A leadership team spent an afternoon mapping which of their roles would disappear under AI. By the end, they had a list of tasks that could be automated, a nervous silence, and no clearer sense of what the actual work would look like in two years.

This is the mistake that repeats across organisations. We keep asking which jobs AI will take, as if the answer were a neat list. The more useful question is narrower and less dramatic: which tasks within a role change, which ones stay, and what does that mean for the person doing them?

Most roles will not vanish. They will be re-formed around a different mix of activity. The parts that involve pulling together routine information, drafting standard responses, or processing predictable requests will increasingly be done by software. The parts that involve judgement, context, relationships, and accountability will remain, often becoming a larger share of the role. The work does not disappear; it shifts, and the person in the role has to shift with it.

Take a caseworker in a public service. AI can draft letters, summarise case notes, and flag inconsistencies in a file. It cannot sit across from someone who has just been given difficult news and decide how much to say, or when to stop talking and just listen. That is not a task that can be automated; it is a human skill that becomes more central as the administrative load shrinks. The same pattern appears across sectors. A finance analyst who used to spend days reconciling spreadsheets can spend those hours explaining what the numbers mean to a non-financial leader. A procurement manager who used to chase paperwork can invest more time in supplier relationships and the trade-offs those conversations carry.

None of this happens automatically. It requires a deliberate effort to identify which tasks are genuinely automatable, which are not, and what new skills the people doing the work will need. That effort is not a technical exercise; it is an organisational one. It means looking at a role as a bundle of tasks, some of which have a certain future, some of which do not. It means being honest that a role may shrink before it grows, and that the interim period, when the automation is in place but the new ways of working are not, is when the pressure on people is highest.

It also means recognising that the skills that become more important are harder to train for. We are good at teaching people to use a new system. We are less good at helping someone develop the confidence to have a harder conversation, or the judgement to know when to override a recommendation, or the patience to build trust with a colleague who is resistant. These are not bolt-on skills. They develop through practice, coaching, and time, and they demand as much attention as any software rollout.

For leaders, this reframes the conversation. Instead of asking which roles will disappear, ask which tasks in each role are most ready to change, which would survive any technology, and what that means for the skills you need to build now. That is a conversation that engages every function: HR in redesigning roles and career paths, digital leads in choosing what to automate, finance in understanding the value of the shift, and the board in overseeing a transition that will not end in a single quarter.

This is not a comfortable message. It does not offer the clean narrative of jobs saved or jobs lost. But it is closer to what most organisations will actually experience: a gradual, uneven, often messy adjustment in which roles evolve, year on year, and the work that remains is more human, not less. Those who plan for that will have more influence over the direction of change. Those who wait for the list of doomed jobs may find the change has already happened around them.