At a recent all-hands, the only question in the Q&A was: "How do we deal with AI anxiety?"

Everyone in the room was worried about the same thing — whether their job would still exist. Engineers, designers, product managers. The common thread: "AI can already do most of what I do."

The same week, a friend of mine at ByteDance got promoted and received a 30% salary increase. His team also cut new grad hiring from six spots to two.

He told me this in a completely flat tone: "My work is still beyond what AI can do. But the grunt work? They don't need people for that anymore."

I sat with that for a moment. Because that sentence contains a problem that most people haven't worked through yet.

The Wrong Frame

For the past two years, almost every conversation about AI and employment has been circling the same question: Will AI take our jobs?

The answer, technically, is yes.

But the real story isn't humans vs. AI. At least not in the labor market. It's actually senior employees + AI vs. junior employees. That's a different conflict, and a more uncomfortable one.

I manage a team. Since I started using Claude Code seriously, the number of times I've needed to pull in junior staff has dropped significantly. I'm not unusual.

What the Data Says

Microsoft AI CEO Mustafa Suleyman predicted in 2024 that AI would reshape large portions of white-collar work within 18 months. Many dismissed it as hype.

The data arrived before he said it would:

  • Entry-level job postings have declined by roughly 13%, while senior role postings have held flat or grown slightly
  • New grad starting salary growth is running well below the growth in compensation for workers with 5+ years of experience
  • In consulting, law, and financial analysis, junior-level work is being heavily substituted while senior partners and advisors are charging more than ever

The reason is structural. AI excels precisely at what junior employees are typically hired to do: compiling research, writing first drafts, cleaning data, running baseline analyses, summarizing meeting notes, handling template correspondence.

None of that requires experienced judgment. It's execution-heavy work. That's AI's home turf.

Why Seniority Is Protected (For Now)

The natural question: aren't senior employees at risk too?

Yes — but the nature of the risk is different.

Junior employees' core value is execution. Their job security depends on the implicit bargain: "Human labor is more expensive than AI, but not that much more expensive." Once AI quality crosses a threshold, that bargain breaks instantly.

Senior employees' core value is judgment. Knowing when to push forward and when to stop. Reading what the client didn't say. Making the right trade-off when two reasonable approaches conflict. Deciding how much revenue loss is acceptable to protect long-term retention.

That last question isn't science. It's art.

More importantly: one senior employee + AI is now doing the work of one senior employee + several junior employees. An experienced product manager using Claude for requirements documents, Cursor for prototyping, and AI for user feedback analysis can produce what used to require a team.

This isn't hypothetical. It's already happening in many organizations.

Who's Actually Winning

Senior employees. Their scarcity has been amplified. In the old model, companies hired ten junior employees plus two senior ones — the senior premium was bounded because "more hands" still mattered. Now one senior person plus AI can match ten junior employees. That changes the negotiation entirely.

Goldman Sachs, McKinsey, and similar firms are already showing the pattern: fewer total staff, higher revenue per person.

People who move fast. There's a window here, and it may be shorter than it looks. Using AI tools fluently is a genuine advantage today. In two to three years it'll be a baseline expectation — like Excel. Not impressive; just required.

The people in the worst position are those caught in the middle: two to three years of experience, past pure execution, but not yet at the judgment level where they're truly irreplaceable. AI already handles 80% of their work. They lack the seniority to command the senior premium.

That group is absorbing the most pressure right now.

The Question Worth Asking

I can't give you a five-step plan. But two things seem clear:

Stop asking "will AI replace me" and start asking "what is my value actually built on." If your value comes mainly from execution and pattern-following, that's a real exposure. If it comes from judgment, relationships, creative synthesis, and navigating ambiguity, AI is your amplifier in the short term — not your replacement.

Use AI to expand your range, not to substitute your thinking. Using AI to think for you is a slow path down. Your judgment atrophies. Using AI to do things you couldn't do before is a path up — the scope of what you can contribute increases, and your value increases with it.

My friend's line — "my work is still beyond what AI can do" — is true today.

But that word still deserves attention.

The capability frontier moves every six months. What AI can't do now doesn't mean what it can't do in two years. The real question isn't whether you're safe today. It's whether you're moving your value toward things AI can't reach — and whether you're moving fast enough.