The idea that humans will increasingly supervise AI agents can sound like an entirely new problem created by generative AI when in fact, it isn't.
Human-factors researchers have spent decades studying what happens when technology becomes capable enough to perform work that a person previously performed directly. The terminology came from environments such as aviation, industrial process control, transportation, robotics and military systems, but the underlying change was remarkably similar to the one now appearing in offices: the human stops executing every step and begins planning, directing, monitoring and intervening in automated work.
That field is known as human supervisory control.
Its history matters because many of the problems now being discovered around AI agents were visible in earlier generations of automation: loss of situation awareness, reduced skill, inappropriate reliance, unclear function allocation, difficulty taking back control and the uncomfortable reality that automation can make the remaining human work harder rather than easier.
AI agents bring those questions into a different operating environment. The supervisor is no longer necessarily a pilot, plant operator or robotics specialist. At any number of companies it could be an accountant, developer, analyst, procurement manager, marketer, security professional or executive.
That makes supervisory control an important intellectual foundation for understanding human-agent work. Let's unpack it here.
Human supervisory control is a model of human-machine interaction in which automated systems perform parts of a task while a human sets goals, provides instructions, monitors performance, intervenes when necessary and learns from the results.
Thomas B. Sheridan was one of the foundational researchers in the field. Supervisory-control research developed around the recognition that increasingly capable computers and automated systems were changing the role of the human operator. Rather than manipulating every control directly, the person increasingly provided higher-level objectives and constraints while automation performed lower-level activity.
This was already an established research problem by the 1970s. A 1976 NATO symposium edited by Sheridan and Gunnar Johannsen described supervisory control in complex human-machine systems where computers were beginning to share significant functions with human operators. Sheridan subsequently developed the model extensively across robotics, transportation, aviation and other automated systems. Monitoring Behavior and Supervisory Control
The basic supervisory-control model describes five human functions:
Plan → Teach → Monitor → Intervene → Learn
In Sheridan's formulation, the person develops a plan, communicates or programs that intent into the automated system, monitors execution, intervenes when performance diverges from what is required and learns from the outcome to improve future interactions. The National Academies' discussion of Sheridan's supervisory-control model provides a particularly useful description of these five functions and the feedback relationships among them. Generic Paradigm for Supervisory Control
It was developed for an earlier technological era, however the underlying architecture looks surprisingly contemporary.
Consider what happens when an employee delegates a meaningful task to an AI agent.
They decide what they want to accomplish. They give the agent instructions, context, constraints and access. They observe what it produces or does. They determine whether the work remains acceptable. They intervene, redirect or stop the system when necessary. Ideally, they incorporate what happened into the next delegation.
The interface is different from an industrial control room, but the behavioral structure is familiar. A knowledge worker using an increasingly autonomous agent is beginning to perform a form of supervisory control.
That change is easy to underestimate because the transition may happen gradually. Someone begins by asking an AI system to draft an email or summarize a document. Later they ask it to research an issue, synthesize evidence, recommend an action, update a system or execute part of a workflow. Eventually, the person may specify the desired outcome while the agent determines much of the process required to reach it.
The technology has moved from assisting the employee with execution toward performing work under human direction.
The employee's job changes with it.
This is the wider operating transition we explore in Human Resilience Engineering for Agentic AI: as agents perform more of the execution, human performance increasingly depends on whether people can delegate appropriately, maintain enough visibility, recognize meaningful deviation and intervene effectively when conditions require it.
One of the most durable findings in the human-automation literature is also one of the most important for today's AI programs.
Automation does not simply subtract tasks from the human role.
It changes the role.
Raja Parasuraman, Thomas Sheridan and Christopher Wickens made this point explicitly in their influential 2000 model of human interaction with automation. They argued that automation can operate across four broad functions:
| Automation function | What the system may do |
|---|---|
| Information acquisition | Gather, filter, prioritize or present information |
| Information analysis | Integrate, interpret, summarize or predict |
| Decision and action selection | Recommend or select what should happen |
| Action implementation | Carry out the selected action |
Crucially, each function can exist at different levels of automation. A system might automate information gathering heavily while leaving decisions to the human, or it might analyze information, choose an action and increasingly execute that action itself.
The authors' central observation was that automation changes human activity and can create new coordination demands for the human operator. Parasuraman, Sheridan & Wickens, A Model for Types and Levels of Human Interaction with Automation
That distinction is highly relevant to AI agents.
An organization's governance problem is very different when AI retrieves information than when it analyzes that information. It changes again when AI recommends a decision, and again when the system is permitted to execute one.
The workforce capability required at each stage changes as well.
