In the future, there are no ICs

Amazon cut 13% of its managers this year. Google cut manager and VP roles by 10%. Meta is replacing product managers with AI builders. Across the tech sector, 59,000 jobs have been eliminated in the first three months of 2026 alone, and the loud narrative in every boardroom is the same: we don't need as many managers.

Let’s be clear on the first irony: the vast majority of those people were not hired as managers in the first place. They were top contributors who were promoted to lead as a result. Now they’re being fired because of their title.

I think that's exactly wrong.

I think the future has no individual contributors.

Here’s why. Jensen Huang just described Nvidia's future as 100 AI agents for every human employee. Not 100 tools, but 100 agents: each with judgment, each producing work that needs to be evaluated, redirected, and held accountable. That's not using software. That's management. The moment you are overseeing an AI agent's output, questioning its assumptions, and providing context it doesn't have, you are doing the job of a manager. Maybe not of managing a human, but requiring many of the same skills.

And it's not going to be optional. HBR published a piece in February called "To Thrive in the AI Era, Companies Need Agent Managers." The punchline: defining tasks, reviewing output, and handling exceptions when the agent gets it wrong. That's not a new job. That's what we've always asked managers to do. We just never gave them the training or the space to do it well.

So let me say the quiet part out loud. Every company that is cutting managers right now is eliminating the exact skillset that every single employee is about to need. They are firing the people who know how to question assumptions, give feedback, and make judgment calls under ambiguity at the precise moment that those skills are becoming universal requirements. It’s a reflexive, spreadsheet-brained decision to cut what we don’t yet know how to measure.

Not everyone agrees on the org chart outcome. Rebecca Hinds, founder of the Work AI Institute at Glean, argues that AI may actually push teams toward flatter, more IC-heavy structures as agents absorb the coordination overhead that managers traditionally handled. She's probably right about the structure, but her research points to a critical irony: as AI absorbs the coordination overhead, the hard skill of the next decade isn't technical execution; it's relationship fluency.

When your team is a fleet of agents, your job is to manage the seams—the high-stakes handoffs between machine logic and human stakeholders. You aren't just managing code; you're managing trust, context, and the honest feedback loops that keep a machine-led system from drifting off course.

The real shift

The question isn't whether you need fewer managers. The question is whether you're ready for a world where everyone is one.

The designer managing three AI design agents needs to evaluate their output the same way a VP evaluates a team's quarterly work. The junior engineer debugging an AI-generated codebase is doing code review, a management function, whether their title says so or not. The salesperson who just delegated prospecting to an AI agent and is now deciding which of its recommendations to trust? That's judgment. That's leadership. That's management.

The skills that separate good managers from bad ones are the same skills that will separate people who thrive in an AI-mediated workplace from people who drown in it: critical thinking, curiosity, empathy, the courage to tell the truth kindly, and the humility to question your own assumptions. And right now, almost nobody is being trained in them.

Three things I'd ask every leader to do right now:

1. Treat the agent as a direct report, not a utility. If you use AI like a calculator, you get an answer. If you manage it like a contributor—questioning its assumptions and providing the context it can't see—you get a result. The former is a task; the latter is a leadership function.
   

2. Evaluate people on humanity, not output volume. When AI handles the cognitive load, the differentiator is no longer who produces the most; it's who thinks the clearest. I have two young engineers on my team who operate far more as senior leaders than many senior leaders I've worked with. They notice when a teammate is stuck, name it, and have the conversation. They use AI to free up their energy for the stuff that requires a human. That's the new bar.
   

3. Stop cutting managers. Start building management capability everywhere. Every employee who works alongside an AI agent is a manager now. The question is whether they're a good one or a bad one. An occasional workshop and LinkedIn Learning is slop. The moment deserves more.

A high-judgment leadership culture doesn't care about the number of layers on an org chart. It recognizes that management is no longer a job title; it is a universal relational skill. The organizations that figure this out will have an insurmountable advantage over those still flattening their way toward obsolescence.

That's what we built Ren for. Not a training program. Not a course. A new kind of infrastructure designed to help everyone develop the skills—feedback, judgment, and honest conversation—that every employee is about to need, regardless of their title. If that's the gap you're trying to close, come take a look.

— Jonathan 

On May 7th, I'll be in San Francisco speaking at Candidly’s HR leaders summit on exactly this question.