Mike Pilawski · SVP of Product, Miro
A role declared dead 5 times — and promoted every time.
Product management in software was improvised. There was no PM community. No PM bootcamp. No “empowered product team.” You reported to marketing, engineering or strategy, depending on who won the last argument. You wrote 200-page PRDs. You wireframed in PowerPoint. You were the product.
The role traces to a single 800-word memo written in 1931 by Neil McElroy at Procter & Gamble. He proposed a new role where one person takes absolute responsibility for one product, as a “business within a business.” That idea took 70 years to reach software.
In 2004, Nokia offered me a PM role. I had a finance degree, an engineering scholarship in Finland, and no idea what “product manager” meant.
My first project: putting internet radio on feature phones. Within months we had 40 million active users. No UX team. Our UI designer just pivoted his career from nuclear engineering. There was no one to support, no playbook. You were on your own.
Early PMs needed to be strong across three domains simultaneously. There was no separate UX function. No Growth team. No Product Ops. You did commercial strategy, technical decisions, and user experience design. All at once!
Then two earthquakes hit: Agile rewired PM from “writer of requirements” to “voice of the customer in delivery.” Cloud economics (AWS S3/EC2, 2006) collapsed shipping costs and made continuous delivery possible. The iPhone (2007) elevated UX from afterthought to obsession.
The PM was all three — overlapping, undefined, improvised.
The PM role was born from necessity, not design. When nobody else exists to do the work, the PM does everything.
As software complexity exploded, the full-stack generalist model became untenable. Functions the PM historically owned got peeled off into specialist roles.
Melissa Perri heard “UX Designer” for the first time when her company hired a Director of UX to do the work she’d been doing as PM.
Build-Measure-Learn. MVPs. Hypothesis-driven development. PMs become scientists, not scribes.
Dedicated Program Managers take the timelines. The PM triad (PM + Design + Eng) becomes standard.
Mind the Product (2011). Product School (2014). Cagan’s Inspired. PM becomes a real career path.
As complexity grew, PM shed responsibilities — but lost ownership in the process. The “what and why” got cleaner; the holistic view got harder.
The triad separates. Each gets its own leadership.
Each specialization emerged because a new constraint became the binding bottleneck in scaling software.
App Store rules, device fragmentation, touch UX. The first specialization wave.
Funnels, A/B tests, retention. Owns metrics, not features. Born at Facebook, spread everywhere.
APIs as product. Developer experience. Twilio, Stripe, Salesforce. Highest salary premium today.
Data as product. Internal stakeholders as customers. BI platforms, quality, governance.
The operating system for PM orgs. Process, tools, intake, enablement. 75% of companies now use it.
Privacy-by-design became roadmap work, not a legal checklist. Every PM’s scope expanded permanently.
Specialization tracks the dominant bottleneck. When you see a new constraint emerge, the next PM specialization — and salary premium — is forming.
Every specialization wave follows the same cycle. Early movers capture outsized returns. Then the window closes.
US avg total pay, Glassdoor early 2026. Small samples for AI/Platform/Mobile — directional, not definitive.
The cycle: Mobile was premium in 2012, table stakes by 2018. Growth commanded premiums in 2016, expected by 2022. AI skills are premium now, but will normalize by ~2028.
“Mobile PM in 2010. Growth PM in 2015. AI PM in 2025. The window won’t stay open forever.” — Aakash Gupta
Surprise: Platform PM commands the largest premium — 51% over baseline. Systems-level leverage across multiple teams is the most valuable scarcity, not just hype-cycle proximity.
Every technology wave declares PM dead. Every technology wave makes PM more valuable.
This is why AI PM isn’t just “harder”, it’s structurally different. You’re no longer specifying features. You’re architecting behavior.
| Old Toolkit | New Toolkit | What Changed |
|---|---|---|
| PRDs with exact specs | AI-native PRDs with model criteria | What it does → how it chooses |
| Wireframes | Prompt engineering | Fixed layouts → boundaries of variability |
| User stories (happy path) | Behavior guardrails | Expected paths → unacceptable outcomes |
| A/B testing UI | Evals (LLM judges) | Conversion → truthfulness & drift |
| Fixed roadmaps | Outcome-based prototypes | Quarters → hours |
If your PM toolkit still assumes deterministic outputs, you’re managing a world that no longer exists. Start learning evals and guardrail design.
Andrew Ng’s team proposed a 2:1 PM-to-engineer ratio. The inverse of the traditional 1:8.
