Article details
Recap from Ravi's Reforge Event: Product Management is Not Dead
Every few years, someone declares that the job you do is about to disappear. Ravi — Reforge course creator has lived through enough of these moments to be skeptical. But he understands why this time feels different. The AI era has produced three specific fears that have taken hold across the industry: that creativity no longer matters, that software is obsolete, and that product management itself is dead.
His session was a direct argument against all three – not just a reassuring dismissal, but a structured case for why each fear, examined carefully, points in exactly the opposite direction.
Myth one: Creativity doesn't matter
In 1965 Birmingham, a 17-year-old welder named Tony clocked in for his last day at a sheet metal factory. It was his last day because he was leaving to tour with his band, chasing a dream of becoming a rock star. Distracted by excitement, he lost focus on the guillotine press he'd been assigned — a machine that pounds sheet metal with crushing force — and the press came down on his fretting hand. When he yanked back, the tips of his middle and ring fingers were torn clean off. His doctor told him he would never play again.
A few days later, a foreman came by with a record. Tony refused to listen — the last thing he wanted was to hear music. But the foreman insisted, and when Tony listened, his mood lifted. The playing was extraordinary: fluid, virtuosic, alive. The guitarist was Django Reinhardt, and the foreman had one more thing to tell Tony: Reinhardt was playing with only two fingers on his fretting hand. He'd lost the use of his fingers in a fire.
Tony melted down plastic thimbles, attached them to his missing fingertips, and started relearning to play. He couldn't move as fast as before, so he leaned into slow, thick, heavy power chords. Nobody had heard that sound before. Tony Iommi became the lead guitarist of Black Sabbath. In the process of working through his limitations, he invented heavy metal.
This pattern, Ravi argues, is not an aberration. Freddie Mercury's four-octave range came from being born with extra teeth. Kanye West recorded "Through the Wire" rapping through a broken, wired jaw. Beethoven composed his most celebrated works while profoundly deaf — forced to feel vibrations rather than hear sound, he wrote music more powerful than anything he'd created before. Henry Matisse, bedridden after surgery and unable to stand at an easel, began cutting shapes from colored paper with scissors. Those shapes are now considered some of his greatest work.
The reason this pattern repeats has to do with how creativity actually functions. Creators are always navigating a tension between what is expected and what is exceptional — a constant push and pull between the average and the anomaly, not just within themselves, but in the eyes of their audiences. When everything trends toward the polished and perfect, something raw and authentic fills the gap. Instagram became impossibly beautiful, so Snapchat built for authenticity. YouTube became cinematic, so TikTok went raw and immediate. Streaming made music frictionless, so vinyl — imperfect, intentional, full of pops — came back to record-high sales.
AI is inside this same dynamic. The more capable it becomes, the more it floods the world with content that looks similar, sounds similar, and reads as slop.
"In a world flooded with pretty AI output," Ravi argues, "the work that breaks through comes from human weirdness, flaws, quirks, lived experience, and virtuosity. These are things that AI just can't average into existence. AI will always generate at the mean."
The practical implication for product builders: stop asking what AI can do in your product, and start asking what the human's role is in the process. The answer breaks down along two axes — craft (can the person actually create what they're imagining?) and taste (can they judge what's good?). Professionals have both and want AI to amplify them. New creatives have taste but lack craft and want AI to bridge the gap. Some users just need a job done and will delegate entirely. Apprentices have emerging craft but need help developing their eye.
Most AI products fail to be specific about which of these users they're building for and end up serving none of them well. Granola, the AI note-taking app, succeeded precisely because it got this right. When the founders looked at the market — Fireflies, Otter, Fathom, Zoom, Teams — they noticed that every existing product was solving for the same use case: delegation. You want someone else to take your notes. But there was a second category of user that nobody was serving: the professional note-taker. For PMs, researchers, anyone who lives and dies by their notes, the experience isn't about offloading — it's about amplification. "I want to take better notes, not have someone take them for me." That one insight — amplification, not autonomy — is what enabled Granola to break through in a crowded market.
Myth two: Software is obsolete
If AI can write its own code, generate its own tools, and spin up software on demand, why do we need SaaS? Why do we need products at all? This fear — which others have called the “SaaS-pocalypse” — has driven a multi-billion-dollar sell-off in Wall Street's valuation of software companies, with investors betting that AI will make the entire category redundant.
But zoom out to look at the actual arc of AI breakthroughs over the past few years, and a different picture emerges. The biggest gains haven't come from better models alone. They've come from what happens when AI gets access to tools.
Perplexity is the clearest early example. A few years ago, if you asked Claude what happened in the news yesterday, it would decline — that was past its training cutoff. Perplexity's insight was simple: what if the model could run a Google search when it needed current information, pull in those results, and then answer the question? The underlying model was OpenAI's. The innovation came from the tools the model could access. Better answers didn't come from a smarter model. They came from a richer set of hands.
Claude Code followed the same logic. When it launched, GitHub Copilot and Cursor already existed as coding tools, both capable of using Anthropic models. What Claude Code added was a command-line interface that gave it access to hundreds of other tools already on your machine: reading files, running packages, using Git, launching apps, and running shell commands. The revolution wasn't the model. It was the harness. OpenClaw extended this further, all of it a harness that allowed an AI to use shell commands, browsers, and APIs to act on the world.
