For management and day-to-day operations
Not more AI agents, but fewer human robots
Why companies that talk about AI agents should first ask whether their organization allows people to take ownership.
Most companies don’t need more AI agents. They need fewer human robots. This is not a rejection of reliable execution, and I want to make that clear right at the start. The title is pointed. What I mean is a contradiction I see in the discussion about AI in companies, and it concerns how much independence people have in daily work.
Isn’t reliable execution exactly what a company needs?
Yes. A company cannot consist only of people who question everything, reinvent every process and constantly run in their own direction. There is real value in people who are loyal, disciplined and precise and who execute well.
The problem begins when execution becomes the entire operating model.
How can you tell that execution has become the whole operating model?
By situations that look alike in many organizations:
- The safest person in the room is the one who agrees.
- The fastest career path goes through avoiding friction.
- “Alignment” quietly means in practice: do not question the plan.
- Ownership is welcome only until it leads to disagreement.
In such situations, people learn that disagreement has a cost and agreement does not. What emerges is an operation that executes very well and shows little judgment.
Where is the contradiction?
In exactly these companies, people then talk about AI agents, autonomous workflows, robotics, software-defined products and transformation. That is the contradiction: machines are expected to act independently, while people in the same company are not welcome to.
What is meant by independence?
Independence does not mean rebellion. It means the ability to understand a context, form a judgment, take responsibility and keep going under uncertainty.
In other words: independence is not a property a company can buy at the push of a button. It grows in daily work, in meetings, in decisions and in how disagreement is handled.
This can create friction. Sometimes it slows down a meeting. Sometimes it questions a manager’s favorite idea. Without this ability, though, no adaptable company emerges. What emerges is a polite execution machine.
Is an execution machine bad?
No, it can be useful. It completes tasks. It protects stability. In reports it can even look healthy. Its limit shows when the market changes faster than the task list. A task list describes what was considered right yesterday. When the situation shifts, you need people who see that the list no longer fits and who are willing to say so.
What is the core in one sentence?
Anyone who expects independence from AI agents should check whether people in their own company are allowed to show the same independence. Otherwise the expectation of the technology and the expectation of the workforce do not fit together.
What follows for the use of AI?
Anyone thinking about AI agents, autonomous processes or robotics can first ask how much independence is possible in their own company today. Who is allowed to disagree here, and what happens then? Who takes responsibility when a plan no longer fits the situation?
These questions concern leadership and culture, and they come before choosing tools. A company that has answered them can build on that. A company that leaves them open gets, with new technology, mostly more execution but not more judgment.