Someone supervising automated research needs to understand whether relevant evidence was missed. Someone supervising automated analysis needs enough expertise to challenge interpretation. Someone overseeing AI-supported decisions needs to understand when judgment should remain human. Someone supervising execution needs visibility and intervention mechanisms that still work after the system begins acting.
Talking about an employee simply "using AI" collapses all of those working relationships into one category.
For risk management, that is increasingly inadequate.
Classical supervisory-control research tended to focus on systems where the stakes and automation were obvious: aircraft, nuclear facilities, military systems, transportation networks and industrial processes.
Generative AI democratizes the pattern.
A person may now delegate increasingly complex work through an interface that looks like ordinary business software.
The risk can therefore develop without anyone thinking of the employee as a supervisor of automation.
A procurement professional might use an agent to gather supplier information, then to evaluate submissions, then to rank candidates and eventually to initiate routine follow-up actions. A software developer may move from code suggestions to an agent that investigates issues, modifies files and opens pull requests. A financial analyst may begin with summarization and gradually rely on the system to perform much of the analysis that informs a recommendation.
The job title has not changed.
The distribution of work and judgment has.
Our Guide to AI Workforce Risk Intelligence looks at why that distinction matters. Usage data can tell an organization that employees are using AI; it tells us much less about whether their roles are shifting from execution toward delegation, supervision, verification and exception handling.
Those changes belong in the risk picture.
In 1983, human-factors researcher Lisanne Bainbridge published a paper with a title that has aged extremely well: Ironies of Automation.
Her argument addressed a recurring problem in highly automated systems. Designers automate the tasks that technology can perform reliably, while leaving people responsible for the unusual, difficult or abnormal situations automation cannot handle. Yet those are precisely the moments when the human may be least prepared to take over because routine involvement in the process has disappeared.
Automation can therefore leave the human with less practice and a harder job. Bainbridge, Ironies of Automation
That problem deserves renewed attention as organizations deploy AI agents.
The apparent operating model is attractive: agents perform routine work while humans handle exceptions.
The human-factors question is what happens to the human's ability to handle those exceptions after the agent has successfully performed the underlying task thousands of times.
Expertise is maintained partly through doing.
Context is maintained partly through participation.
Situation awareness develops through contact with the work.
If automation removes enough of that contact, the human may still possess formal responsibility for intervention while becoming progressively less equipped to exercise it.
This is one reason effective human oversight of AI agents cannot be reduced to placing an approval button at the end of a workflow.
The person needs enough capability and context to know when approval would be a mistake.
Research in the 1990s gave this problem another useful name: out-of-the-loop performance.
Mica Endsley and Esin Kiris studied what happens when people supervising automation lose sufficient involvement to respond effectively when the automated system fails. Their 1995 research linked automation with losses in situation awareness and slower recovery when participants needed to resume manual control. The paper discussed several contributing mechanisms, including loss of skill, vigilance problems and the movement from active toward passive information processing. Endsley & Kiris, The Out-of-the-Loop Performance Problem and Level of Control in Automation
An AI workforce can develop an analogous problem without any dramatic failure occurring.
Someone who once performed an analysis may become the reviewer of AI-generated analysis. Over time they spend less time constructing the reasoning themselves and more time evaluating a completed result. If the AI is consistently competent, review can become lighter and the person's detailed familiarity with the underlying work may decline.
Eventually the organization can arrive at an uncomfortable configuration: the human remains responsible for recognizing when the system is wrong precisely because the system's success has reduced the human's opportunities to maintain the expertise required to recognize it.
That is a supervisory-control problem, and it is also a Human Risk Management problem.
The established supervisory-control literature gives us an important foundation, but contemporary agentic work introduces additional requirements.
Sheridan's model can be summarized as:
Plan → Teach → Monitor → Intervene → Learn
For modern enterprise agentic work, Cybermaniacs adapts these supervisory-control and human-automation principles through what we call the HRS Agentic Supervision Lifecycle:
Define → Observe → Detect → Reconcile → Control → Reauthorize → Own
This is not presented as a replacement for Sheridan's model or as a claim that the underlying principles are new. It is an operational adaptation for the particular governance and behavioral problems created when knowledge workers delegate work to generative and agentic systems.
The emphasis changes because enterprise agent supervision has to deal explicitly with changing authority, changing workflows, ambiguous boundaries, evolving agent capability and continuing human accountability.
Effective supervision begins before the agent acts.
The person and organization need sufficient clarity about the objective, task, boundaries, authority, expected human role and conditions under which the work should return to human control.
This becomes particularly important as agents gain access to tools and can perform actions rather than merely generate information. An ambiguous prompt attached to a highly capable agent is materially different from an ambiguous prompt attached to a drafting assistant.
