The agentic-AI coworker

They have no employee number, no onboarding plan, no salary review, yet AI agents are already performing tasks that, only a few months ago, defined what it meant to be an employee: sorting cases, answering policy questions, screening candidates, compiling decision material. The question now is who is steering it – and that makes HR's role in shaping the answer urgent.

A dual mission

It is easy to think that AI in HR is about making the function itself run more smoothly through fewer manual cases, faster recruitment, and less administration. That may be true, but it is only half the picture. While HR is digitalising its own engine room, the rest of the business is waiting for answers to far bigger questions: Which tasks will disappear? Which skills will become critical? How do we reskill employees at the pace the change requires?

Rather than succumbing to the temptation of looking at these two missions as a binary choice, they must be pursued in parallel. An HR function that only fixes its own processes does become more efficient, but no additional value is created for the business. Waiting to get your own house in order before daring to advise the organisation will always leave you one step behind.

The trap most HR functions fall into

The most common misconception is that the solution will come from the outside – that Workday, SAP, or the next provider will eventually deliver the module that solves everything for you. The providers are undeniably building powerful functionality, but a new module leaves fragmented data, unclear ownership, and outdated processes exactly as they were. And it does nothing to supply the ambition you still need to bring yourself.

"HR cannot outsource its imagination to a software provider" is an often-cited line of reasoning in conversations with HR leaders, and for good reason: HR itself must now decide what the employee experience should feel like, which decisions AI should support, and where human judgement must remain – or run the risk that the provider decides it for them.

A clearer conversation

Part of what makes the AI conversation so difficult to manage is that it is being held at the wrong level. "Will AI take the jobs or not?" is rarely answered meaningfully. And maybe it cannot be.

Instead, it is more useful to break down roles into tasks and to look at the expected outcome of those tasks by using four different definitions:

• Fully automated – tasks AI can take on end to end
• Human + AI – tasks that improve when solved together
• Human only – tasks that require empathy and ethical judgement
• Entirely new – tasks that did not exist before, and the new roles and career paths that need to be built around them

Task by task, this distinction can help move the conversation from abstract anxiety to concrete prioritisation and grant a deeper understanding of how roles in the organisation are affected. It also happens to be the kind of work that HR is best placed to lead – if the function dares to take the lead, that is.

The skills gap no one wants to talk about

As we all know, the change is moving fast, and many businesses are already letting AI take over an increasing number of tasks – but where is HR in the conversation? Where does human judgement need to take up more space? How is employee skill development impacted, and what impact do agents have on the working environment? Many decisions around AI development are deeply human. The challenge is that HR is often left off the invitation list and, unfortunately, still lacks the confidence to invite itself.

It is true that HR understands people, culture, and leadership better than any other function – some would argue it is the whole point of HR. But that does not exempt HR from stepping forward and experimenting more with technology. HR needs to be able to ask providers and IT the sharp questions required to own the direction. This is an argument for fostering digital curiosity and accepting that a seat at the table, when the business shapes its AI strategy, now requires getting comfortable with feeling uncomfortable.

What efficiency can buy

Another trap, more treacherous than the first, is celebrating freed-up time as the end goal in itself. A function that only automates its own transactions and reduces its headcount ends up smaller and stays exactly as tactical as before.

The true win only comes when the time freed up is actually invested in what makes HR business-critical: advising leaders through transformation, designing new and changed roles, and taking responsibility for keeping the human side of the workplace central to a technology-driven change.

The position is being claimed now

The organisations that already have agents in their workflows have moved past asking whether AI will change work and now ask who should own the change as it unfolds.

For HR, it is a question of positioning: to be the function that shapes how people and AI work together – or to be the function that manages the consequences of someone else having done it first.