Leadership and Management in the AI Era

Risk and uncertainty are different. With risk, you know the possible outcomes and can estimate their likelihood. With uncertainty, even that is impossible. AI brings both: errors in an application can be measured, but nobody knows how work will change over the next few years. Leaders who handle uncertainty well choose a direction for the coming months, state what they do not know and agree when to review it. They distinguish between what is fixed, such as values and the promise to the customer, and what can shift, such as methods and systems. A trial can therefore stop without changing the overall direction: new insights change the approach without immediately changing the vision of the future.

A vision of the future describes where the organisation wants to be in a few years and why: for which customers, in which role, with what work. An AI vision that only covers systems misses the heart of it: what does the organisation want to do differently or better with AI? Scenarios help to test such a vision. A scenario is not a prediction but a coherent picture of a possible future: one in which AI is available everywhere at low cost, for instance, alongside one with strict rules and little trust. Hold the vision up against three or four scenarios and you see which choices stand in every scenario explored and which depend on one particular future. That is how the leadership decides which steps are already sound today.

Taking the initiative means starting before all the answers are in. In many organisations middle management waits for a strategy from above, while the top waits for ideas from below. A leader who starts breaks that deadlock. At the outset, the leader uses AI for a few weeks on a fixed part of their own work, such as a monthly report or preparing a performance review. That experience shows where AI helps and where it does not. A further trial within the team shows what an application contributes to the work they do together. The conversation about AI then has an example from inside the organisation instead of a supplier's presentation. Even a trial that delivers little shows where expectations and practice part ways.

An AI-first organisation starts with a leader who makes choices about work. AI and new technology do the work people need not do, including the repetitive and predictive work: compiling reports, updating schedules, forecasting next quarter's demand. People focus on what they do better: exercising judgement, making choices and working with customers and colleagues. A team that leaves the gathering of figures to AI has more time for the customer who has doubts and the colleague who is stuck. Whether that time really materialises depends on how the leader divides the work anew.

More teams work alongside AI agents: AI systems that take steps independently within an assignment, such as compiling a case file or updating a schedule. Leaders reallocate the work, identifying what an AI agent can do and where a person is needed. They set its authority and decide who checks its output. Performance assessment changes because some of the work no longer rests with one person alone. New roles emerge, including people who configure and monitor AI agents. Alongside directing agents, the human side of leadership remains: coaching, exercising judgement and having the difficult conversations. The Agentic AI keynote examines authority and oversight in more depth.

Psychological safety means a team believes people can ask questions, express doubt or report mistakes without punishment or humiliation. This matters when AI changes work, because it affects people in ways no project plan covers. An experienced employee wonders what their expertise is still worth; a colleague sees extra work. When people are afraid to voice doubts, they try little that is new. In customer service, a leader can let the team choose which enquiries AI drafts responses to and which staff answer themselves, with time set aside to practise. The leader also highlights what remains the same: professional judgement, contact with customers and knowledge of the work.

As AI takes on more tasks, the definition of a good employee changes. Expertise still matters, alongside the ability to give clear instructions, assess results critically and redesign one's own work. Talent is increasingly judged by what someone can do, with less emphasis on job titles or qualifications. That affects every career conversation a leader has. Learning goals shift with the work. Recruitment places more weight on how candidates use AI. Retaining good people means discussing how their work is changing and the development that fits it. The start of a career changes too, as AI takes over some entry-level work.

An executive team or board that decides on AI must know enough itself to ask the right questions. Learning therefore starts at the top: fixed moments to try out new applications, conversations with specialists and an honest view of what their own knowledge is still worth. On top of that, the leadership takes charge of learning across the whole organisation. Education often lags behind what is changing in the work, and a one-off training course dates quickly. More and more organisations are therefore building their own future academy, where colleagues from their own teams teach and the programmes fit their own processes. That changes the leader's position too: they decide on a technology whose capabilities shift again while they are still learning.

In many organisations employees use AI on their own initiative, often with a personal account. That signals two things: there is demand, and there are risks, from company data leaving the building to answers nobody checks. A ban only pushes that use further out of sight. It works better to stay curious about what people have already discovered and make clear which data may and may not be used; safe tools follow from there. Those employees are often the best guides: they know where the work gets stuck and what AI is already changing there. Leaders who ask about it and share the good examples have something to build on.

Innovation in an established organisation means connecting something new to systems that remain in use for years, rules that once served a purpose and habits nobody questions any more. AI makes this tangible: a new application has to work with old data and procedures. Leaders making progress in that setting do not start by replacing everything. They choose a process with a clear owner, build a working solution alongside the old system and demonstrate the result. Retiring the old method then becomes a smaller step. They also remove things deliberately: an approval stage that adds nothing, a report nobody reads. Making room for something new often starts with removing something old.

It is tempting to put IT, an innovation team or someone in a new AI manager role in charge of AI. But the work that changes happens in operational departments: the department that prepares quotations, the team that helps customers and the planners in the warehouse. Specialists build; leaders in operational departments apply it and own the result. They decide how the work changes, who checks what and when an application is good enough. An innovation team that builds a pilot without an owner in an operational department produces a demonstration, not a new way of working. When operational departments take ownership, a pilot can become part of everyday work, and the knowledge is more likely to stay in the organisation when the project team leaves.

AI makes a lot of work cheaper and faster. But cheaper is not always better. The payments company Klarna let AI take over a large part of its customer service. Its chief executive later admitted that the focus on cost had led to lower quality; the company is investing in human staff again, so customers can always speak to a person if they want to. The lesson is not that AI fails in customer contact, but that leaders must choose deliberately where people make the difference. That is often in moments of doubt, emotion, complaints and big decisions. Organisations that get this right identify what the customer really values in each interaction and automate the rest.

No organisation can hold all the knowledge about AI in-house. The technology is being developed in many places at once: by start-ups, researchers, partners, governments and organisations in other sectors working on the same question. A leader's ecosystem therefore helps determine how quickly an organisation learns and renews itself, and that ecosystem does not stop at the national border. It does not build itself. Leaders who build one visit organisations that are further ahead, look for partners for a joint trial and talk openly with peers about what failed. A healthcare organisation learns from scheduling in a port, a local authority from customer service in retail. Who a leader spends time with is a strategic choice, not a minor detail in the diary.