1. Research Question
If execution authority is migrating from humans to systems, where is governance?
2. Research Context
The preceding experiments had established a shift in the object of inquiry. The question was no longer only what artificial intelligence could do, but what happened when increasingly capable systems acquired the capacity to produce consequences in the world.
Money could move. Access could be granted. A candidate could be rejected. A transaction could be blocked. A resource could be allocated. A process could begin. A machine could move. A system could change.
These were not merely computational outputs. They were exercises of authority.
I began to think about authority in a particular way. Authority is the power to produce consequence.
Historically, that capacity had overwhelmingly resided with humans and the institutions humans created. A board approved a decision. An executive authorised expenditure. A manager allocated work. A compliance officer restricted an activity. A bank released funds. A government official issued a licence.
Authority had an address.
The Infrastructure Condition complicated that architecture. If systems could interpret conditions, make decisions and execute actions without requiring human intervention at every point, then authority itself could begin to change location. Execution authority was migrating from humans to systems.
This brought my professional experience directly into the research.
I had worked as a Chief Financial Officer and Chief Compliance Officer. I had spent years working with controls, delegations of authority, approval limits, risk management, accountability, segregation of duties, policies, compliance processes, boards, audit committees and financial controls.
These mechanisms exist because organisations have long understood that humans cannot be given unlimited authority to execute. Authority is therefore allocated and constrained.
A person may be authorised to approve expenditure up to a particular amount, while another level of authority is required beyond that threshold. Certain transactions require two signatories. Particular decisions are reserved for the board. Some activities are prohibited entirely. Transactions are recorded so that they can be reviewed. Exceptions are escalated when they fall outside established parameters. Actions are audited so that accountability can be established.
An organisation therefore establishes who may act, what they may do, under what circumstances they may do it, what limits apply to their authority and who will oversee the consequences.
Generations of governance architecture have been developed around human authority.
This produced the central problem for the experiment. If execution authority was moving into systems, where were the equivalent governance arrangements?
3. Research Method
The research approached the question through a combination of professional experience, governance theory and examination of the changing relationship between authority and execution.
The starting point was the governance tradition associated with Professor Mervyn King, whose work had significantly shaped my understanding of governance before this research began. I approached his work not simply as a source of definitions, but as an intellectual foundation for understanding how organisations exercise authority, create value, maintain control and remain accountable within wider systems of relationships.
King’s contribution to governance provided an important frame for the experiment because it expanded governance beyond compliance and formal corporate structures. His work encouraged an understanding of organisations as interconnected systems whose activities affect and depend upon stakeholders, society, resources, institutions and the environment.
That perspective made it possible to examine the technological system not merely as a tool used by an organisation, but as part of the environment through which organisational authority increasingly operates.
The research then examined the delegation chain through which authority traditionally moves. The board delegates to executives. Executives delegate to management. Management delegates to employees. Employees execute within defined mandates, policies and controls.
The experiment compared that established architecture with an emerging execution chain in which a human organisation can delegate activity to software, automated workflows, artificial intelligence systems, agents, APIs and interconnected infrastructure.
The purpose was not to determine whether systems possess legal authority equivalent to humans. The purpose was to identify whether the operational location at which authority produces consequences is changing, and whether existing governance arrangements adequately account for that change.
4. Research Findings
The research found that the central governance problem created by artificial intelligence is not simply the capability of systems to make decisions. It is the increasing capacity of systems to execute decisions and produce consequences.
The distinction is important.
A system that generates an answer is performing computation. A system that can use that answer to move money, grant access, reject an application, initiate a process, allocate a resource or alter another system is participating in execution.
Execution introduces authority.
The governance architecture I had known throughout my professional career had developed around organisations in which the actors exercising delegated authority were ultimately human. The board did not execute everything itself. It delegated responsibility to executives. Executives delegated responsibility to managers. Managers delegated activities to employees. Employees acted within mandates, policies and controls.
Governance followed those delegations.
Policies established what could be done. Mandates established who could do it. Approval limits constrained how much authority an individual could exercise. Reporting structures created visibility. Committees provided oversight. Assurance mechanisms tested whether controls were functioning. Audit trails allowed actions to be reconstructed. Accountability connected consequences back to the person or body responsible for the authority that had been exercised.
