Why PMs Who Build Win

The Divide Is Already Here

Look at who's hiring AI PMs right now. OpenAI, Anthropic, Google, Meta. Check their job listings. "Hands-on experience building AI products" isn't a nice-to-have. It's requirement #1.

Meanwhile, companies are cutting traditional PMs who excel at stakeholder management but can't differentiate between prompting and fine-tuning. They're not bad PMs. They're PMs for a job that's shrinking.

The math is simple:

• AI PM roles: $300K-$500K average, growing 40% YoY
• Traditional PM roles: $180K-$280K average, flat or declining

Same title. Different job. Different pay. The gap isn't closing — it's widening.

---

What "Building" Actually Means

Let me be specific, because "build" gets thrown around loosely.

Building means:

1. You can prototype. Before any engineer writes a line of code, you've already tested the AI capability. You know what works, what doesn't, and where the sharp edges are. You're not speccing something you haven't touched.

2. You can eval. You know how to tell if an AI feature is good. Not vibes — actual evaluation methodology. Test sets. Rubrics. Metrics that make sense for non-deterministic outputs. If you can't eval, you can't ship.

3. You can talk the talk. When your ML engineer mentions context windows or model drift or prompt injection, you don't nod along blankly. You have opinions. You push back. You collaborate, not just coordinate.

4. You can ship incrementally. AI features don't waterfall. You ship small, measure fast, iterate constantly. The PM who needs a 20-page PRD before touching anything? They're not building. They're blocking.

Building doesn't mean you're an engineer. It means you're not helpless without one.

---

The "Strategy vs Execution" False Choice

Someone's going to object: "But PMs should focus on strategy! Let engineers execute!"

This is a comforting lie that lets you stay in your comfort zone.

Here's reality: the best AI product strategies come from people who have actually built AI products. You cannot effectively strategize about what AI should do if you've never touched what AI actually does.

Ever had a PM confidently promise a feature that was technically impossible? That's the non-building PM trying to strategy their way through AI.

The best AI PMs I know operate in a loop:

• Build something → Learn what's possible
• Learn what's possible → Form better strategy
• Form better strategy → Build something better

You cannot skip the building step and expect the strategy to be grounded in reality.

---

"But I Don't Have Time to Learn Technical Stuff"

Then you don't have time to be an AI PM.

Sorry. I know that sounds harsh. But companies are hiring AI PMs specifically because they need someone who bridges business and technical. If you can only do one side, you're half the value.

The good news? The technical stuff you need to learn is actually learnable. We're not talking about getting a PhD in machine learning. We're talking about:

• Spending a weekend getting good at prompting
• Understanding what evals are and why they matter
• Learning the basics of how LLMs work (conceptually)
• Knowing the tradeoffs between different models
• Being able to prototype features in tools like v0, Replit, or Claude

This is weeks of effort, not years. And it's the highest-ROI investment you can make in your career right now.

---

The Building PM Advantage

When you build, you get superpowers:

Speed. While other PMs wait for eng time to test an idea, you've already validated three variations and know which one works.

Credibility. When you can demonstrate understanding — not just articulate it — engineers trust you. That trust translates to better collaboration, faster execution, and less "PM vs Eng" dysfunction.

Better Specs. When you've built the prototype, your PRDs aren't theoretical. They're grounded in what the AI can actually do. Fewer "this seemed like a good idea in the spec" moments.

Career Leverage. You're not dependent on any single company. You have skills that transfer. If your company pivots or lays off, you can build something yourself, join a startup, or move to a company that values builders.

The non-building PM is at the mercy of their org chart. The building PM has options.

---

The Counter-Argument I'll Concede

Yes, there are successful PMs who don't build. They exist in orgs where PM is a coordination function, where eng capacity is abundant, where AI is a small part of a larger product.

If that's your situation and you're happy, more power to you.

But if you want to:

• Work on AI-first products
• Command AI PM compensation
• Have leverage in your career
• Not be anxious every time layoffs are announced

Then you need to build.

---

How to Start Building (Today)

Not next quarter. Today.

Step 1: Open Claude or ChatGPT and start using it for work. Not just asking questions — actually building things. Write prompts. Create prototypes. Make it your daily tool.

Step 2: Build an eval for something you're working on. Even if it's informal. How do you know if your AI feature is getting better or worse? Figure that out.

Step 3: Read the ML org's docs. Not the marketing pages — the actual technical documentation. Understand how their models work. What the limits are. What they recommend.

Step 4: Prototype your next feature before writing the PRD. Even if it's rough. Even if no one asked you to. Do it anyway.

Step 5: Talk to your ML engineers about what they wish PMs understood. Then understand it.

These aren't huge asks. They're habits. The PMs who build start small and compound over time.

---

The Manifesto

So here's what PM the Builder stands for:

We build things. Not because we have to, but because building makes us better. Better strategists. Better collaborators. Better PMs.

We ship. Strategy without shipping is fantasy. We'd rather ship something imperfect than strategize something perfect that never exists.

We stay technical. Not to become engineers, but to partner with them effectively. We speak the language. We understand the constraints. We add value, not just process.

We're honest about uncertainty. AI is probabilistic. We don't pretend otherwise. We communicate uncertainty clearly and build products that handle it gracefully.

We prepare for the interview. The AI PM interview is different. We do the work to pass it, not hope our traditional PM experience carries us.

We help each other. This stuff is new. Nobody has all the answers. We share what we learn.

This newsletter exists to make that easier. Every issue: practical tactics for PMs building AI products.

No fluff. No frameworks-for-frameworks. Just stuff that helps you build and ship.

---

Now go build something.

---

Key Takeaways

1. The AI PM job is different — and it rewards PMs who build, not just PMs who strategize
2. Building is learnable — you don't need a CS degree, just willingness to get hands-on
3. Start today — prototype, eval, learn the tech; the gap is widening and waiting won't help