yet prove finally reach the people who can act on them.

Each of these gives someone standing, competence, or safety. None of them assumes people are the problem. All of them assume people are the solution and deserve conditions to match.

Good friction and bad friction

There is also a structural reason this matters more now.

In a purely human organisation, weak conditions announce themselves. Unclear strategy produces questions. Missing capability produces visible struggle. A poor climate produces attrition. This friction is uncomfortable, but it carries information, and it has always arrived for free.

Non-social agents do not generate it. An algorithmic agent does not ask which objective wins when two conflict, does not push back on an unclear instruction, and does not resign. It carries out an unclear mandate quickly, at scale, and without complaint.

This changes what the Clearing state asks of leaders. Removing friction remains one of the most valuable things a leader can do. But not all friction is waste. Some of it is how an organisation notices that something is wrong. Human governance helps leaders tell the two apart, so they can clear the bad friction boldly, knowing the good friction has been deliberately designed back in.

Making sure it is real

A fair concern about any governance argument is that it can turn into paperwork: named owners who are only signatures, dissent routes that are only forms, climate surveys that nobody reads.

The safeguard is simple. Enabling conditions prove themselves through use. An override that has been used, and whose user was thanked, builds more trust than any policy. A dissent route that once changed a decision tells everyone it is real. A climate conversation that changed how an executive team works is worth more than any score.

So the most useful habit in human governance is not writing more rules. It is looking for evidence that the conditions have been exercised, and celebrating it when they have.

Stability precedes speed, starting at the top

Conditions at the top travel down through every decision they shape, and that works in both directions.

When Clarity is strong at the top, people across the organisation know who decides and can act without waiting. When Capability is strong at the top, new technology is approved with understanding, so it can be scaled with confidence. When Climate is strong at the top, the organisation learns that honesty is valued, and information arrives while it is still cheap to act on.

Human governance is how condition architects give themselves the same foundations they are asked to give everyone else. It is what turns human leadership from a risk to be managed into the driver of a fast, reliable organisation.

What to check this week

Take one live AI decision your organisation is making. Not a policy. A decision, with money and consequence attached. Then ask four questions.

Who, by name, can stop it, and do they know they can? Can that person evaluate the system, or only receive its output? What would it cost them to use their authority tomorrow, and how could that cost be brought close to zero?

And the fourth question, the one that shows whether governance is working as a driver: what has that person's authority already made possible? A faster recalibration, a wider rollout, a problem caught while it was small?

If you can answer the fourth question, your governance is enabling your organisation. If you cannot, you have found the place to start.

Leaders design the conditions of execution. Human governance is how they receive well-designed conditions in return, so that people can lead with confidence, speak with safety, and carry strategy at the speed the age of AI demands.

Note

This article is part of a series extending the CCC framework (Clarity, Capability, Climate) introduced in The Core Conditions for Organisations in the AI Era, and follows Why Strategy Will Remain Human. It leaves one question open for the next piece: if friction no longer reports on the state of our conditions, how can organisations sense them continuously and use that sensing to move faster? From there the series leads into AUREA™ — the Strategic AI Operating Blueprint, and later into what I currently call The Healthy Organisation and The Continuum Organisation.

Selected References

Edmondson, A. (1999). Psychological safety and learning behavior in work teams. Administrative Science Quarterly, 44(2).

Elish, M. C. (2019). Moral crumple zones: Cautionary tales in human-robot interaction. Engaging Science, Technology, and Society, 5, 40–60.

European Union (2024). Regulation (EU) 2024/1689 (Artificial Intelligence Act), Article 14 (Human oversight) and Article 26 (Obligations of deployers).

Galbraith, J. (1977). Organization Design.

Helmreich, R. L., Merritt, A. C., & Wilhelm, J. A. (1999). The evolution of Crew Resource Management training in commercial aviation. International Journal of Aviation Psychology, 9(1).

ISO/IEC 42001:2023. Artificial intelligence — Management system.

Liker, J. K. (2004). The Toyota Way.

Meadows, D. (2008). Thinking in Systems.

NIST (2023). AI Risk Management Framework 1.0.

Parasuraman, R., & Manzey, D. (2010). Complacency and bias in human use of automation. Human Factors, 52(3).

Salas, E., Sims, D., & Burke, C. S. (2005). Is there a "Big Five" in teamwork? Small Group Research, 36(5).

Schneider, B., Ehrhart, M., & Macey, W. (2013). Organizational climate and culture. Annual Review of Psychology, 64.

Snowden, D., & Boone, M. (2007). A leader's framework for decision making. Harvard Business Review.

Swiss Code of Obligations, Art. 716a (Non-transferable duties of the board of directors).

U.S. House Committee on Transportation and Infrastructure (2020). Final Committee Report: The Design, Development & Certification of the Boeing 737 MAX.

Additional data sources

The Conference Board (2026). Governing AI: Board oversight and disclosure in the S&P 500.

Deloitte Global (2025). Governance of AI: A critical imperative for today's boards.