“Things that used to take six engineers three months, my friends and I will just build on a weekend.” — Andrew Ng, No Priors podcast, August 2025
90% of engineering teams now use AI coding tools. 50% increase in code shipped. Cursor hit $500M+ ARR by mid-2025.
Ng’s metaphor: the typewriter made writing easy but created writer’s block. AI coding solved the “how” leading to “builder’s block.” The bottleneck is now: what should we build?
The ATM Effect: ATMs didn’t reduce bank employees. They made branches cheaper, so banks opened more. Cheaper development may make more products viable, expanding PM scope.
The scarce resource in software is no longer engineering capacity. It’s knowing what to build and why. That’s a PM skill.
“Vibe coding” was Collins Dictionary’s Word of the Year 2025. PMs are building functional prototypes in hours, not requesting engineering sprints.
“It was only when I showed my working prototype that I suddenly got everyone onboard.” — Staff PM at Duolingo, who built it over a weekend with Cursor
Claire Vo built ChatPRD — now used by 100,000+ PMs — single-handedly over Thanksgiving weekend. Google, Stripe, and Netflix are introducing AI-prototyping rounds in PM interviews. Carnegie Mellon now requires vibe-coded prototypes.
Cagan’s “Product Creator” framework: anyone shaping the product across the four big risks (value, usability, feasibility, viability) is a creator. Process-oriented PMs face existential pressure. Builder PMs are more valuable than ever.
In 2004, I shipped internet radio to 40M users on feature phones. I was tasked with UX, strategy, and distribution because nobody else existed to do it.
Over the last 10 weeks, I have built seven products alone and with friends. I am doing again UX, strategy, tech and marketing because now one person can do it again. Twenty years apart. Same shape. Infinitely more leverage.
2026: The circles reconverge — but now with 100× the leverage.
Designs workflows where multiple AI agents collaborate — researcher, coder, analyst. Masters sequential, parallel, and supervisor patterns and invents your own.
Owns the full spectrum from product to market. Blends product, product marketing, marketing, and revenue operations into one role.
As agents move from analysis to execution, governance becomes an operating model.
The paradox: PM hiring is down 14%, but PM compensation outpaces most functions. Companies are retaining experienced PMs while pulling back sharply on external junior hiring.
AI automates precisely the tasks junior PMs used to do to learn the craft: writing specs, organizing feedback, basic analysis, user story creation. The “broken rung”: we’re removing the ladder rungs that train junior PMs into senior PMs.
As one engineering leader warned, the talent pipeline crisis will become “devastatingly clear in 24–36 months.”
The implication for leaders: if you’re building a PM org, invest in apprenticeship models now. The people you don’t develop today are the senior PMs you can’t hire in 2030.
The industry is eating its own seed corn. If you lead PMs, create pathways. If you’re junior, build. Your portfolio of shipped AI products is your new resume.
What survives every cycle: curiosity, the ability to learn fast, and the capacity to zoom in on detail and zoom out to strategy — the same skill at every inflection point.
Nokia → Vungle → Typeform → Smallpdf → Lokalise → Miro → Building with AI.
Each company was a different inflection point. Monetization architecture. PMF-to-scale. M&A integration. AI pivot. Multi-product portfolio. The constant across all of them: curiosity, learning fast, and the ability to zoom in and out. You need to be able to see the big picture and go down to addressing the details. That’s the skill AI can’t replace. It can only amplify it.
“Product creator” outlasts every cycle. The title changes. The capability endures.
Cagan, Doshi, Rachitsky, and Torres all converge on the same conclusion: judgment and product sense are the irreducible core. AI amplifies what you bring to the table. If you bring process, AI replaces you. If you bring vision, AI supercharges you.
Pick a user problem. Prototype with Miro Prototypes or Lovable and build with Claude Code or Codex. Get 20 users. This is the highest-ROI PM activity right now.
Evals are to AI PM what A/B testing was to Growth PM — the defining competency. Start with Aman Khan’s guide on Lenny’s Newsletter.
Every AI feature must create value ≥ 3× its compute cost. Understand the P&L of tokens and infrastructure.
Shreyas Doshi: “THE main skill in the AI age.” The 10-30-50 framework: top 10% in one sense, top 30% in another, top 50% in a third.
Reforge: “Flip the script. Focus on what AI enables for you that you couldn’t do before.” That’s the zoom lens in action.
Map your product as data flows and agent interactions, not screens. The stochastic mindset is the new table stakes.
The 6th time is yours to write.
Mike Pilawski
mikepilawski.com/talks/the-full-circle
Builder