When AI takes over tasks, the mandate of HR changes

The general AI debate has by now thoroughly established that automation changes individual tasks rather than simply replacing entire professions. For HR, the real work begins one step later: what happens to a role when exactly those tasks disappear on which people build experience, form relationships and practise responsibility?

The focus therefore shifts from the automation rate to role architecture. HR must clarify which work a system may carry out, which preparatory work people must assess, where relationship and situational judgement remain indispensable, and how junior staff will still get from knowledge to experience at all.

This article therefore begins where the technical automation question ends: at workforce planning, recruiting, performance management and HR's responsibility to actively help shape the new division of work between people and AI.

A task is not a role

A professional role never consists merely of a list of tasks. It combines routine, expertise, relationships, judgement, escalation, learning and responsibility.

That is exactly why the statement "40 percent of this position is automatable" is only the starting point for HR. What matters is which 40 percent is meant. Is it repetitive tasks? Is it the activities junior staff learn from? Is it situations where customer relationships are formed? Or is it preparatory work that a person must later assess?

Anyone who only looks at time saved does not see the whole role. HR must therefore think more strongly in terms of task architectures: what can AI carry out, what may it prepare, what must a person assess, and where must responsibility remain strictly human?

The central HR question is not: which position can AI replace? It is: which components of a role can we automate without losing competence and responsibility?

When routine disappears, a learning path often disappears with it

Many activities that are especially well suited to automation are, at the same time, entry-level tasks. They are not always technically spectacular, but they fulfil an important function: people use them to learn the normal case.

A junior employee does not develop judgement through training alone. They develop it because they see many cases, recognise patterns, make mistakes, receive feedback and take on more responsibility step by step.

When a system takes over exactly these cases, we gain efficiency in the short term. At the same time, the previous path from junior to senior can disappear.

From an HR perspective, this means: whoever automates learning fields must build new learning architectures. Mentoring, case reviews, simulations, job rotation, joint analysis of AI outputs and deliberately assigned levels of responsibility become more important.

Workforce planning becomes the design of a human-AI system

Workforce planning is changing too. In future we will no longer plan only positions, FTEs and skills. We will plan a work system made up of people, technology and decision rights.

For relevant roles, HR and the business units must jointly clarify:

Which tasks can AI take over completely?

Which tasks may it prepare, but which must be checked by people?

Where is relationship, negotiation or situational judgement needed?

Which tasks are today's learning fields for tomorrow's specialists?

Where must a person be able to stop, correct or decide in a critical situation?

Workforce planning thus becomes more a matter of organisational design. The relevant unit is no longer just the position, but the combination of task, competence and decision right.

Recruiting must look for different signals

As standard work becomes more automated, the criteria by which we select people also change. Pure execution competence remains important, but in some roles becomes less differentiating.

Skills such as contextual understanding, the ability to learn, judgement, dealing with uncertainty and the willingness to question a seemingly perfect system answer become more important.

This does not mean that expertise counts for less. On the contrary: whoever is to assess AI outputs needs enough technical depth to recognise errors and contradictions.

Performance management needs a new logic

When AI prepares part of the work, performance becomes harder to measure. An employee produces a report in ten minutes because AI has pre-structured most of it. Are they more productive? Or are they simply using a more powerful tool?

For HR, it therefore becomes crucial not to measure output alone. We must assess more strongly how people work with AI: do they check results? Do they recognise limitations? Do they escalate when uncertain? Do they use the system as support – or do they delegate the thinking?

HR belongs before the automation decision

For us, a clear consequence follows: HR must not join AI projects only when it comes to communication, training or workforce reduction.

We must help shape things when tasks are automated, roles are redrawn and decision rights are assigned. Because every automation decision is at the same time a decision about competence, learning paths and responsibility.

Our task is not to defend people against technology. Our task is to ensure that the organisation, even after automation, still has the capabilities it needs to decide, to learn and to intervene when it matters.

If HR only joins in at the training or workforce-reduction stage, the most important part of organisational design has already happened.