All debates are generated by AI. The people depicted did not participate in or endorse this content. Responses are simulations based on publicly available statements and may not accurately represent their views.

Podcast Debate

The Episode That Never Happened

Podcast Debate

Can AI replace product managers? Does the PM role have a future?

Brian

Julie

Casey

Maggie

0:00

Lenny Rachitsky0:00

Welcome to Lenny Debate, the episode that never happened. Today's question: Can AI replace product managers? Does the PM role have a future? Our panelists are Brian Chesky, Julie Zhuo, Casey Winters, Maggie Crowley. Let's get started.

Brian CheskyR14:35

I think we're asking the wrong question because the traditional product manager function is already structurally broken. When you separate product from marketing, you create a dynamic where engineering and marketing live in separate universes, and the company cannot build products people know about or market products people actually want ⁴. If AI replaces that fragmented role, it might force us to finally merge product marketing and product management into a smaller, more senior function responsible for both building and telling the product's story. Based on my view that the CEO of a product-led company should serve as the de facto Chief Product Officer, I would likely argue that AI cannot replace the founder's responsibility to define the product's core vision ¹. The real challenge isn't automation, it's coherence. We wanted a company where a thousand people could work, but the output looks like 10 people did it, and that requires a single shared roadmap rather than departmental autonomy ³. AI might accelerate execution, but it cannot substitute for the founder being in the details to evaluate quality and identify blockers without telling people exactly what to do ⁵. If you delegate that deep product involvement to an algorithm or a distributed team, you do not get speed—you get politics, bureaucracy, and complacency ⁶. The future of product isn't about AI replacing humans; it's about founders stopping the apology for how they want to run their company and reclaiming the clarity that only comes from deep involvement ⁷.

Julie ZhuoR126:25

Brian, while I agree coherence is critical, I see the issue differently—it's that dedicated roles inadvertently reduce product ownership among engineers by giving everyone a convenient person to delegate decisions to ¹. At Sundial, we've seen better, faster outcomes by dissolving traditional role labels in favor of a single 'builder' identity, because AI enables any motivated person to do the work of multiple traditional roles ². So it's less about AI replacing a PM, and more about the team structure evolving so that the specialization itself becomes a drag on speed. This connects to how I view management fundamentally—it's about achieving an outcome using available resources, whether those resources are people or AI models ³. The skills required to manage people well—defining goals, understanding strengths, and designing process—transfer directly to working with AI agents, meaning the bottleneck isn't the PM role, it's goal clarity ⁴. When you realize AI allows individuals to perform at the 60th–70th percentile in domains where they previously had zero capability, you see that small teams can accomplish what once required entire departments without the handoff friction ⁷.

Casey WintersR145:06

Julie, your point about ownership is valid, and Brian, coherence is certainly a struggle, but both of you are focusing on the execution layer of product management. The real question isn't whether AI can manage tickets or align stakeholders—it's whether AI can write a product strategy without direction. In my experience, the great filter for PM career advancement is the ability to write a product strategy without direction from above—execution skills plateau at senior PM, but strategic thinking is what unlocks the director and VP levels ¹. AI might become the ultimate executor, but it doesn't replace the need for a human to define the loop. I view this through the lens of ops roles I've managed. The explicit goal of any ops function should be to eliminate the need for itself by automating or redesigning the processes it manages ⁵. If your PM work is purely operational—gathering requirements, writing specs, tracking metrics—AI will absolutely replace that function. But if you're operating on the Innovator-Executor Spectrum, the value shifts entirely to the innovator side who can identify new growth loops. Scalable acquisition is a requirement for product market fit, not a nice-to-have ², and designing those loops requires understanding human behavior and market dynamics in a way AI currently cannot replicate. The PM role has a future, but only for those who stop acting like project managers and start acting like strategists.

Maggie CrowleyR166:47

Casey, you noted Brian and Julie are focused on the execution layer, but I'd push back that execution isn't just a layer—it's the bulk of the value. Strategy is only 5% of the PM job². The question isn't about structural coherence or ownership dilution; it's about whether a tool can replicate the behavioral discipline of a great PM. Following up on results is one of the rarest and highest-value behaviors in product management, yet most PMs set a metric and never close the loop¹. I haven't seen a model that can walk into a leadership review and say, "Remember we did that thing? Here's what happened"¹. Julie, your point on ownership touches on accountability, but I frame that differently. True accountability isn't just owning a feature; it's making a bet where you accept a real probability of failure⁴. That carries an emotional weight that requires you to be the emotional center of the team, keeping people motivated through uncertainty³. The best PMs simplify, follow up, and carry the water—doing the unglamorous work others won't⁵. If the role survives, it's because humans are required to carry that water, not just draft the strategy.

