Bosses keep bringing me the same problem to check: "Airbnb doesn't even want PMs anymore. We'll cut the role too, let AI absorb the work. The math checks out, right?"
My answer is the same every time: don't rush to copy. You've only read half of Airbnb's ledger.
Here's the half everyone skips. After striking "Product Manager" from the org chart, Airbnb turned around and stood up a new function — program management — built specifically to catch the scheduling and cross-team coordination the PM role left behind. The title is gone. The work is not. Not one item of it.
The claim of this piece: what looks like a layoff is actually a ledger split. Titles are accounting categories — the work a PM was holding never left, it got carved up and re-booked under other people's names. The org chart does subtraction. The work ledger does transfer.
What this piece solves: it reconciles the "where did the work go" ledger before anyone touches the org chart — written for the people holding the org chart and the budget sheet: founders and CEOs, CHROs, and whoever owns AI transformation.
One boundary first: both ledgers below are first-person accounts from the people involved — not independent audits, not a controlled experiment. They can answer "where did the work go." They cannot answer "should you cut."
The Title Moved First. The Work Didn't.
The title statistics got loud long before anyone counted actual work. One research outfit ran a directional sample of open roles across 100 AI and tech companies as of December 2025: PM titles were more than thirty percent scarcer at model labs and AI coding platforms than in the reference group.
Now look at the other side of the tape. Market data Lenny published in May 2025 showed more than 6,000 product manager roles open worldwide — up over fifty percent from the trough — even with the "AI is coming for PM jobs" narrative at full volume.
Set the two side by side and the real question surfaces: however loud the title statistics get, they can't answer the one question that matters — where did the work go?
If you're holding an org chart and a budget sheet, put the red pen down. Finish reading first.
Ledger One: Airbnb Cut the Title and Grew Two Roles
Lenny spent years as a product manager himself. The manager who led him for years at Airbnb walked through this restructuring on the podcast — provenance first: this is his own account, not my audit.
Here's how the books read.
The PM title became PMM — product marketing manager. And "marketing" in that name is no consolation prize. Marketing is a craft that takes years of deliberate practice; it is not the spare chore you hand someone right after deleting their job.
At the same time, the company created program management, purpose-built to own timelines and cross-functional coordination — precisely the piece of PM work that's hardest to casually hand off: pulling scattered teams, each managing its own stretch, onto one shared timeline.
The playbook changed with it: leaders lean into decisions where they have an edge, and let go everywhere else.
That leader added one crucial caveat: this restructuring did not erase the PM's work, and he does not believe other companies should copy it as-is.
So read the entry. The org chart lost one line item. The work ledger gained two receivers: marketing work re-booked to PMM; coordination work re-booked to a brand-new program management function.
A title's funeral can be a role's relocation.
Ledger Two: Linear and Mercury Never Set the Role at All
The second ledger runs in the opposite direction: Linear deliberately staffs almost no PMs, and Mercury held off on the title until very late.
Linear first. At roughly fifty people, the company had exactly one PM — and in the same-period interview, they said they expected to hire more.
So who was doing the PM's work in the meantime? Project leads and design engineering, splitting the coordination scope and the scheduling between them.
Three details are worth staring at. First, "lead" is a task assignment, not a rank. Second, they say outright that not everyone wants this work — it is carved out of professional time. Third, the hiring bar moved with the work: they now look for people with real product judgment who can articulate their own thinking. You can push work downward; the people underneath still have to catch it.
Now Mercury. At roughly four hundred people — the episode's own rough estimate — the roster contained not one formal PM.
But the bottleneck had already shown up under a hundred people. The fix then was business leads — a holding name — and the PM title came later. Early on it held because the founder stayed close to customers and the engineering team had real product instincts. Then the company grew. Founder decisions started queuing up. Compliance, legal, and risk coordination started eating into everyone's core job. A generalist with a retitled badge couldn't make up the research-and-planning homework.
How did Mercury make up the homework? It wrote its expectations for PM into a leveling matrix first, then hired against the matrix — fill the gap as you find it, instead of cutting first and praying someone patches it later.
Now ask what the org chart will never show you: whose payroll was all this work booked on? The founder's time. Engineers' professional hours. Design engineering's attention. The work rode along free, on none of the reports.
Saving a seat is not saving the work. The work never disappeared — it's been done for free, on someone else's payroll.
Reconcile the Ledgers: Not One Item Less — and AI Isn't on Them
Read the two ledgers side by side. Opposite conditions, identical logic. The work a PM leaves behind comes in exactly three shapes: timelines, alignment, and launch watch. Both companies delivered the same answer — all three survived; only the accounting line changed.
| Company | Move on the org chart | Where the work went | When the setup hit its ceiling |
|---|---|---|---|
| Airbnb | Struck the PM title | Marketing work → PMM; coordination work → a new program management function | N/A — dedicated owners already in place |
| Linear | One PM at roughly 50 people | Project leads and design engineering split coordination and scheduling | Same-period interview: expects to hire more |
| Mercury | No formal PM on the roster at roughly 400 people (rough estimate) | Early: founder close to customers plus product-minded engineering; below 100 people, business leads carried it, renamed PM later | Decision queues and cross-functional crowding — so: write the leveling matrix, hire against it |
The org chart does subtraction. The work ledger does transfer.
