Leadership in the AI Age - Part 2

Author: Paul Harman, Partner nem Australasia

Last updated: 22 October 2026

A note from Paul: This series reflects my own views, experience and judgement on leadership in the AI age.

Part 2.  Leading People Through Uncertainty

In Part 1 we explored judgement and how to take responsibility for decisions based on AI generated work. In Part 2 we look at how AI changes the landscape for an employee and how a leader is required to create a platform of communication to ensure they get the benefit of the new opportunities that exist with AI. 

On Monday morning, a manager announces that the team will begin using AI to improve productivity. The message is upbeat. There is talk of saving time, reducing administration and creating space for more valuable work.

Around the table, people hear different things.

One sees an opportunity to escape repetitive tasks. Another wonders whether years of hard-earned expertise are becoming less valuable. Someone else quietly calculates how much of their role consists of the work now being described as suitable for automation.

Leading this change starts with recognising those different interpretations.

A leader does not need to predict every consequence of AI. They do need to create conditions in which people can ask reasonable questions and receive honest answers. Reassurance becomes fragile when it promises certainty the organisation does not have.

A more credible starting point is specific: here is the work we are exploring, here is why, here is what has been decided, and here is what remains open. If changes to roles are possible, acknowledge that possibility and explain how people will be involved in the process.

Trust also depends on what happens after the announcement. A team told to experiment may reasonably hesitate if every mistake attracts criticism. Equally, an invitation to “try AI” offers little help if people have no approved tools, no time to learn and no clarity about acceptable use.

Leaders need to make learning practical. Begin with a bounded task, such as drafting a routine internal update using non-sensitive material. Give people an opportunity to compare the result with their own work. Discuss what improved, what was lost and how much checking was required.

Those conversations build shared understanding. They also make room for different kinds of expertise. A confident user may know how to get a useful first draft. An experienced colleague may recognise an omission that makes the draft unsuitable. Both contributions matter.

Leaders should participate in this learning themselves. Showing an example where AI misunderstood a request can open a more useful discussion than presenting only polished successes. It signals that reviewing and correcting outputs is part of competent use.

There is also a deeper question about development. If AI takes on some first drafts and routine analysis, how will less experienced colleagues learn the underlying work? Leaders should deliberately preserve opportunities to practise, explain reasoning and receive feedback.

People are more likely to contribute when they can influence how their work changes. Ask them where effort is wasted, which tasks require personal attention and what they would do with time saved. Their answers can reveal opportunities that a management-led efficiency exercise might miss.

The measure of leadership is how well the team can learn, question and adapt together. Enthusiasm is useful, but it should leave room for informed scepticism and different starting points.

On Monday morning, take three practical steps:

• Hold a candid conversation. Ask the team what excites them and what concerns them about AI. Explain what you know, acknowledge what you do not, and commit to a date for revisiting unanswered questions.

• Create one supported experiment. Choose a low-risk task, confirm the approved tool and information boundaries, and allocate time to test and review it together.

• Agree where saved time should go. Identify a worthwhile priority, such as customer follow-up, coaching or deeper analysis. Revisit whether the experiment actually creates capacity for it.