There is a reassuring way to talk about AI in organisations. We can say that it is a skills issue. We can launch training programmes, publish guidance, invite people to experiment responsibly and build a dashboard. No transformation is complete, after all, until someone has built a dashboard.
None of this is wrong. Much of it is necessary. It is just not enough. AI adoption is often presented as if the main challenge were to teach people how to use new tools. This is a convenient frame. It gives HR, L&D and business leaders something clear to do. Training can be planned. Attendance can be tracked. Satisfaction can be measured. The organisation can show that it is moving. Everyone can breathe for a moment and agree that the future has been addressed.
The problem is that AI does not only change skills. It changes work.
It changes tasks, workflows, role boundaries, decision rights, performance expectations and the relationship between human judgement and automated support. It changes what junior professionals learn by doing, what managers need to supervise, what experts are expected to validate and what organisations see as productive work.
Treating all this as a generic upskilling issue is rather like renovating a house by buying new furniture while trying not to discuss the walls, the wiring or the fact that the ceiling looks slightly concerned.
For HR leaders, this creates a strategic shift. HR cannot remain only a service provider for recruitment, training, engagement, policy and employee experience. These activities still matter. But they are no longer enough. In AI-enabled organisations, HR increasingly needs to act as a steward of work design.
That expression may sound a little grand. It is meant to. The alternative is less attractive: work will be redesigned anyway, but informally, unevenly and often by overloaded managers trying to make sense of new tools while keeping daily operations alive. This is not necessarily careless. It is simply what happens when technology moves faster than organisational design.
The first implication is that HR should move upstream. Before launching broad AI training programmes, organisations need to understand how work is actually changing. This requires more than asking which tools people use. It requires a practical form of work analysis.
Which tasks are likely to be automated? Which tasks can be augmented? Which tasks still require human judgement, professional responsibility, ethical evaluation, relational sensitivity or contextual understanding? Which decisions can be supported by AI, and which decisions should remain clearly accountable to a person? Which parts of a role become more complex because AI can produce more information at higher speed?
These are operational questions. They affect job descriptions, learning priorities, delegation, supervision, performance management and risk. They also affect identity. People do not experience work only as a list of tasks. They experience it as competence, recognition, autonomy, contribution and professional meaning. When AI changes the task structure of a role, it may also change how people understand the value of their work.
This is why HR should avoid two tempting simplifications.
The first is technological optimism without work design. In this version, AI will free people from routine tasks, raise productivity and allow everyone to focus on higher-value work. This may happen, in some roles and under some conditions. But higher-value work does not appear by magic once lower-value work has been automated. It has to be defined, distributed, supported and recognised. The fairy godmother of productivity is, sadly, not part of most operating models.
The second simplification is defensive caution without redesign. In this version, HR focuses mainly on risks, rules, acceptable use and compliance. These are essential. No serious organisation can ignore confidentiality, bias, data protection, intellectual property and responsible use. But if governance remains only a list of restrictions, people will still improvise. They will simply do so more quietly, which is not usually the gold standard of governance.
A more mature HR approach connects governance, learning and work design.
This begins with task analysis. HR and business leaders should identify where AI is already entering work, even before official policies have fully caught up. In many organisations, employees are already experimenting with drafting, summarising, analysing, translating, coding, preparing presentations, benchmarking and generating ideas. The question is not whether AI is present. It usually is. The question is whether the organisation understands what this presence is doing to roles and responsibilities.
The second step is role redesign. Once tasks are mapped, HR can help managers distinguish between automation, augmentation and human responsibility. Some activities may be safely automated. Others may be improved by AI but still require human review. Others should remain centred on judgement, trust, empathy, negotiation, creativity or ethical accountability.
This distinction matters because it prevents AI adoption from becoming a vague invitation to “use the tool when useful”. That sentence sounds reasonable. It is also a fine way to leave people alone with uncertainty. A role-based approach is more precise. It clarifies where AI can support a role, where it changes expectations, where it creates risks and where it should not replace professional responsibility.
The third step is learning design. Training should follow role redesign, not sit beside it as a separate activity. Generic AI awareness has its place, especially at the beginning. But real transfer requires learning objectives connected to work situations. A manager does not need the same AI learning path as a recruiter, a consultant, a finance specialist or a customer service professional. Even within the same function, seniority changes the learning need.
Junior professionals may need to learn how to use AI without bypassing the slow acquisition of expertise. Managers may need to learn how to evaluate AI-supported work without becoming either naïve users or suspicious gatekeepers. Senior experts may need to define quality standards, review practices and accountability boundaries. HR teams may need to rethink competency models, performance criteria and development paths.
The fourth step is managerial accountability. This is often the most delicate point. Organisations sometimes speak about AI transformation as if managers simply need to encourage experimentation. Encouragement helps, but it is not a management system. Managers need support in discussing how work changes, how expectations shift, how performance is evaluated and how people can raise concerns without being labelled as resistant, old-fashioned or, the modern corporate insult, “not future-ready”.
If HR leaves this entirely to line managers, outcomes will vary widely. Some managers will redesign work with care. Others will delegate the issue to the most digitally confident person in the team. Some will overuse AI because it looks efficient. Others will underuse it because they fear mistakes. The result is not innovation. It is organisational patchwork, with a modern label.
The fifth step is performance and wellbeing. AI adoption should not be evaluated only through usage metrics. The fact that people use AI tools does not prove that work has improved. It may only prove that people have understood the hint. Better questions are needed. Has the quality of work improved? Are decisions faster and still accountable? Are employees learning, or are they simply producing more? Are managers able to supervise AI-supported work? Are risks visible? Has workload decreased, or has productivity simply become a polite name for more demand?
This last point matters. AI may increase efficiency, but efficiency gains do not automatically become wellbeing gains. Without deliberate choices, they may become shorter deadlines, higher expectations and more invisible pressure. HR should therefore connect AI adoption to sustainable work design, not only to productivity.
This is where the strategic role of HR becomes clear. HR leaders are not expected to become technologists. That would be unnecessary, and in some cases possibly alarming. They are expected to understand work, people and organisations. That is exactly the territory AI is now reshaping. The opportunity is not to own the technology agenda. It is to make sure that the technology agenda does not bypass the human and organisational architecture of work.
In practical terms, HR Directors can start with a limited number of questions for each major role family:
- What parts of this work are already being changed by AI?
- Which activities should be automated, augmented or protected?
- What new judgement, review or supervision practices are required?
- What do managers need to discuss with their teams?
- What learning is role-specific rather than generic?
- What performance indicators need to change?
- What risks may remain invisible if we only measure adoption?
These questions move the conversation from “Are our people ready for AI?” to “Is our organisation redesigning work responsibly?” That is a better question. It is also harder to answer, which is usually a sign that it is worth asking.
For years, HR has argued that people strategy should be closer to business strategy. AI offers a practical test of that claim. If HR responds only with training catalogues and acceptable-use policies, it will remain useful but peripheral. If HR helps the organisation analyse, redesign and govern work, it becomes strategically central.
The future of HR in AI-enabled organisations will not be defined by who writes the best policy on generative AI. Policies matter, but they are not enough. It will be defined by who helps organisations decide how work should be structured, how responsibility should be distributed, how people should learn, and how performance should be evaluated when human and artificial capabilities increasingly interact.
AI adoption is a work design challenge, not just a skills gap. HR should treat it accordingly.
