AI, Managers and The Accountability Gap 

The narrative around AI and managers quickly goes straight to headcount and how AI could be responsible for layoffs and organisational restructuring. Can we get by with fewer layers, wider spans of control, less middle management? Well, plainly, plenty of organisations are over-layered, and a real share of management work is coordination overhead that AI absorbs well, like aggregating status, routing decisions, scheduling, first-pass synthesis of what’s happening below.

As with most questions, however, the real answers are uncovered upon deeper examination. In real organisations, management is one of the places concentrated judgement accumulates. If you use AI as a convenient excuse to strip the layers in an organisation without distinguishing overhead from judgement, you save a penny today but spend dollars later. Such moves look, sound, and feel like efficiency, but the real cost is paid later.

We should be asking what management looks like when the rote duties of coordination, reporting, and status rollups are gone. I think it looks like duties are getting narrower and heavier, with fewer people spending less time on synthesis and more on judgement-dense work that doesn’t delegate to a model. Most organisations don’t have this yet. They instead have the cost-cutting instinct without the redesign, which is how you get the layoff without the rethink, and the organisation without the muscle of context and judgement. 

AI Can’t Have Accountability

It’s tempting to characterise the defining value of a great manager as something AI is currently bad at. That’s a shaky place to plant your flag because that frontier keeps moving. Reading context and generating decent coaching language are the things models got good at fast. Pin a manager’s worth to a capability gap and you’re betting on the gap staying open. 

At the end of the day, though, someone still has to be answerable for a call or a bet. AI can generate feedback signals, but it can’t be held responsible for them. A great manager takes that noisy signal, decides what it means, but most importantly, owns the decision that follows. 

Part of that job is keeping the signal honest - making sure a dashboard doesn’t quietly become the truth, or that a feedback number doesn’t flatten into a one-dimensional score that someone’s career rides on. The harder part is making a call people can understand and trust when there’s no clean metric to point at and standing behind it when it’s wrong. 

If the job degrades into “confirm what the dashboard says,” you’re not really a manager. You are just a reporting machine. A human PC monitor.

What higher leverage management looks like in practice 

Consider a day in an organisation that does AI right. At the start of the day, an agent has already summarised what moved overnight. Management doesn’t have to spend the first hour reconstructing status from six places. They instead get to ask “is this read of the situation correct, and what's it missing.” That’s the shift, with almost no time gathering, and almost all of the time reaching conclusions and making judgements.

One-to-ones between a boss and a subordinate change the most. The discussion becomes about what the models think is going on with work, and the back and forth is about the next actions and course correcting, if necessary. Sometimes the most useful thing managers do all week is tell someone the dashboard is wrong about them, because the signal missed the context. That conversation might not have existed before, but now it’s central. 

Managers often do work and provide significant value to an organisation without it appearing on a calendar. An agent could draft the performance summary, but managers are the one who decides whether it’s fair, and who owns it when they deliver it. AI can structure a development plan, but it can’t sponsor someone in the room they’re not in or notice that a quiet person is about to leave.

AI gave managers back the hours they used to spend finding out what was happening. The whole job becomes what you do with those hours. Done poorly, that time evaporates into more meetings but done well, it goes into better decisions and better people, both things that pay dividends that compound over time. 

The biggest risk is performance management without an owner 

Where organisations are most at risk is introducing AI into performance management and people operations. Organisations cannot automate the measurement of performance without automating the accountability for it. The result is a system that confidently rates people, and no human actually owns those ratings. 

Performance is contextual and multidimensional, but a model can only act on what it can see. The system quietly starts rewarding the legible work and discounting the invisible work like the person who unblocks everyone else, mentors the new hire, or calms the angry client. None of that shows up in the data, so the system reads them as low performers. It might even go unnoticed until the best employees are the ones leaving. 

The deeper trap is that the standard fix makes it worse. Everyone says “human-in-the-loop,” but in practice the human becomes a rubber stamp, approving what the system surfaces because to override it is work, friction, and the system looks authoritative. Doesn’t it know better than I do?

The organisations getting AI right are the ones where an employee can see what fed a judgement, challenge it, and reach a human who genuinely owns the outcome rather than ratifying the machine’s. 

The next three to five years will sort organisations by trust 

There is a new requirement for load-bearing skills which is about interrogating the output. You don’t need to be an ML engineer to ask questions of a model. You aren’t smarter than the computer.  Everyone needs the habit of asking models questions and challenging their responses even when the output is fluent, confident and easy to accept. 

AI removes the excuse to keep mentorship as a primary element of the manager-subordinate relationship. The moment an agent can triage reports, summarise work, and draft feedback, it looks fantastic and efficient on paper to give a manager 30 direct reports instead of six. But at 30 directs, mentorship is impossible; the connection gets crowded out by a span of control that AI made look affordable. 

The ones that succeed ask what management becomes when the admin layer is gone. They’ll choose what AI is allowed to decide versus only recommend, who is accountable when it’s wrong, how someone challenges a decision that lands on them, and whether overriding the system is genuinely easy or just theoretically allowed. 

Trust is downstream of accountability. The organisations that keep a human visibly answerable for the decisions that shape people’s lives will earn it. The ones that hide behind the system will spend three years looking efficient and the fourth wondering where everyone went.