MIT's motto is mens et manus — mind and hands. Intelligence without hands to act is inert. AI is the mind; software is the hands. They need each other.
Every AI product is really a combination of three things: model capabilities, data, and functionality (the harness of tools and integrations you plug the AI into). The right way to think about your product isn't that AI is making it obsolete — it's that your proprietary data and proprietary functionality are giving the AI superpowers that competitors can't access. The AI gets smarter; your product gets better; the moat deepens.
Granola again illustrates this. They didn't train their own models — they used Deepgram for transcription, and they used Anthropic and OpenAI for note generation and summarization. What they innovated on was how these pieces fit together. Notes you take during a meeting provide context to the AI, so when it combines those handwritten notes with the transcript, it knows what was important to you. That output becomes part of a growing database. Better notes lead to better summaries, which build a better knowledge asset for you and your team, which creates compounding switching costs over time. They also made a counterintuitive call to build a Mac app rather than a web product — because a desktop app can listen for system-level sound, detect when a meeting starts, and surface a prompt to take notes without requiring an agent to join the call. That's a software innovation, not a model innovation.
Software isn't obsolete. It's what gives AI the ability to act on the world.
Myth three: Product management is dead
The third myth is that if software can be generated on demand and everyone can be a full-stack builder, product managers are redundant — AI will write the specs, build the prototypes, and prioritize the backlog. Ravi's response: this is the least correct of the three myths.
To understand why, you have to understand why so many PM best practices exist in the first place. Most of them weren't designed to help teams build the best products. They were designed to help teams avoid expensive mistakes. The minimum viable product isn't something customers asked for — nobody wants the product that just barely meets their needs. MVPs exist because building software used to be expensive, and you needed to validate demand before investing further.
The entire product lifecycle was architected around scarcity. You climbed a ladder from low-cost artifacts (specs, which cost a few hours to write) to higher-cost ones (wireframes, paper prototypes, functional prototypes, MVP), investing a little more effort at each rung to earn a little more signal. Every gate existed to justify the cost of the next stage.
When AI collapses the cost of building working software, that ladder collapses too. It's no longer more work to create a functional prototype than to write a spec. In some cases you can have working software faster than you could write a two-page requirements document. And when that's true, the assembly-line model — PM hands to design, design hands to engineering, with weeks of lag between each stage — stops making sense.
What Ravi sees in the most AI-native teams is something closer to a jazz band. There's no linear handoff. Working software sits at the center. Research, design, engineering, and the people closest to customers all play their instruments and riff off each other. Sometimes a PM builds a prototype, sometimes an engineer does, and sometimes a designer does. If a prototype looks promising, someone writes a spec. If it surfaces more questions, you go back to research. The process is fluid, not sequential, and working software — something actual customers can react to — is the artifact of communication rather than a slide deck or a PRD.
Not everything is moving at AI speed, though. Ravi draws a useful distinction between the things that have accelerated — writing code, shipping software, analyzing data, writing PRDs, generating tickets — and the things that inherently move at human speed. Talking to your customer takes the same amount of time it always did. Customers have limited cognitive bandwidth for onboarding and activation; you can't rush that. Getting stakeholders aligned requires them to actually understand what you're building. Brand recognition requires repeated exposure. Establishing culture requires time. These things are all fundamentally human-paced.
And so in a world where building is cheap but humans are still humans, the PM's job changes in a specific way: judgment shifts from deciding what to build to deciding what deserves to ship. The primary role moves from prioritizer — gatekeeping what gets built at all — to curator: taking everything that's now possible, everything the team has built and tested, and applying the strategic judgment to determine what should actually reach customers and what should be cut.
"We can move from minimum viable product to minimum lovable product," Ravi argues. "We can finally create the products we always wanted to build." Judgment, taste, curation, leadership — these were always the heart of product management. AI makes some of the surrounding work easier, but it doesn't replace these things. It demands more of them. There's more to filter, more to decide, more to steer. "If you're going very slow, steering isn't important. If you accelerate to 60 or 100 miles an hour, steering becomes absolutely fundamental. Our organizations are going faster because of AI, so steering and leadership don’t get less important — they become essential."
The optimistic vision
The pessimistic vision of the AI era tells a coherent story: creativity doesn't matter because AI generates content better and faster than humans; software is obsolete because AI builds whatever it needs on the fly; product management is dead because AI does the planning, scoping, and shipping. Ravi takes this vision seriously and says it's not crazy to be concerned.
But look carefully at what's actually happening. Creativity is a self-correcting system, and the more AI floods the world with averaged output, the more it creates demand for things that are strange and human and specific. Software acts as the “hands” of AI — the only thing that lets intelligence act on the world — and the breakthroughs of the past three years have consistently come not from better models but from better software surrounding them. Some PM practices seem threatened but – on deeper reflection – these practices were mostly artifacts of scarcity. The key PM traits – judgment, taste, understanding of users, strategic clarity — are more valuable now than they have ever been.
The teams that will win in this era aren't the ones that ship the most. They're the ones that ship the best. And figuring out what "best" means, for which customers, at what moment, in a world where you can build almost anything — that's never been a more consequential question to get right.
Watch the full recording here
Find out more about Ravi's Reforge courses: AI Prototyping, AI Strategy, and Product Leadership