The chain could become complicated, but its basic unit remained remarkably stable.
The human remained the actor at the point of execution.
The Infrastructure Condition changes this arrangement.
A board can delegate responsibility to an executive. The executive can delegate responsibility to management. Management can authorise a system to perform a process. That system can invoke an artificial intelligence agent. The agent can interpret information and determine an action. An API can transmit that action into another system. That system can execute it. The resulting action can trigger another automated process.
Eventually, a consequence occurs.
The human organisation may still originate the mandate. The governing body may still retain ultimate accountability. Management may still retain responsibility.
However, the operational path between mandate and consequence can increasingly pass through technological systems.
This does not mean that a system suddenly possesses legal authority equivalent to a director, executive or employee.
It means something more precise.
The operational location of execution is changing.
This revealed a distinction between the formal location of accountability and the operational location of execution.
Accountability may remain assigned to a human or institution while the execution that produces the consequence increasingly occurs through systems.
The problem is therefore not necessarily that accountability has disappeared. The problem is that the architecture connecting authority to execution and execution to accountability is becoming more complex.
The research also confirmed the continuing relevance of the governance tradition associated with Mervyn King.
King’s work does not become less relevant because execution is becoming technological. It becomes more important because its central insight is that governance must respond to the actual organisation and the system in which that organisation operates.
If the organisation changes, governance must be capable of seeing the change.
If the relationships through which value is created change, governance must account for those relationships.
If the mechanisms through which authority is exercised change, governance must account for those mechanisms.
The emergence of technological execution therefore does not sit outside governance. It becomes part of the governance object itself.
This led to the central finding of the experiment.
Governance has to follow authority.
5. Research Implications
If execution authority is migrating from humans to systems, governance cannot remain designed exclusively around human execution.
The traditional mechanisms of governance remain necessary. Approval matrices, financial controls, delegations, policies, audit trails, segregation of duties, boards, risk management and compliance remain important.
However, they can no longer be assumed to provide sufficient visibility merely because a human remains formally accountable.
Governance must increasingly account for where execution actually occurs.
This means understanding what systems are authorised to do, what conditions activate that authority, what resources they can access, what other systems they can instruct, what constraints operate on them, whether intervention is possible, how actions are recorded and how consequences can be traced back through the execution chain.
The implication is larger than artificial intelligence governance.
As execution becomes distributed across software, infrastructure and interconnected systems, governance itself has to become capable of governing distributed execution.
This also changes the meaning of control.
A control that exists only as a policy may be insufficient if the system capable of producing the consequence operates faster than the human responsible for reviewing it.
A delegation that is formally approved may be insufficient if its practical scope is not visible.
An audit trail may establish what happened without establishing whether meaningful intervention was possible before it happened.
Accountability may remain formally assigned while practical control has migrated elsewhere.
The governance question therefore moves from simply asking who is accountable to asking whether the architecture of authority, execution, intervention and accountability remains connected.
This is where the research begins to move beyond conventional AI governance.
The question is not simply whether humans remain responsible.
The question is whether humans still possess sufficient visibility, authority and capacity to govern the execution for which they remain responsible.
6. Conclusion
The experiment began with a simple question: if execution authority is migrating from humans to systems, where is governance?
The answer was not that governance has disappeared.
The answer was that governance has not yet fully followed the migration of execution authority.
Mervyn King gave me a way of understanding governance as an architecture of relationships, responsibility, control, accountability and legitimacy. The Infrastructure Condition revealed that the architecture through which execution occurs was changing.
The significance of this shift is that governance can no longer be understood only through the human actors who formally hold authority. It must also account for the systems through which authority is increasingly exercised.
Execution Authority became the bridge between these two bodies of thought.
The question was therefore no longer simply how artificial intelligence should be regulated.
The question became how execution authority should be governed when execution is no longer exclusively human.
That led directly to the next experiment.
What happens when authority moves faster than governance can follow?
7. References
No formal references are included in this experiment. Mervyn King is treated here as an intellectual foundation for the research rather than as a source being formally cited.