The supervisor needs to understand what has actually been delegated.
A person cannot supervise work they cannot meaningfully see.
Observation includes enough visibility into the agent's progress, actions, outputs and relevant state to understand whether execution remains consistent with the intended task.
This does not necessarily mean exposing every machine-level step to the employee. More information can create its own monitoring problem. The objective is useful observability: enough evidence for the human to maintain an appropriate model of what is happening.
Observation creates the possibility of recognizing deviation.
The person needs to identify when something about the work has moved outside expected conditions: an unusual result, changing scope, inappropriate assumption, unexpected action, higher consequence or evidence that the agent's operating model no longer matches the task.
This is partly a technical design problem and partly a workforce-capability problem. The human must know what meaningful deviation looks like.
Agentic work creates many situations that are not obviously right or wrong.
An unexpected action may be legitimate. New information may require the objective to change. The agent may identify a better approach than the one originally specified.
Supervision therefore requires judgment rather than simple anomaly detection.
Reconciliation is the process of comparing what was intended, what occurred and what the current environment requires before deciding whether the work should continue.
When intervention is necessary, the human needs a usable means of exercising it.
That could mean correcting instructions, changing constraints, withholding approval, narrowing permissions, pausing execution, rolling back an action, escalating the issue or taking over the work.
A theoretical ability to intervene is not equivalent to practical control. If the system acts too quickly, the human lacks sufficient information, intervention is operationally painful or organizational culture discourages interruption, the control may exist primarily on paper.
This step becomes particularly important in agentic systems.
Once something material changes, organizations should not assume that the original delegation remains appropriate indefinitely.
Changes in objective, task scope, system access, decision authority, consequence or autonomy can require a new judgment about what the agent should be allowed to do and what role the human should play.
Our discussion of delegation drift examines why this matters. Delegated work can gradually move away from what the person and organization originally intended without producing an obvious point at which somebody formally changed the arrangement.
Reauthorization creates a deliberate point at which that changed relationship can be recognized.
The lifecycle finishes with accountability and learning.
Someone still needs to understand what happened, whether the outcome was acceptable, what should change next time and where responsibility sits for the result.
This is broader than simply assigning blame when something goes wrong. Ownership allows the organization to learn from agent behavior, improve the workflow, adjust controls, develop workforce capability and decide whether the delegation itself should change.
Without that feedback loop, supervision becomes a sequence of isolated transactions rather than an improving human-agent system.
It is important to distinguish the level at which this model operates.
Technical agent systems may contain sophisticated planning, feedback, reasoning, monitoring and control mechanisms of their own. Agent platforms, IAM systems and runtime-security technologies may observe actions, enforce policies, constrain permissions or stop prohibited behavior automatically.
Those controls are important.
The Agentic Supervision Lifecycle describes the human and organizational supervisory relationship around the agent.
Its questions include whether the person knows what was delegated, can maintain meaningful awareness of the work, recognizes conditions requiring attention, possesses usable intervention authority and revisits the delegation when material conditions change.
This complements the technical control plane rather than duplicating it.
Our Guide to AI Agent Governance and the Missing Human Risk Layer explores that distinction in more detail. Agent identity, authorization, runtime controls and observability can tell an organization a great deal about what an agent is permitted to do and what it actually did. They cannot independently establish whether the human supervisor understood the work, exercised meaningful judgment or retained the capability required to intervene.
Both sides of the system matter.
The Parasuraman-Sheridan-Wickens model also gives organizations a useful way to think about increasing agent autonomy.
Automation does not have one universal "level." It can operate differently across information gathering, analysis, decision selection and execution.
That produces very different human jobs.
An AI system that retrieves documents mainly changes information acquisition. A system that synthesizes them moves into analysis. A system that recommends which supplier to select begins participating in decision selection. A system that then approves the supplier, updates the procurement system or sends instructions begins participating in action implementation.
The employee may be present in every version of the workflow.
Their supervisory burden is not the same.
As automation moves further into analysis, decisions and action, organizations need to revisit what the human must understand, how much attention supervision requires, where intervention can occur and whether the person's retained skill remains appropriate for the responsibility they still hold.
That is why broad labels such as human in the loop can become misleading. They describe presence rather than supervisory effectiveness.
The better question is what the human is expected to do in the loop.
Many AI enablement programs still concentrate primarily on adoption: how to prompt effectively, use approved tools, protect sensitive data and identify obvious limitations.
Supervisory work requires additional capabilities. People need to understand appropriate delegation, establish useful constraints, monitor without drowning in detail, recognize abnormal or consequential behavior, maintain skepticism without reflexively distrusting automation, intervene when conditions warrant it and know when a changed task should require renewed authorization..
