There is a sentence I recently read in an article about the state of the technology industry that has stayed with me:
“There’s always another agent that waits on your response.”
It perfectly captures how I feel in this new agentic era.
Until recently, the amount of work that could be performed by a team was constrained by several very human factors. People needed to sleep. They had working hours, weekends, holidays, families, illnesses, moments of inspiration and moments of exhaustion. Even the most committed team eventually went home.
Artificial intelligence agents do not.
They can work at night, during lunch, while we are attending a meeting or while we are working on something entirely different. One agent can analyze a problem while another writes code, a third reviews documentation and a fourth investigates an alternative approach. Provided that the work has been organized properly, something can always be moving forward.
This does not mean that we must pay constant attention to them. Quite the opposite. One of the main advantages of agents is precisely their ability to operate with a certain degree of autonomy. We can assign a task, provide the necessary context, define the expected result and return later to evaluate what has been done.
In theory, this should make our working lives easier.
In practice, it introduces a new and rather subtle form of pressure.
The team that never goes home
The virtually unlimited availability of agents means that work can now be distributed across twenty-four hours a day, 365 days a year, without demanding our continuous presence.
If we organize ourselves well, there can always be someone working.
A task assigned before dinner may be completed by the following morning. While we review its results, another agent can investigate the problems that emerged. Once those problems have been clarified, yet another agent can implement the corrections. The cycle can continue almost indefinitely.
This is an extraordinary capability.
It is also where the trouble begins.
When the people around us have limited working hours, inactivity is normal. If a colleague has finished the tasks that could reasonably be completed that day, nobody considers the remaining hours of the night to be wasted productive capacity.
With agents, the situation feels different. An idle agent can appear to be a missed opportunity. If it is ready to work and is doing nothing, perhaps the problem is not the agent. Perhaps the problem is that we have failed to prepare its next assignment.
There is always another task that could be delegated, another investigation that could be launched, another document that could be reviewed, another possible solution that could be explored.
And there is always another agent waiting for our response.
From worker to team leader
Once we begin using agents seriously, our role inevitably changes.
We may still write code, design products, analyze data or prepare documents, but a growing part of our contribution no longer consists of performing the work directly. Instead, we define objectives, divide problems into manageable tasks, provide context, establish constraints, evaluate results and decide what should happen next.
In essence, we become team leaders.
Anyone who has held this responsibility in a professional organization will recognize the pattern. A good team leader does not merely distribute tasks. Their most important responsibility is to create the conditions in which other people can perform at their best.
Team members need clear objectives. They need access to the relevant information. They need to understand how their work fits into the broader mission. They need boundaries within which they can act autonomously. They need feedback that is specific enough to help them improve. Above all, they need someone to remove the obstacles that prevent them from making progress.
Agents need much the same things.
They perform poorly when objectives are ambiguous, context is fragmented or success has not been defined. They waste time when they lack access to the right tools and information. They make locally reasonable decisions that are globally wrong when nobody has explained the larger architecture, strategy or purpose.
The quality of their work depends heavily on the quality of the environment we build around them.
This shifts our attention away from a familiar question:
How much work can I complete today?
It replaces it with a more unsettling one:
How effectively can I enable my team to work without me?
The agentic syndrome
This question can easily become an obsession.
If agents are constantly available, should we always have work ready for them?
Should we spend the final minutes of every day preparing tasks for the night? Should every completed activity automatically generate a queue of follow-up actions? Should an agent ever be idle when there are still unresolved problems, unexplored ideas and unfinished projects?
This is what I call the agentic syndrome: the persistent feeling that we are not making full use of our artificial team.
It is the suspicion that somewhere in our digital organization, productive capacity is being wasted because we have failed to provide the right prompt, the right context or the next sufficiently well-defined task.
The syndrome presents itself as a question:
Am I good enough at supporting my team members to allow them to work twenty-four hours a day?
Notice how easily the promise of automation can turn into a new source of anxiety.