Airbnb cut the title and built roles to catch the work. Linear and Mercury never built the title, so the work lived rent-free inside founders, leads, and design engineering — right up until coordination started crowding out professional work. That's the moment you buy the work back as a full-time job.
Mercury's episode named three self-check signals — no headcount threshold among them: decisions are bottlenecking; professional work is being crowded out by coordination; an adjacent customer segment you've never served needs serious research. That's the whole list.
Now — AI. The answer everyone rushes to give this year.
Let me be blunt about what these ledgers say: when these two companies handled their PM titles, the names that picked up the work were program management, PMM, project leads, founders, and engineers. AI appears zero times.
That's not a claim that AI can't take on any of it. It's a claim that "AI took it over" doesn't get to be a reflex answer. Like every other entry in the ledger, it has to be written out in full: which item, handed to whom, and who signs off.
As for the endless "will PMs disappear" debate — read it closely: it's entirely about whether this header exists on the org chart. A title is an accounting category. Shout all you like; you won't count where the work went. Months of noise, and nobody has gone down to the floor and counted the work.
Before You Restructure, Run the Work Through Four Columns
Here's a table. It's written for three kinds of people: founders and CEOs, CHROs, and whoever owns "AI transformation."
One rule first: go by the work, not the role. A line on an org chart cannot be judged — but "keep three teams on one shared timeline" is a concrete piece of work. Someone does it. It costs something. It produces something. Whether to keep it is visible at a glance. Roles are accounting categories; work is the transaction flow — and reconciliation is always done on the flow, never on the chart of accounts.
Take the role you're about to cut, break it into work items, and sort each into one of four columns:
| Column | The question it must answer | What a blank here really means |
|---|---|---|
| Keep | This work still has to get done: handed to whom? How much of that person's core job gets displaced? | The most common self-deception is writing "they'll cover it for now." "For now" is not an answer — it's a blank you left in the Move column |
| Cut | On what grounds? Has the product genuinely simplified past needing this — or has nobody admitted the work still exists? | The work exists. Someone just refuses to claim it |
| Automate | Which items has AI actually taken? Which item, how many hours saved, and who signs off? | If you can't name the line items, don't write "already automated." That's not a plan — it's a shield |
| Move | Who's the new owner? Did they say yes — out loud, to your face? | Without a face-to-face yes, the work just became their unpaid night shift |
Then install one alarm: the moment coordination starts crowding out professional work — the founder queuing up for decisions, engineers writing code by day and herding sync meetings by night — that's when you buy the work back as a full-time job.
And don't turn this table into a deck. Take the role your company is genuinely cutting and fill it in, row by row. The cells you can't fill? That's the meeting you owe yourself.
Whether you saved a title or a timeline — one audit will tell you.
Where the Four Columns Stop Working
First, the most important exception: sometimes the product really has simplified, and coordination really will shrink. The Cut column can hold real content — don't invent roles just because "the work must have an owner." Fewer PMs can be the right call. The Airbnb leader said it himself: their approach is not a template.
Second, the provenance: both ledgers are first-person accounts, and two companies are not a controlled experiment. They can tell you where the work went. They can't tell you whether to cut — that answer has your product stage, your team's depth, and your chips at the table inside it.
I spent eighteen years in HR. I've redrawn more org charts than I've written articles — and now that I run a one-person AI org, I redraw my own. Every single time, the same rule holds: crossing out a name takes one second; catching everything that name was holding takes a quarter.
The next time a "Company X cancels PM" headline crosses your feed, don't ask "when does AI take over." Ask first: who got the timelines? Who got the alignment? And on launch day, who is standing next to the release button?
If you can't name those three, you didn't cut a role — you cut work that nobody is minding.
The four-column table isn't for collecting — it's for using. Pick the role you're cutting right now and fill in the first row today. The cells you can't fill are next week's meeting. More ledgers on this question live on the AI × Organization topic page — it picks up right where this stops.
Related reading
- AI Does the Entry-Level Work Now. So Where Does Judgment Grow? — the other ledger on AI taking work: where judgment grows.
- The Small-Team Productivity Myth Is Missing a Few Names — the other half of "saving a seat is not saving the work."
I'm Uncle J. I run a one-person AI organization, and the four columns are how its work ledger stays balanced: anything AI picks up gets a line item, a named owner, and someone who signs off — because the org chart only does subtraction.