I'm Directing Five Agents Right Now. In Six Months, That Number Doubles.

Right now, in my daily work, I’m actively directing five to ten AI agents across different workflows. Content orchestration, knowledge management, code development, compliance analysis, research synthesis. Each agent has a defined role, defined boundaries, and specific capabilities.

The cognitive load is real. I’m holding multiple system states in my head, making cross-cutting decisions, managing risk across workflows, evaluating whether the output meets the bar. It’s demanding work — different from managing a human team, but no less intense.

Here’s what I’ve noticed: the bottleneck isn’t the agents. It’s me.

The agents can handle more. They could move faster, take on more complex tasks, operate more autonomously. What constrains the system is my ability to effectively oversee, steer, and validate at the pace they’re capable of. My bandwidth is the limiting factor.

That tells me two things. First, the scaffolding matters enormously — the better my infrastructure for monitoring, validation, and control, the more agents I can effectively direct. Second, this ratio is going to shift. As the tooling improves and as I build better oversight infrastructure, the number of agents I can effectively manage will increase.

Six months from now, I expect to be directing twice as many. Not because I’ll be working harder. Because the systems supporting the human role will be better.

That’s the investment that matters: not just building better agents, but building better infrastructure for the humans directing them.