Leading leaders who lead agents

There is a role quietly appearing inside large organisations that has no name, no job description and no development pathway. It is the person who used to manage a team of eight, and who now manages three people who each direct a fleet of AI agents. Their team's output has gone up. Their headcount has gone down or held flat. And their old management playbook — built on assigning work, checking work, developing people through the work — has stopped describing their week.

Now go one level up. Somebody leads that person. What does that job look like?

This is the question I find most senior executives have not asked yet, and it is the one I think decides organisational performance over the next five years. Not "how do we adopt AI" — that question is everywhere. The harder question is: what does it mean to lead leaders whose teams are part-human, part-agent?

The middle layer is where the change lands

Start with what is actually happening one level down.

When capable agents enter a team, the first-line manager's job transforms in specific, observable ways. Task allocation stops being about matching work to people and becomes about designing the split between people and systems. Quality control stops being "review the work" and becomes "design the verification regime" — because the volume of output now exceeds what any manager can personally read. Development conversations change, because the junior tasks people used to learn on are precisely the tasks the agents absorbed. And accountability sharpens uncomfortably: when an agent-produced error ships, whose error was it?

These managers are, mostly, figuring it out alone. They were promoted for being good at a job that no longer exists in the same form, and the organisation above them is still measuring them with instruments built for the old job — utilisation, headcount, activity. The gap between what their role has become and what their organisation thinks their role is generates a very particular kind of quiet stress. I hear it in coaching conversations every week.

What the layer above must now do

If that is the first-line reality, then leading that layer — being the leader of leaders — is a genuinely different discipline. Four shifts stand out.

From reviewing outcomes to reviewing systems. You can no longer evaluate your managers by sampling their team's work, because the work is increasingly machine-produced and volume has exploded. What you must evaluate instead is the system each manager has built: how they decide what gets delegated to agents, what their verification looks like, where human judgement is mandatory, how failures surface. A manager with a sound system and a bad week is fine. A manager with good numbers and no articulable system is a risk you have not priced. The skill this demands — auditing the design of work rather than the output of work — is closer to how good engineering leaders review architecture than how business leaders traditionally review performance.

From capacity maths to judgement maths. For a century, organisational capacity has been arithmetic: people times hours times skill. Leaders of leaders did resourcing. When agents supply effectively unlimited execution in some domains, the scarce input stops being hours and becomes judgement — the ability to decide what is worth doing, to specify it well, and to recognise quality. Your resourcing question changes from "do we have enough people" to "do we have enough judgement, and is it positioned where the decisions are". Those are different maps. A team can be fully staffed and judgement-bankrupt.

From managing performance to managing accountability. Agents create an accountability vacuum by default. Work happens, output ships, and when something goes wrong the explanation dissolves into "the system produced it". The leader-of-leaders' job is to refuse the vacuum: every consequential output has a human owner, every delegation to a machine has a named person who stands behind it, and everyone knows this. This is not a compliance nicety. Organisations run on the felt weight of ownership, and that weight does not transfer to software. If you do not deliberately re-anchor it, it evaporates.

From developing skills to developing people through a skills discontinuity. Your managers built careers on expertise that is being repriced in real time. Some of what made them valuable is now abundant. Leading them through that is not a training problem — it is closer to pastoral work: helping capable, proud professionals renegotiate where their value sits without pretending the change is smaller than it is. The leaders who can hold that conversation honestly will keep their best people. The ones who hide behind reassurance will watch them leave — or worse, stay and calcify.

The failure modes I am already seeing

Three patterns recur in organisations that get this wrong.

The first is inversion blindness: executive teams planning as if the current ratio of humans to agents is roughly permanent, when every trend line says the ratio inverts — more agent-hours than human-hours on knowledge work — within a few years in exposed functions. Structures, spans of control, career ladders and leadership development are all being quietly invalidated, and the planning horizon does not acknowledge it.

The second is span-of-control nostalgia. If one manager now delivers what a department used to, the reflex is to flatten and stretch: fewer managers, wider spans. Sometimes right. But it misreads what the remaining managers are doing. Their spans have not narrowed in effort — they have deepened in consequence. Each report now runs a human-machine system whose failure modes are faster and less visible than a human team's. Loading more of those systems onto one leader because "the agents do the work" is how you build an organisation that fails suddenly instead of gradually.

The third is delegating the transition itself to the middle. Executives announce an AI-forward strategy, procure the tools, and leave the middle layer to work out what it means for how work runs. This is the transformation-program mistake of the last two decades replayed at higher speed. The middle layer cannot redesign the organisation's operating model from the middle. That design work — what gets delegated to machines as a matter of policy, where accountability sits, what managers are now for — is executive work, and it is being widely left undone.

Where to start

If you lead leaders, three moves are available this quarter, none requiring a program office.

Sit with one of your managers and have them walk you through, concretely, how their team's work runs now — what the agents do, what the people do, where they personally intervene. Not a status update: an operating review of the human-machine system they have built. You will learn more in ninety minutes than from any dashboard, and they will learn that the system design is the job now.

Re-anchor accountability explicitly. Pick the outputs that matter most from that team and confirm, in writing if needed, which human stands behind each. If the answer is unclear to you, it is unclear to everyone.

And start the honest conversation about what your managers are for. Not with slogans — with the real question: as execution gets cheap, what is the judgement only you can supply, and how do we build the next generation of it now that the old apprenticeship tasks are gone? That last clause is the sleeper issue of the decade. The junior work agents absorbed was also the training ground where senior judgement used to grow. Every organisation running on agents is quietly living off its existing stock of judgement. The ones that thrive will be the ones that worked out how to grow more.

This is the territory our leadership academies work in. If your organisation is meeting it now, talk to us.