AI was supposed to relieve us of work. Instead, it can confront us with the possibility of work that never needs to stop. It removes one of the natural boundaries that once protected us: the unavailability of other people.
The office used to close. Colleagues went home. Messages stopped arriving. Progress had to wait until morning.
The agentic workplace has no closing time.
The bottleneck moves upward
When execution becomes abundant, direction becomes scarce.
This may be one of the most important consequences of agentic AI. The bottleneck does not disappear. It moves. We are the new bottleneck.
Writing the code may no longer be the limiting factor. Deciding what the code should do may be.
Producing a document may take minutes. Determining what deserves to be said, to whom and for what purpose remains difficult.
Exploring ten alternatives may become inexpensive. Understanding which trade-offs matter is still our responsibility.
Agents increase the speed at which decisions can be converted into artifacts. But this also increases the demand for decisions. The faster the team can execute, the more frequently it returns to us for direction.
A poorly prepared decision that once affected a few hours of personal work may now misdirect several agents simultaneously. Conversely, a clear objective, a well-designed process or a carefully written piece of context can enable an entire artificial team to operate productively for hours.
This makes leadership more valuable, not less.
It also makes our weaknesses more visible.
Ambiguous thinking produces ambiguous instructions. Unclear priorities generate impressive but irrelevant results. An inability to distinguish what is urgent from what is important creates a very efficient machine for accelerating confusion.
Agents do not merely amplify our productivity. They amplify the quality, or the inadequacy, of our organization.
Utilization is not necessarily the objective
Is an available agent that remains idle evidence of a failure on our part? Perhaps it is. If productive capacity is available and meaningful work remains to be done, failing to use that capacity may reveal a weakness in how we define tasks, communicate objectives or organize the workflow.
But is that always the case?
We inherited from industrial production the idea that an idle machine represents lost output. Yet knowledge work is not a factory, and it is unclear whether agents should be treated as production lines whose utilization must constantly approach one hundred per cent.
Should the objective be to keep every agent busy, or to make meaningful progress?
Sometimes the best decision may be to launch five parallel investigations. At other times, it may be better to stop and understand what the first investigation has taught us. Another implementation may open a valuable path, or it may merely create one more result that requires attention and review.
Generating work is easy. Determining whether that work is useful is much harder.
An agent waiting for a task may represent wasted potential. Its inactivity, however, may also indicate that no further execution is currently needed, or that the next bottleneck is human judgment, conversation, observation or simply time.
So how should we assess good leadership in an artificial team? By how continuously its members work, or by whether their work serves a coherent purpose? And are these objectives necessarily in conflict?
Perhaps learning to lead agents means living with this unresolved tension: idle capacity may be a missed opportunity, but keeping every agent occupied may not be the same as moving forward.
Learning to leave agents idle
The abundance created by agents requires a new kind of discipline.
We must learn how to formulate better tasks, provide better context and build systems that allow useful work to continue without constant intervention. But we must also learn when not to delegate another task simply because doing so is possible.
The challenge is twofold.
First, we must become better leaders. We need to clarify intentions, establish priorities, design feedback loops and create environments in which agents can act autonomously without drifting away from the objective.
Second, we must resist the idea that permanent availability creates a duty of permanent utilization.
Otherwise, the technology that promised to give us more time will merely expand work to occupy every hour it has liberated.
There will always be another agent waiting for our response. There will always be another prompt we could write before going to sleep, another process we could start over the weekend, another opportunity to ensure that progress continues while we are elsewhere.
The agents will not become tired.
But we will.
Perhaps the ability to make agents work around the clock is not the ultimate test of the agentic era. Perhaps the real test is whether we can benefit from their tirelessness without adopting their inability to stop.
The most effective leader may not be the one whose artificial team never rests.
It may be the one who knows when there is nothing worth asking it to do.
Credits
The Pragmatic Engineer, The state of the tech industry in 2026, Gergely Orosz
