For years, we have drawn organizations as boxes.
CEO.
VP.
Director.
Manager.
Engineer.
Designer.
Analyst.
A neat hierarchy.
Humans reporting to humans.
Responsibilities attached to roles.
Accountability attached to names.
That model is beginning to change.
Not because humans are disappearing.
Because something new is entering the org chart without actually being on it.
Agents.
AI that can research.
Write.
Analyze.
Code.
Plan.
Execute.
And increasingly, delegate work to other agents.
The Strange New Team
Imagine you are a manager.
Your team has eight people.
But those eight people collectively supervise thirty AI agents.
Some work continuously.
Some wake up when triggered.
Some make recommendations.
Some execute tasks.
Some check the work of other agents.
Suddenly, your team isn’t eight people anymore.
But the org chart still says eight.
That’s the problem.
**Our organizational structures are human-shaped.
The work is becoming hybrid.**
So Who Is Actually Managing?
This sounds like a technology question.
It isn’t.
It’s a management question.
When an agent makes a mistake:
Who owns it?
When an agent makes a decision:
Who approves it?
When an agent’s output is wrong:
Who notices?
When an agent is given too much autonomy:
Who is responsible?
And perhaps the most important question:
Who decides what the agent should never be allowed to decide?
We Used to Manage People.
Now We May Manage Systems of Agency.
A manager traditionally works through people.
Set direction.
Create clarity.
Remove obstacles.
Develop capability.
Give feedback.
Build trust.
With agents, another layer appears.
You don’t just manage the person.
You increasingly manage the system through which that person works.
The workflow.
The permissions.
The escalation rules.
The review points.
The quality thresholds.
The boundaries of autonomy.
That changes the manager’s job.
It doesn’t make management less important.
It makes management more architectural.
The Danger Is Not Too Much AI
The danger is unclear accountability.
Imagine a workflow where:
Agent A researches.
Agent B analyzes.
Agent C creates the recommendation.
Agent D executes it.
A human reviews the final result.
Something goes wrong.
Who made the decision?
Everyone.
And therefore, potentially, nobody.
This is what happens when autonomy grows faster than accountability.
The system becomes powerful before it becomes governable.
The New Management Skill: Knowing Where to Stand
Recent research describes this as a new management problem: as agents begin managing or coordinating other agents, humans have to determine their “managerial altitude.”
Stand too close and every decision waits for you.
You become the bottleneck.
Stand too far away and errors accumulate silently.
You discover problems only after they become expensive.
This is remarkably similar to managing people.
Except the feedback loops are faster.
And the consequences can scale much faster.
Trust Gets Complicated
We often say:
“We need to trust AI.”
I think that’s incomplete.
The real question is:
What exactly are we trusting?
The model?
The workflow?
The data?
The permissions?
The person supervising it?
The evaluation mechanism?
Recent research from Harvard Business Review identifies trust as a central barrier to meaningful AI-agent adoption. Employees are unlikely to give agents significant autonomy when they don’t understand their reliability, intentions, or behavior.
Trust therefore becomes a design problem.
Not simply an emotional one.
And Humans Are Still the Most Complicated Part
There is an irony here.
The more autonomous AI becomes, the more important human understanding may become.
A recent California Management Review analysis argues that managers need to understand human behavior not only to manage people, but increasingly to manage AI agents effectively. Emerging evidence suggests that skills associated with good human management, including asking questions and engaging in meaningful interaction, also correlate with better AI-agent management.
That’s fascinating.
We thought AI would make management more technical.
Perhaps it will make human understanding more valuable.
The Anxiety Layer
There is another part leaders cannot ignore.
People don’t experience AI as an architecture diagram.
They experience it as a threat to identity.
Their expertise.
Their career.
Their relevance.
Gallup’s recent research found that frequent AI users were more than twice as likely to fear their jobs being eliminated within five years compared with infrequent users. Importantly, supportive management practices were associated with much lower AI-related anxiety.
This means the manager has another job.
Not selling AI.
Not resisting AI.
Helping people make sense of what AI means for them.
The Manager Becomes the Translator
The manager of the future may need to translate in both directions.
To the organization:
Here is what humans still need to own.
To the team:
Here is what AI can now take off your plate.
To the agent:
Here are the boundaries within which you can act.
And sometimes, simply:
Stop. A human needs to decide this.
What Should Stay Human?
I don’t think the answer is:
“Everything important.”
That’s too vague.
The better question is:
Where should human judgment remain the final authority?
Perhaps:
Values.
Accountability.
Relationships.
Trade-offs.
Exceptions.
Consequences.
Meaning.
AI can recommend.
AI can execute.
AI can optimize.
But someone still has to decide what is worth optimizing.
And that may be the most important management question of the AI era.
The Human Layer
We spent decades designing organizations around human limitations.
Now we’re beginning to design organizations around machine capability.
That is going to change more than workflows.
It will change:
Roles.
Hierarchy.
Management.
Career paths.
Trust.
Accountability.
And eventually, our definition of a team.
The org chart may still show eight people.
But the real team may contain dozens of agents working underneath them.
The question is no longer:
“How many people do you manage?”
It may become:
“How much agency are you responsible for?”
That is a very different definition of leadership.
And we are only beginning to understand it.
Leadership Reflection
Ask yourself:
- If AI agents become part of your team, who owns their decisions?
- Where should human approval remain mandatory?
- Are you designing workflows around AI capability or around human accountability?
- Are your managers becoming bottlenecks—or disappearing too far from the work?
- What part of leadership should never be delegated?
The Human Layer Insight
AI may change the size of our teams without changing the number of people on our org chart. The future manager won’t simply manage people. They will manage a system of human and machine agency—and remain accountable for what that system does.
— Majid Nisar
Read the full issue on LinkedIn →
THE HUMAN LAYER publishes on LinkedIn. Subscribe here.