Technology makes work cheaper. But what about the organization?
Until recently, organizational design followed a relatively stable logic:
task → capabilities → roles → team
A task required certain capabilities. Those capabilities shaped roles, and roles shaped the team. For decades, this sequence underpinned organizational design and workforce planning.
That logic has changed significantly. When a single specialist working with AI tools can perform an amount of work that previously required several people, there is an obvious temptation to do the math: one employee can handle more tasks — so fewer people are needed.
At the level of an individual operation, this looks perfectly rational. Organizations, however, are more complicated. Teams do more than produce output. They bring different perspectives together, preserve context, surface problems before they become costly, transfer knowledge, and distribute responsibility.
And this raises a question that goes beyond individual productivity: if AI takes over part of what a team used to do, what becomes less valuable within the organization — and what, instead, becomes more valuable?
The question is gradually moving beyond individual functions. It concerns the architecture of the organization itself.
The Illusion of Proportionality
Professional discussions, including InfoQ Culture & Methods Trends 2026, point to a shift from traditional cross-functional teams toward more compact configurations in which one or several specialists work with AI systems and agents. This is particularly visible in engineering, where AI is already changing how work is distributed and how professional roles are defined.
The principle behind team formation is changing as well. Team size is increasingly determined not only by the capabilities required, but also by the level of risk, task complexity, the cost of error, and the degree of human oversight an organization is prepared to maintain.
A new variable has entered the old equation: some capabilities and operations can now be provided by AI. But when individual activities become cheaper, it does not necessarily follow that the function itself becomes cheaper.
The first symptoms are already visible in the market. Klarna brought human involvement back into customer service after a period of aggressive AI automation. Commonwealth Bank of Australia cancelled planned staff reductions after reassessing the initial need for those roles, and similar examples are becoming more common.
Both cases illustrate the same organizational challenge: the consequences of automating individual operations proved more complex than the initial calculation suggested.
An organization is not a collection of individual operations. It is a system of connected capabilities.
Automation can eliminate certain activities while simultaneously weakening a function’s ability to handle exceptional situations, transfer expertise, or make decisions quickly. Reducing automated work therefore does not necessarily produce a comparable reduction in organizational complexity.
The Management Paradox
AI increasingly shifts the organizational bottleneck from production toward management, coordination, and judgement.
When producing an output becomes fast and inexpensive, the constraint moves to what happens before and after execution: defining the task, reconciling competing priorities, checking the result, and taking responsibility for the consequences.
This is already visible beyond IT. Engineering teams are better understood as an early indicator of a broader shift that is gradually reaching other corporate functions. As production approaches real-time speed, organizations generate more outputs, more options, and more parallel workflows. The ability to select the right option, validate it, and integrate it into the wider process does not emerge automatically with AI.
We observe the same logic in our work with clients on process and organizational optimization. When the productive capacity of an individual employee increases sharply, the existing configuration of roles and interactions needs to be reconsidered. Sometimes this genuinely allows redundant layers to be removed. In other cases, the freed-up capacity is better directed toward the areas on which the resilience of the system depends.
As AI makes production cheaper, the value of collaboration increases.
This changes the object of organizational optimization itself. The question is no longer only about headcount or the allocation of functions. It is about the architecture of work: which tasks can genuinely be delegated to AI, which capabilities have become less scarce, and where human involvement creates the greatest value.
Organizational redesign therefore starts with the process, not the headcount plan. First, we need to understand what has changed in the work and which capabilities the organization actually needs now. Only then can we determine how roles, interactions, and structures should be designed.
The old question:
“How many people do we need to do this work?”
is gradually giving way to another:
“What system of collaboration do we need to build so that the increased productivity of each individual becomes sustainable value for the organization as a whole?”
🔗 InfoQ. Culture & Methods Trends 2026: The Human Side of AI Engineering. August 7, 2026.
