Hiring the leader who can redesign work around AI

Many AI initiatives begin at the level of the individual: a professional drafts faster, a manager summarises information or an analyst reviews a larger body of material. These gains may be useful. They do not automatically change service quality, cost or customer value at organisational scale. That change requires choices about the flow of work, the point at which human review is needed, the capabilities people must develop and who owns errors.

The OECD's analysis of AI and skills describes AI's potential to automate tasks, create new ones and improve productivity, with different effects across occupations. For an executive appointment, the implication is practical: the leader must be able to reason about tasks and systems rather than assume that a job title will remain unchanged or disappear wholesale.

Start with the work itself

A strong candidate can take a workflow apart. Which steps require judgement, empathy, negotiation or accountability? Which are repetitive but still sensitive to context? Where do delays arise because information is inaccessible or decisions are unclear? Which tasks could AI assist, and what new verification work would that create? The answers should lead to a redesigned process, not a generic adoption target.

Consider a hypothetical professional-services team preparing a client recommendation. AI may accelerate retrieval and drafting. Yet the senior professional still needs to verify sources, interpret the client's situation, challenge weak assumptions and own the advice. If the organisation measures only drafting speed, it may miss a decline in quality or an increase in review burden. The leader's job is to redesign the whole service, including the human work that remains essential.

Assess change leadership, not only tool sponsorship

The search should look for experience in changing roles, incentives and collaboration across functions. Has the candidate worked with employees who feared loss of expertise or status? Did they involve users in design, train managers to review new forms of output and create a way to report failures? Did they measure whether a new process improved the customer or employee experience, not only whether the tool was used?

It is important to examine how the candidate handled work that became more valuable. If routine preparation is reduced, senior time may shift toward complex judgement, client dialogue or coaching. Those benefits do not arise automatically. The executive must decide how capacity is redeployed and how quality will be maintained.

Make the mandate executable

An organisation may ask a new leader to deliver AI-enabled productivity while leaving process ownership, technology budget and talent development in separate silos. That is a mandate problem. Before searching, the client should define which workflows are in scope, which leaders must cooperate, what investment is available and how outcomes will be judged.

The appointment brief should then ask for evidence of system-level change: a baseline, a redesigned workflow, revised decision rights, a credible learning process and measured effects. The strongest candidate will not necessarily promise the fastest adoption. They will show how to make the work better while keeping the people responsible for it capable and accountable.

An effective first-year mandate would choose a small number of workflows with different risk and value profiles. It would compare intended benefits with actual quality, cost, employee experience and customer effects, and use the findings to decide whether to scale, redesign or stop. Such an agenda gives the new leader room to learn while making progress observable. It also prevents the appointment from being judged solely by the number of tools introduced.

The leader should also know how to prevent a two-tier workforce in which a few experts design AI-enabled work while everyone else is expected to adapt without explanation. In assessment, ask how the candidate would involve the people who know the process, how they would communicate changes in role expectations and how they would develop managers who must review work produced with AI assistance. These questions reveal whether the candidate sees people as a constraint to overcome or as a source of operational knowledge.

A useful counterfactual is to ask what improvement could be achieved without AI. If a workflow is slow because authority is unclear, a model may accelerate the wrong step. A leader who can compare redesign, conventional automation and AI on the same business terms is more credible than one who assumes the technology must be the centre of every solution.