AI Won't Replace Product Managers — Just Their Worst Half

 

      On August 2, Lenny's Podcast — basically required listening for product people in Silicon Valley — dropped an episode with a spicy title: "This CPO regrets that the product manager role exists."

      The guest, Verrilli, is no outsider. He's the Chief Product Officer at Whatnot, previously CPO at Twitch, and before that, director of product growth at Twitter. A guy who climbed the product ladder all the way to the top, turned around, and said: this job should never have existed.

      And his company makes it hard to dismiss as venting. Whatnot is a livestream shopping platform — hosts run live auctions; it started with trading cards and collectibles, now it sells everything. In 2025 it did $8 billion in sales at an $11.5 billion valuation — with just 20 product managers out of 1,200 employees. Silicon Valley's rule of thumb is one PM for every five engineers. Over the past two years, 31,832 people applied for a PM job at Whatnot. They hired one.

      This isn't a dying company cutting costs. It's one of the fastest-moving companies around saying: I don't need many product managers — and I'm still outpacing you.

1. Why he hates the role

      Verrilli gives three reasons, each sharper than the last.

      First, the role rewards the wrong skills. What's valued most isn't understanding technology or customers — it's politics: driving alignment, building relationships, telling good stories upward. People do politics and call it product management. He calls it an anti-pattern.

      Second, it's not the people — it's the system. He points to Twitter's endless "alignment meetings" and offers the most quotable line of the whole interview: your system breeds systems. Build an organization that rewards performative alignment, and it will grow you a generation of PMs whose profession is alignment. These people aren't bad; the org simply pays them to that standard.

      Third, PMs turn engineers and designers into children. They're perfectly capable of making good calls — they just never have to, because a PM is always there to babysit. Junior PMs have gone from "CEO of the product" to "button babysitters."

      I've lived this myself. At Cheetah Mobile — a Chinese app company whose founder, Fu Sheng, branded himself "China's number one product manager" — product management was the only path to promotion. And how do you get promoted? By grabbing authority: "No, you don't understand product. I'll tell you how it's done." Over time, the whole system becomes a political system.

2. He's not against alignment — he's against alignment as a career

      One thing needs to be clear, or you'll misread him: he's not advocating chaos. At Whatnot, you get autonomy over execution; nobody gets autonomy over strategy. His own words: "Without alignment, autonomy is wasted."

      What he opposes is a class of people who coordinate for a living. What he wants is a small number of senior, hands-on PMs who dare to make the call — because in a big, inertial organization, going with the flow is the easy path, and being the decision-maker is exactly what you hire a PM to do.

3. The job was born political — in 1931

      On May 13, 1931, a P&G advertising manager named Neil McElroy wrote an 800-word memo. He was pushing a soap called Camay, and his biggest rival wasn't on the market — it was inside his own company. Ivory, P&G's flagship, held all the resources, and Camay got blocked by its own colleagues at every step.

      The memo looked like a staffing request — "I'm drowning, give me two more people" — but what it really proposed was this: every brand needs one person who owns it end to end, watching sales, running product, ads, and promotions, and getting out into the market to meet real customers. His boss read it and reorganized all of P&G around brands. Every P&G CEO since has come up through brand management. McElroy himself became president in 1948, and later Eisenhower's Secretary of Defense.

      See the point? This role was built from day one to fight internal wars — to coordinate resources inside a company. The politics and alignment Verrilli rails against aren't a corruption of the job. They're its birthmark.

4. The internet grew the role a second leg

      So why does everyone today think of PMs as "the person who defines the product"?

      P&G's brand managers defined products too: analyze competitors, claim a lane — say, "smoothness" — then get packaging from design, the formula from the lab, and airtime from advertising. That's how the shampoo brand Rejoice was born. But what they delivered was a brand, not a feature. Today's internet PMs ship apps and features, lead their own engineers and designers, and decide "who calls the shots" on their small team. The classic case in China is Tencent — its early nickname among users was "the damn penguin," because it was so good at copying. And "copying well" is exactly what tests a PM: tear a rival's product apart, understand it completely, then redefine it with your team.

       So the job has always walked on two legs: one is coordination and politics, the other is understanding and judgment. In P&G's era, the first leg was stronger. Internet companies emphasized the second. But the first leg never went away — the bigger the company, the faster it grows back.

5. AI is taking exactly that first leg

      For a while, the trendy way to use AI was role-play: "You're the designer, you're the tester" — and let several agents discuss like a human team. People who tested it found the results were bad, because it ports humanity's bureaucracy onto AI. Humans need meetings and handoff docs because we can only hold so much information at once. AI just calls all that "context" — it can absorb a million tokens (roughly hundreds of thousands of words) in one go. No relay required.

      So frontier labs like Anthropic changed course: one unified agent that takes the task, breaks it down itself, delegates to sub-agents, and delivers the summary.

      Notice — "break down, coordinate, summarize" is exactly the PM's daily grind. The most toxic half of the role is precisely the half AI is best at absorbing. The other half? AI can't take it.

6. What's left for humans

      Engineers: from writing code to drawing boundaries. With vibe coding — describing what you want in plain English and letting AI write it — building real systems actually demands more: you don't have to write memory management by hand, but you must understand it, and you must draw boundaries. Tools like Claude Code and Codex default to "band-aids": the smallest possible fix, because they don't know where their boundaries are. There's a sci-fi story about a computer that could answer anything; asked "what should humanity do?", it thought for ages and answered "42." We don't want that answer — so humans draw the boundaries.

      Designers: from hands to eyes. AI can crank out a hundred drafts in minutes. Who picks? Still the designer — who then combines and refines. The skills that matter now: understanding what the client actually wants; taste — it used to be about the hand (can you draw it), now it's about the eye (can you pick it); plus the communication skills to explain the choice.

      Product managers: keep the judgment, add the labeling. Understanding business, customers, and data, then making the call — that leg stays. The new job is data labeling: AI can't tell which market feedback is a good signal and which is noise. PMs turn real-world data into signals AI can learn from, feed them back in, and iterate fast alongside AI.

7. What really gets rewritten: "the product" itself

      Push one step further: what gets disrupted may not be any job, but "the product" itself. We're used to products having boundaries: separate Android and iOS versions, compatibility headaches, app store rules. But what we actually want is the service and the result, not the product-shaped shell — and the productivity wasted maintaining those boundaries often exceeds what goes into the result.

      When we built Clean Master, a phone-cleaning app, fewer than 10 people actually worked on "cleaning." The product team had two to three hundred. What did the rest do? Maintained boundaries — and monetized, by showing users ads. It's like buying a bottle of Rejoice: you want softer hair, but an army of people is debating whether the bottle should be green. In the AI era, product boundaries may dissolve — or get redrawn entirely.

8. Wrapping up

      So in the age of AI, what actually gets replaced?

      My answer: the defaults. The default of one PM for every five or six engineers. The default that meetings are the job. The default that products must have today's boundaries. A role invented in 1931 has grown parts that are actively harmful to organizations — and those parts are exactly what AI absorbs best.

      The question worth asking isn't "will my job disappear?" It's: strip away the job title — what service and value do I actually deliver? Or more bluntly: how much of my day produces results, and how much just maintains boundaries? I'd love to hear your answer.