This CPO regrets that product management exists | Tom Verrilli (CPO of Whatnot)

Tom Verrilli, Whatnot’s chief product officer and formerly a product leader at Twitch and Twitter, joins me to rethink strong product work. Whatnot is a fast-growing live-shopping marketplace, and Tom’s provocative premise is that its product team “regrets” product management exists.

That does not mean PMs are useless, mate. Product began with founders, design, and engineering working out what to make. The role grew because internet companies became too large for one person to hold every decision. But if a PM handles every decision, engineers and designers never build that judgment muscle.

So the issue is not product management itself; it is assuming every pod needs one because a staffing formula says so.

Exactly. Great PMs create clarity, absorb ambiguity, and force calls. But companies scaled into neat units: engineers, designer, PM, manager. Soon even infrastructure has a PM, while capable builders expect someone else to babysit decisions. Specialization cuts both ways, so hire a PM for a concrete need, not the org chart. At Whatnot, anyone can lead product work, but everyone does the real work and passes the same review bar.

PMs work on meaningful problems rather than belonging permanently to teams. An engineering group may run product work for a year without one; PMs move toward the hard, important work. This was not born from AI. At Twitch and Twitter, I saw ratio-heavy organizations and alarming quantities of alignment meetings, often compensating for management not seeing reality. Leaders with real context can decide quickly instead of multiplying process.

AI gives experienced ICs absurd leverage: inspect code, pull nuanced cohorts, and decide while the problem is warm. But it only works with judgment. It also means senior leaders can sometimes stop an alignment debate—if they genuinely know what is happening on the ground. Whatnot has just over twenty PMs, loosely covering buyers, sellers, trust, and risk. Every six months leadership defines outcomes and crucial projects, then assigns direct accountability.

Of more than thirty-one thousand PM applicants over two years, Whatnot hired one. What is falling out of favor?

Candidates whose story is stakeholder management and alignment are trending down. I used to be that person: the specialty was politics, not customers or technology. We want people who can hold the macro system in their head while being impatiently specific about the next validation. Have you built and iterated something difficult, made calls, and changed its trajectory—not merely polished an established machine?

Systems thinking is practice. For every experiment: if it succeeds, then what? If it fails, then what? If neither changes the plan, we have not thought hard enough. Consider a thousand times the usage, what breaks, and second-order effects. Then know, then go: reason through risks, but do not let imagined objections freeze you. High-growth work seeks compounding moves, not static five-percent gains.

Senior PMs should remain ICs. Old ladders promoted strong doers into coaching and organizational weather. At Whatnot, even PM managers spend nearly all their time hands-on; I still spend about half of mine there. IC work is whatever ships the thing: support tickets, data pulls, sitting with design and engineering, understanding code, writing the spec, running the stand-up. I have put some code into production, mostly to learn; I am not pretending to be an L4 engineer.

One owner for connected decisions eliminates politics. At Twitch, ads and discovery fought over feed impressions. Put both under one accountable PM and optimize the business result, not a local metric. You can pay unusually impactful ICs at VP-level rates rather than funding a huge management pyramid. Senior ICs see more of the board, make better trade-offs, and stay connected to truth.

AI’s immediate gift is data access. It can trace a user’s experience, find similar users, estimate impact, detect regressions, and query code instead of interrupting engineers. But AI analysis is not automatically good analysis. Without sound data engineering, tracking, attribution, and labeling, people will confidently misunderstand a messy system. The leverage belongs to people who verify. Cheaper experimentation may mean more small, technically led teams taking serious swings at work once too expensive.

The durable PM skills are customer understanding, business understanding, technical fluency, and translating among all three—not strategy decks and storytelling. I do not buy “hire brilliant people and disappear.” We operate more like verify, then trust. Leaders go deep enough to understand reality: our founder might clear his calendar, pull up tickets, data, and code, and find the answer with the team. Shared evidence is unblocking, not micromanagement.

With product-minded founders, do not add a redundant layer. Let them own the problem and cover elsewhere. Product leadership translates founder intuition into reality, not competition for vision. Push back from curiosity: ask what context informs the view, bring data when requested, and if it is only two opinions, check your ego.

Our rhythm is “play the accordion.” Pull outward: what are we trying to accomplish, what did we learn, and what follows? Then compress into the smallest useful version and ship it. Roadmap theatre loses software’s learning advantage; pure iteration loses direction. At Whatnot, sellers can hold up AirPods and auction them. Formal listings would add friction, but without listings search cannot route an AirPods buyer. Zoom out, then back in, so you do not fix one local problem while breaking the marketplace.

Agentic commerce will be terrific for air filters, light bulbs, urgent black shoes, and other high-intent tasks. But much shopping is discovery, taste, and social experience. Live commerce brings internet scale to that human part. Fifty people in a stream sounds tiny in media economics; a physical shop with fifty customers would never want to close.

Twitter taught me genuine product-market fit and the damage weak leadership does. People loved it, yet an expanded character limit endured working groups and design cycles because nobody made the call. Sometimes the issue is not complexity; leaders will not decide. My recurring failure is trusting averages too much. A feature used by three percent may be everything to a specific group. Go beneath the metric; averages lie to individuals.

This model fits Whatnot, not necessarily every company. But it is a great moment to return to actual product work and leave more theatre behind. Assume half of what I said is wrong, like any product bet. Still, I would rather have fewer, senior, deeply engaged builders finding truth and shipping than a larger machine optimized for talking about it.

For the lightning round: The Hard Thing About Hard Things, The Purpose Driven Church for community builders, and Babel. I enjoyed Apple TV’s Star City, and Services New South Wales is a genuinely excellent government app. My motto: “I’ll figure it out.” And my most entertaining purchase was a live California spiny lobster from a Whatnot seafood stream. I did not wake up intending to buy a lobster. It was delicious. Find me as TD Robo on X or LinkedIn for product thoughts, Warriors complaints, and candid Whatnot feedback.