WHEN 'THE BENCH' IS NO LONGER HUMAN

My job, in one form or another, has always been to match capable people to work that needs doing, to see far enough ahead that the right person is free at the right moment, and to notice when someone is being slowly worn down by work nobody else can see. Lately I keep getting asked a version of the same question from a new direction. When some of a firm’s capacity is no longer human, when its resources start to include AI agents that can research, draft, analyse and act, what happens to resource management? Will resource managers end up scheduling agents alongside people? Or will the agents, in the end, manage each other?

I’ll be honest, a lot of the scheduling, matching and chasing that fills resource managers’ weeks could and should be done by software. And I’m genuinely excited about a lot of it; a great deal of what fills a resourcing week is exactly the kind of work I’d happily hand to an agent tomorrow. But the part of this work that was always the most valuable is the part that software cannot do, and I think an agent-heavy firm makes that part more important, not less.

Start with what does change. The most immediate shift is that the bench gets a new kind of occupant. A resourcing plan has always been an answer to three things: what work is coming, what capability it needs, and who we have to do it. Agents add a new class of capability to that last part. They take no holidays, they have no notice period, and you can bring a hundred of them online in an afternoon. For a resource manager that is both liberating and disorienting, because most of our craft assumes capacity is scarce, slow to build and human. When part of it becomes elastic and instant, the arithmetic of capacity planning shifts underneath you.

But only part of it, and the old tensions come straight back in new form. One of the things I say most often is that hiring is not always the answer. In an agent-heavy firm, spinning up more agent capacity will not always be the answer either. The temptation to meet every gap by adding capacity, human or synthetic, is the same temptation it always was: it treats a question of judgement as a question of supply.

Then the next layer of the question. If agents can do the work, can they also do the allocating? In principle, yes. Orchestration already exists, where one agent breaks a task into pieces, hands them to others, monitors progress, retries what fails and assembles the result. That is, recognisably, a manager’s work. So it isn’t fanciful to picture a firm where much of the scheduling and matching that fills a resource manager’s week is handled by software, with agents managing agents and no human in that particular loop.

So what is left for the humans? The part that mattered most all along.

An orchestrator can keep a fleet of agents perfectly busy, every hour, on tasks completed to spec, and it can do all of that while pointing the whole fleet at work that was never worth doing. I have watched human firms make exactly this mistake. A firm can be told its utilisation is too low, drive it up, and destroy value in the process. An agent fleet can make the same mistake faster and more cheaply, and be congratulated for doing so. The judgement about what is actually worth doing, which work genuinely serves the client and the firm, was never a scheduling problem, and it is the one an orchestrator has no way to originate for itself. Someone human still has to own the commercial outcome and answer for it.

And then there are the people. Half of resource management has always faced the employee lifecycle: careers, growth, overstretch, the person who is already halfway out the door. Agents have no careers to nurture, so it is tempting to assume that half of the job simply shrinks. I suspect the opposite. In a firm where agents absorb more of the routine, the humans are left holding the work that is hardest to see and hardest to measure: the judgement, the relationships, the difficult calls. That is precisely the work visibility bias has always overlooked, and precisely the kind of overstretch that never shows up on a report. The human half of the job becomes more important in an agent-heavy firm, not less, because the people who remain are carrying the part of the work that matters most and shows up least.

So will resource managers manage AI agents? Probably, in part, and sooner than many expect. Will agents manage agents? In the mechanical sense, very likely. But neither of those answers the question sitting underneath, which is who decides what the work is for, who judges whether it was any good, and who looks after the people whose working lives are being rearranged around it. Those were never really scheduling questions. They are questions of judgement and stewardship, and they are the reason resource management was always a human discipline.

The tools will keep changing, as they always have. The real work, I suspect, will stay exactly where it has always been. Not in keeping the capacity busy, but in deciding what it is for.

have new articles delivered

I write here every so often about resource management, utilisation, visibility and the commercial value of people in professional services. If you'd like new pieces sent to you as they're published, leave your details on the form.

 

 

 

 

I hereby agree that this data will be stored and processed for the purpose of establishing contact. I am aware that I can revoke my consent at any time. I'll only use your address to send you new articles. No sharing, no noise, and you can unsubscribe from any email in one click.*

* Indicates required fields
Thank you. Please check your inbox and confirm your email, and I'll make sure the next piece reaches you.
Information icon

We need your consent to load the translations

We use a third-party service to translate the website content that may collect data about your activity. Please review the details in the privacy policy and accept the service to view the translations.