We Will All Become Managers of Agents

There's a quote I can't stop thinking about.

"We used to design how people interact with software. Now we are designing how much they need to."

That line comes from Luke Wroblewski's recent piece Finding the Role of Humans in AI Products, and it captures something profound about where we're headed. Not just in software development, but in every profession.

The Six Stages of Letting Go

Wroblewski traces the evolution of AI through coding tools, and the pattern is striking:

1. Humans wrote code. Computers just executed it.
2. Machines started suggesting. Copilot-style tools offered completions, but humans stayed in control.
3. Chat changed the dynamic. We started talking to AI, brainstorming, problem-solving together.
4. Agents picked up tools. AI began searching files, calling APIs, writing and correcting its own code.
5. Orchestration emerged. Coordinator agents started building plans and delegating to specialized sub-agents.
6. Machines running machines. Meta-orchestration, where humans are barely in the loop at all.

Each step moved humans further from execution and closer to oversight. And here's the thing: software developers were just the first profession to walk this path. They won't be the last.

The Management Thesis

Here's what I believe: eventually, everyone will manage teams of AI agents. Not just developers, not just tech workers, everyone.

Think about what happened with computers. In the 1980s, "using a computer" was a specialized skill. By the 2000s, it was a baseline expectation for almost every job. AI agents are following the same trajectory, just faster.

Wroblewski's data supports this. Software developers make up only 3% of U.S. workers but accounted for nearly 40% of Claude conversations in 2025 and roughly 50% of AI agent deployments by 2026. They are the early adopters, the canary in the coal mine. What's happening to developers now will happen to accountants, marketers, project managers, designers, and lawyers within the next few years.

The pattern is always the same: first the tool assists you, then it collaborates with you, then it works for you while you supervise.

The Skills That Matter

This is where it gets interesting. The skills you'll need to thrive in this world aren't programming skills. They're management skills.

As Wroblewski puts it, the durable competencies are "steering, delegation, and awareness: knowing what to ask for, how much autonomy to grant, and when to look under the hood."

Sound familiar? That's exactly what good managers do with human teams. You don't need to know how to do every task your team performs. You need to know what outcomes you want, how to break complex goals into delegatable pieces, when to trust your team's judgment, and when to intervene.

The difference is that your "team" will increasingly include agents that don't sleep, don't take holidays, and can scale instantly.

What This Means for You

If you're reading this and thinking "but I'm not a developer," that's exactly the point. You don't need to be. The barrier to working with AI agents is dropping fast. Describing what you need in plain language is becoming enough.

The real question isn't whether you'll manage agents. It's whether you'll be ready when the moment comes. And that moment is closer than most people think.

Start building the muscle now. Practice delegating to AI tools. Get comfortable with reviewing output rather than creating it from scratch. Learn to specify what "good" looks like without dictating every step.

Because the future doesn't belong to the people who can do the most work. It belongs to the people who can direct the most work, effectively, responsibly, and with the awareness to know when something needs a human touch.

We used to manage people. Soon we'll manage agents. And the ones who do it well will shape what comes next.