In short: When AI agents take over routine work, the organization changes in three places. Roles shift from execution to setting goals, reviewing results and taking responsibility. Teams get smaller and are built around outcomes rather than functions. And leadership takes on new tasks: setting limits for agents, managing workload and bringing in experienced employees as knowledge holders. These shifts are happening anyway. The question is whether the executive team shapes them.
How are AI agents changing roles in the company?
The Bitkom study “Softwarewelt 2036” describes a shift in the software industry that carries over to many knowledge professions: away from execution and toward setting goals, architecture, context, quality and accountability. What’s needed is judgment, the ability to refine AI output and systems thinking. People increasingly become translators between business, data, regulation and technology.
A product manager, for example, no longer manages a feature list. She owns customer outcomes. An account manager no longer explains how the software works but works with the customer’s metrics. Customer service answers fewer how-to questions and helps customers work with agents. If this shift isn’t shaped, you end up with people being paid and measured by old roles while doing new work.
Why are teams getting smaller?
When everyone works faster with agents, coordination between people becomes the bottleneck.
Large teams then lose part of their speed gain to meetings, handoffs and approval loops. That argues for smaller teams built around an outcome that combine the functions they need, from domain expertise and data to technology and customer contact.
That doesn’t mean large teams will disappear. Existing products and processes still need to be maintained, secured and adapted to working with agents. But the balance is shifting, and the executive team has to decide how fast. A small team for a new initiative needs a mandate framed in outcomes and access to the data and customers of the core business. Above all, it needs the freedom to work differently from the rest of the company. That last point creates tension. Leadership needs an answer that sounds like a deliberate division of labor, not favoritism.
What happens to the span of control?
When agents take over part of the execution, leaders can in theory manage more people, because less operational coordination is needed. At the same time, the share of work that calls for judgment goes up, for example when reviewing results or deciding on exceptions. The result is decision density for those who carry responsibility. Wider spans of control therefore only work if decision rights are clearly delegated and employees are allowed to use their own judgment. Widen spans without shifting decisions and you create bottlenecks at the top.
What role do experienced employees play?
A bigger one than many expect. The Bitkom study describes domain knowledge combined with the ability to steer AI systems as the scarcest skill of the coming years. It recommends prioritizing internal learning over permanent dependence on outside expertise and deliberately involving experienced employees.
The obvious reaction, bringing in AI specialists from outside and letting experienced employees carry on as before, is therefore often the wrong one. The most valuable combination emerges when people with decades of knowledge about customers, processes and exceptions learn to direct agents. The study also notes, however, that many experienced employees are not prepared, technically or psychologically, for such fundamental change. A training program alone isn’t enough. It takes a clear answer to the question of what role their knowledge plays in the new model.
What new roles are emerging?
Forward Deployed EngineersThey combine technology, product and customer knowledge and make sure agents actually work within the customer’s workflows and with the customer’s data.
Agent governance leadsa function that defines and documents what agents may do, who approves and how often this is reviewed.
Value engineers in salesThey translate the customer’s business goals into measurable outcomes and make the impact of AI visible.
What does the executive team have to own itself?
The Bitkom study puts it plainly. AI has to be a top-management priority and can’t be delegated to IT. In concrete terms, the executive team owns four things nobody else can:
- the pace of the transformation,
- the allocation of resources between the existing and the new business,
- the role of experienced employees in the new model,
- the rules under which agents and people work together, including working hours, availability and parallel work when tools keep running around the clock.
That also means using AI yourself. You can only make good decisions about agents if you’ve experienced what they can do, where they fail and what working with them feels like.
Frequently asked questions
Do AI agents make teams smaller?
In many areas, yes, because agents take over routine work and coordination between people becomes the bottleneck. Smaller, cross-functional teams built around an outcome make better use of the speed agents provide.
Which skills become more important with AI?
Judgment, goal setting, quality review, systems thinking and domain knowledge combined with the ability to direct AI systems. Pure execution work loses weight.
Who should own the reorganization driven by AI?
The executive team. Roles, budgets and responsibilities can’t be reorganized from within a single department.
Further reading: Decision density, not more delegation · How much AI expertise do leaders really need? · Leading the transformation
Source: Bitkom (2026): Softwarewelt 2036