This is one reason our Agentic Readiness & Change (ARC) work looks at the workforce, roles, workflows and governance surrounding agents rather than treating readiness as a training event. Agent adoption changes who performs work, who supervises it, which decisions remain human and what capabilities the remaining human role requires.
The technology can be ready while the supervisory system around it is not.
The history of automation gives AI leaders an unusual advantage.
We do not have to discover every human-automation problem from scratch.
Decades of human-factors research have already shown that automation changes attention, skill, situation awareness, workload, reliance, coordination and the nature of the remaining human work. Research has also shown that the human being retained as the final safeguard can become less capable of performing that role if system design removes too much meaningful participation.
Generative and agentic AI change the scale of the problem, the kinds of work involved and the speed at which delegation can expand.
They do not repeal the underlying human factors.
The practical implication is that organizations should examine agent deployments as changes to the work system, rather than simply deployments of new software.
Who defines the objective? What does the agent actually perform? Which judgments remain human? What evidence does the person see? What constitutes meaningful deviation? When can they intervene? What happens when the task or authority changes? What capability does the supervisor need to retain? Who owns the outcome?
Those are design questions.
Waiting until thousands of employees have quietly become supervisors of automation is a poor time to discover that nobody designed the supervisory role.
Supervisory control gives us an important intellectual ancestor for the emerging human-agent workplace.
It also helps explain why agentic AI cannot be governed exclusively as either a technology problem or an employee-training problem.
The relevant unit is increasingly the human-agent working system: the person, the agent, the task, their respective authority, the information available to each, the environment in which they operate and the mechanisms available when conditions change.
Human Resilience Engineering builds on that systems perspective.
The objective is not to keep humans artificially involved in every task agents could perform. It is to identify where human judgment, control and adaptive capacity remain important and then design those responsibilities so people can actually perform them.
That may mean less human involvement in some parts of a workflow and much more deliberate human involvement in others.
The research tradition behind supervisory control gives us several decades of evidence about why that design choice matters - Agentic AI makes it everybody's problem.
Human supervisory control is a form of human-machine interaction in which automated systems perform parts of a task while a human plans objectives, instructs the automation, monitors its performance, intervenes when necessary and learns from the result. Thomas Sheridan and other human-factors researchers developed the field through work on increasingly automated systems beginning decades before generative AI.
People working with AI agents increasingly perform a similar role. Instead of directly executing every part of a task, a person may define an objective, delegate work, monitor the agent, evaluate its behavior and intervene when necessary. This makes supervisory-control research highly relevant to contemporary human-agent work.
Thomas B. Sheridan was one of the foundational researchers in human supervisory control. His work described five central supervisory functions: planning, teaching the automation, monitoring, intervening and learning from experience.
Sheridan's model identifies five major supervisory functions: Plan, Teach, Monitor, Intervene and Learn. The model describes a feedback process in which humans establish objectives and instructions, observe automated execution, intervene when needed and use experience to improve future control.
The HRS Agentic Supervision Lifecycle is Cybermaniacs' adaptation of established supervisory-control and human-automation principles for contemporary knowledge work with generative and agentic AI.
It consists of:
Define → Observe → Detect → Reconcile → Control → Reauthorize → Own
The lifecycle focuses particularly on maintaining effective human control as delegated work, agent authority, task consequences and the human role change over time.
No. Supervisory control has a substantial research history in human factors, automation and engineering. The HRS Agentic Supervision Lifecycle explicitly builds on that intellectual foundation and adapts it to enterprise human-agent work rather than claiming the underlying concept as new.
The out-of-the-loop performance problem occurs when automation reduces a person's active involvement enough that they lose situation awareness, skill or readiness to resume effective control when automation fails or conditions change. Research by Endsley and Kiris demonstrated this problem in automated task environments, and the concept has clear relevance for people expected to oversee increasingly capable AI systems.
Human presence does not guarantee effective oversight. The person also needs sufficient capability, context, authority, attention and opportunity to recognize problems and intervene. As agents become more autonomous, organizations need to define the actual supervisory function rather than simply require a human approval step.
Relevant capabilities can include defining appropriate objectives and boundaries, understanding what has been delegated, interpreting agent outputs, detecting meaningful deviation, questioning recommendations, controlling or stopping execution, escalating uncertainty and retaining enough domain knowledge to evaluate the work.
The exact capability depends on the task, agent authority and potential consequence.
Human Resilience Engineering applies a systems approach to the human contribution in AI-enabled and agentic work. Supervisory control provides an important foundation because it treats people and automation as parts of a joint operating system rather than examining human behavior or automation independently.