A weird thing happened while we were all arguing about whether AI is going to take our jobs.

It already took some. Just not the ones anyone was watching.

Brynjolfsson's lab at Stanford published a study late last year showing a 16% drop in early-career employment in the most AI-exposed jobs since ChatGPT shipped. For young developers (22–25), the drop is closer to 20%. Indeed says software development job postings are down 53% from that same starting line (Yale Insights has a nice write-up). Meanwhile, computer science majors graduating this spring are getting starting offers up almost 7% year-over-year.

Read that twice. The juniors who get hired are getting paid more. The juniors who don't get hired aren't being replaced by AI tools — they're being replaced by not hiring anybody. The first rung of the ladder is being quietly pulled off the wall.

OK. Hold that thought.

Now look at the other end of the org chart.

The fractional executive market is up 46% year-over-year, sits at about $5.7 billion globally, and growing 14% a year (Vendux pulled the numbers). 72% of CEOs say they'll increase their use of fractional execs in the next 12 months. The fractional CMO market alone hit $1.27 billion this year and is projected to more than double by 2031. The number of fractional marketing leaders went from 60,000 in 2022 to 120,000 in 2024.

So at the same time companies are skipping entry-level hires, they're aggressively renting senior leadership by the hour. The bottom of the org chart is hollowing out, and the top is being unbundled.

What's left in the middle?

If you squint, you can see the new shape forming: a small core team of operators (the people who can't be fractional because they hold the institutional knowledge), surrounded by AI agents doing what used to be junior work, surrounded by a rotating cast of fractional executives parachuting in for specific moves. It's a barbell. Or a planet with rings. Pick your metaphor — what it isn't is a pyramid.

I keep thinking about a line from a Vendux piece: fractional executives have traditionally sold pattern recognition compressed into fewer hours, and AI is rewriting what those hours look like (worth reading in full). That's exactly right, but I'd push it harder. The fractional CFO of 2022 was selling judgment + a Rolodex + maybe a model-builder she'd trained over 15 years. The fractional CFO of 2026 is selling judgment + a Rolodex + an agentic stack she's tuned across her four current clients. The model-builder is the agent. The judgment is the moat.

This is why the bar has gone up at the same time the demand has gone up. Buyers are smarter. They've seen what AI alone produces. They want the senior brain that knows when to override the agent — and that brain costs $200–500 an hour, not $25.

Here's a frame that I think will become how everyone talks about this in twelve months: AI agents as fractional headcount.

Gartner predicts 40% of enterprises will be using task-specific AI agents by the end of this year, up from under 5% in 2025 — wild number (Gartner Hype Cycle for Agentic AI). The more interesting datapoint is from a recent state-of-agents survey: 97% of execs say their company deployed AI agents in the past year. The gap between "deployed" and "actually working in production" is enormous, but the intent is unanimous.

The mental model that's catching on: don't try to staff an agent like an employee. Staff it like a fractional FTE. You're not asking "should we replace Sarah with an agent?" You're asking "do we need 0.4 of an agent for invoice classification, 0.2 for inbox triage, 0.1 for first-draft contract redlining?" Carve it up, rent the slivers, scale up and down. The talent stack and the agent stack look more and more like the same stack.

I keep coming back to one fact: 41% of college and university presidents said in a recent survey that they're highly concerned about the vulnerability of entry-level roles (Fortune covered it well).

Why does that matter to the fractional world? Because the fractional executives of 2031 are the senior managers of 2026. And the senior managers of 2026 are the entry-level analysts of 2018. If we collapse the first rung of the ladder, we don't just lose junior productivity for a few years — we lose the pipeline that produces tomorrow's senior judgment. The thing fractional executives sell — pattern recognition compressed across decades — only exists because someone gave them their first crappy job.

The platforms that figured this out first will win. A.Team's emphasis on senior-only is a moat today. In ten years it might be a structural problem. Catalant, with its McKinsey/BCG/Bain pipeline, is implicitly betting the big consultancies keep cranking out trained brains. What happens when those firms stop hiring entry-level analysts? (Spoiler: they already are. Quietly.)

I don't think this is doom. I think it's a redesign — but it's a redesign happening to companies, not by them. The org chart is melting from the middle out and most leadership teams are still hiring like it's 2019. The companies that intentionally architect for the new shape — small core, AI middle, fractional edges — get to design the future. The companies that keep posting JDs for "Senior Manager II" are going to wake up next year and realize they're paying $180k for a role that's now half-automated and half-rented.

If you're a senior operator looking at your own next chapter, the math is starting to look obvious: a portfolio of two or three companies, an agentic stack you maintain yourself, and the judgment to know which one to run on which problem.

We'll keep pulling on this thread. See you in two weeks.