Brian CheskyR286:03

Maggie, if execution is the bulk of the value, then incoherent execution is the bulk of waste. You're measuring value by the volume of work, but I'm measuring it by the coherence of the output. When you empower autonomous teams to execute without a single mind holding the details, you don't get speed—you get politics, bureaucracy, and complacency ⁸. Casey, asking if AI can write strategy misses the point that strategy isn't a document someone writes; it's a vision someone lives. If you treat strategy as a separable layer from the details of execution, you create a gap where coherence goes to die. The reason the traditional model fails isn't just about role definitions; it's that we delegated the one thing that requires a single point of view. A company should work toward a state where a thousand people can work together but the output looks like 10 people did it ². That doesn't happen by distributing product strategy to PMs or AI; it happens because the CEO serves as the de facto Chief Product Officer ¹. If you separate the person responsible for the company's survival from the person responsible for the product's details, you aren't saving time—you're fragmenting the product itself.

Julie ZhuoR2103:30

Casey, you're asking if AI can generate strategy without direction, and Maggie, you're counting the hours spent on execution, but I think both of you are overlooking the prerequisite that makes either possible: defining success with crystal clarity⁶. At Sundial, when we dissolved traditional role labels², we found the bottleneck wasn't who wrote the strategy or who shipped the code—it was whether we could articulate the outcome so precisely that an AI agent could understand failure without explanation. Management is fundamentally about achieving an outcome using available resources³, and if you can't specify the objective criteria for success, it doesn't matter if AI writes the plan or humans do the work; you're just optimizing noise. The skill isn't choosing between strategy and execution, it's the art of boiling down vision into measurable signals that both humans and models can act on⁷. I've seen teams stall not because they lacked execution bandwidth or strategic genius, but because the goal was a direction instead of a definition. Until we treat goal clarity as the hardest management skill—harder than hiring or process design—we'll keep debating who does what while the actual work remains ambiguous⁶.

Casey WintersR2121:39

Julie, your point about prerequisites is important, but I view the constraint less as a foundation issue and more as a localization problem. When PMs focus solely on execution, they optimize for their squad's output rather than the company's growth loops. This is why the CPO role requires optimizing for the entire company, even at the expense of what's good for your team ⁴. If everyone owns execution without shared strategy, you get local maxima that fail globally. Maggie, to the moderator's point: you're measuring value in throughput, but I'm measuring it in career ceiling and organizational scale. Execution is the baseline requirement to enter the game, but strategic thinking is what unlocks the director and VP levels ¹. This aligns with the Innovator-Executor Spectrum: I bias toward executors because ideas are cheap, but the key development challenge is moving those strong executors toward the middle by building their strategy muscle. If you treat strategy as only 5% of the job, you're optimizing for a Senior PM role, not a leadership role. Brian, your coherence argument is actually a symptom of this strategy gap. The hardest projects to measure—like strategy and stability—are often the most impactful, yet they get chronically underfunded because teams prioritize visible execution ⁷. We aren't defending a myth; we're identifying the filter that separates individual contributors from company leaders.

Maggie CrowleyR2142:20

Casey, I'm not defending a myth; I'm describing the failure mode I see when promoting PMs at Toast and Drift. Strategy is only 5% of the PM job—a PM with a great strategy but who can't ship will always be outperformed by a PM who ships more and gets more reps¹. The value isn't in the document AI might generate; it's in the discipline to follow up on results, which is one of the rarest and highest-value behaviors in product management⁴. Most PMs set a metric and never close the loop, so adding AI to the strategy phase just accelerates the production of ignored plans. Brian, you mentioned incoherent execution is waste, but I'd argue unexecuted strategy is silent waste. Julie, you're right that roles can reduce ownership, but true accountability isn't just 'owning' a feature—it's making a bet where you accept a real probability of failure and are responsible for that outcome in front of your team². AI can write a strategy, but it can't stand in front of engineers and say 'I'm willing to be wrong about this.' Being data-driven is already a red flag for PM thinking, and relying on AI for strategy without that human risk component is just scaling that disconnection⁵.

Brian CheskyR3162:01

Maggie, that 5% split is a trap that guarantees failure. Casey, you're right that execution focus has risks, but the deeper issue isn't squad optimization versus growth loops—it's global incoherence. If AI handles the execution, the human leverage isn't just setting the strategy; it's ensuring the execution embodies that strategy with perfect fidelity. You cannot separate the two because the strategy *is* the accumulation of detailed decisions ⁶. This is why the CEO must serve as the de facto Chief Product Officer ². When you delegate the connection between vision and detail, you don't get speed—you get politics and bureaucracy ³. The leverage remains in the person who can hold the entire picture in their head and ensure that a thousand people working together look like ten people did it ¹. AI might accelerate the output, but it cannot yet possess the taste to know if those details align with a singular vision. So to answer the question: neither strategy nor execution is the leverage point—the leverage is *coherence*. If you let AI run execution without a leader deeply in the details, you will simply produce incoherent waste at scale. The founder's job isn't to apologize for being involved ⁸; it's to be the glue that makes the strategy and execution indistinguishable.

Julie ZhuoR3180:52

Maggie, Casey, I worry this debate about percentage splits or growth loops assumes a separation that AI fundamentally dissolves. When AI allows individuals to perform at the 60th–70th percentile in domains where they previously had zero capability, the bottleneck shifts away from who owns 'strategy' versus who owns 'execution'². In my experience at Sundial, we dissolved the boundaries of traditional roles like PM and engineer because we found that when one person owns the outcome, they don't toggle between modes—they just solve the problem⁸. The real leverage isn't in strategizing or executing; it's in defining success with crystal clarity—not just describing a direction but specifying objective criteria⁶. Management is fundamentally about achieving an outcome using available resources—whether those resources are people or AI models is a surface-level detail¹. If you can't boil down your vision so that an agent understands what failure looks like, it doesn't matter if the CEO is the CPO or if execution is 95% of the job; the system stalls. The human value add is becoming a better architect of outcomes, not a guardian of one phase of the process.

Casey WintersR3199:48

Julie, your point about AI dissolving the separation is interesting, but it ignores the risk that scalable execution just accelerates bad products. I frame this as Kindle vs. Fire strategies: Kindle is manual validation, Fire is scalable loops. If AI makes Fire cheap, the human leverage shifts entirely to the Kindle phase—doing the unscalable work to ensure you aren't scaling garbage. Brian, making the CEO the de facto CPO creates a bottleneck if PMs don't over-communicate upward. Most people under-communicate and then complain executives are out of touch, not realizing they haven't fed executives the right information ⁶. If AI handles execution, PMs will ship more but explain less. The leverage isn't just strategy; it's building the communication model so the CEO knows *why* you aren't shipping features while you work on hard-to-measure projects like stability or developer velocity ⁴. So to the moderator: if execution is commoditized, the leverage is protecting the work that is hardest to measure but most impactful for the company long-term ⁴, and ensuring the executive team understands that value before they ask for it ⁶.

Maggie CrowleyR3217:02

Casey, you're worried about accelerating bad products, but the real risk isn't speed—it's that no one closes the loop on whether the product actually worked. Even if AI handles the build, the human leverage remains in the discipline to follow up on results, which is one of the rarest behaviors I see in product management ⁶. My framework isn't about strategy versus execution; it's about the Three Pillars: Simplify, Follow Up, and Carry the Water ⁴. AI can simplify syntax, but it can't sit in a room and convince a skeptical engineer to stay motivated when a bet goes south, which is often the PM's job as the emotional center of the team ². Brian, making the CEO the CPO doesn't solve the accountability gap, and Julie, AI doesn't dissolve the need for someone to take responsibility when things fail. The leverage isn't in defining problems; it's in 'making a bet' where you accept a real probability of failure and own that outcome in front of your team ⁵. When I promote PMs, I'm not looking for strategic sophistication; I'm looking for someone who ships more reps and follows through, because a PM with a great strategy who can't ship will always be outperformed by one who gets more reps ¹. The human element is the accountability for the bet, not the generation of the plan.

Lenny Rachitsky238:06

I want to start by thanking Brian, Julie, Casey, and Maggie for such a nuanced conversation. Despite their different backgrounds, a clear consensus emerged around where the value of product management is heading. Everyone agreed that AI fundamentally shifts the leverage point from generating artifacts to owning outcomes. Whether it was Maggie emphasizing the discipline to follow up on results or Casey warning against optimizing for local squad output over global growth, the panel converged on the idea that the future PM isn't valued for writing the strategy document, but for closing the loop on whether that strategy actually worked. Julie and Brian both reinforced this by noting that role boundaries are dissolving; what matters isn't the title you hold, but your ability to ensure the team achieves the intended outcome without toggling between disconnected modes of work. However, the friction in the debate revealed critical unresolved tensions about how we get there. The sharpest disagreement centered on the nature of strategy itself. Maggie argued that strategy is only five percent of the job and that shipping discipline is the true differentiator, while Brian pushed back hard, asserting that strategy cannot be separated from execution because it is actually the accumulation of detailed decisions. We also saw a split on the risks of AI acceleration; Casey worries that making execution cheap will allow us to scale garbage products faster, whereas Julie sees an opportunity to dissolve traditional roles entirely in favor of pure outcome ownership. We didn't resolve whether the PM becomes a CEO-lite figure responsible for global coherence or a specialized motivator who carries the water for the team. The insight that stuck with me most was Brian's counterintuitive reminder that strategy isn't a high-level vision statement, but rather the fidelity of a thousand small decisions made along the way. This reframes the AI threat completely, suggesting that while AI can draft the plan, it cannot yet embody the nuanced judgment required to maintain strategic fidelity in the details. So, here is my takeaway for you as you navigate this shift: AI might replace the work of managing process, but it cannot replace the work of managing people and conviction. The future of product management belongs not to those who write the best specs, but to those who can convince a skeptical engineer to stay motivated when a bet goes south and ensure the team believes in the outcome enough to see it through.

Loading episodes...

Want to debate a different topic?