Ten years ago, I was writing messaging briefs in Word documents and emailing them to team members for review. This week I wrote a bunch of Markdown files and trained my Claude instance to string 3 skills together.

Both within the range of being a PMM. Completely different world.

I’ve been noodling on this a lot in my real job … how much of what we do as PMMs has genuinely been reinvented by AI, and how much is exactly the same as it’s always been? Turns out, the answer to both is “more than you think.”

So here’s my take. Five things in PMM that are completely new … and five things that have not moved an inch.

The New Five

1. Markdown is now a job skill

Ten years ago, a PMM’s “toolkit” was PowerPoint, Word docs, and screen capture tools for demos. Today? If you’re not writing in Markdown, building context files, and understanding how to structure information for AI consumption, you are both leaving enormous productivity on the table, and you are missing what is actually happening today. The job is changing in real-time right in front of you.

2. The blank page problem is (mostly) solved

The first draft used to be the hardest part. The dreaded blank page glowing back at you. Now? You brief the machine, and you get a first draft in seconds. It won’t be perfect (it never is), but it’s something … and editing something is always faster than creating from nothing.

3. Agentic workflows are real, and they’re a PMM superpower

We used to talk about PMM as an orchestrator of people. Now we’re orchestrators of agents. Competitive monitoring, persona stress-testing, content transformation, battlecard drafts … these are no longer things you wait on. They run while you sleep. The “execution tax” of PMM is finally starting to get alleviated, freeing us up to spend time on more valuable activities.

4. Research that took days now takes minutes

Win/loss analysis, competitive landscape sweeps, ICP persona synthesis from raw interview transcripts, reading analyst reports … I used to block whole days for this. Now it’s a well-structured prompt and a strong cup of coffee (side note: definitely still need the coffee). The PMM who learns to brief AI well becomes a one-person research engine.

5. One asset. Many outputs. Zero excuses

Ten years ago, you wrote a whitepaper and it became … a whitepaper. Today that same content gets transformed into a variations for each ICP, a blog, a LinkedIn series, a sales email, a battlecard, and a demo script before lunch. Content leverage has gone from a nice-to-have to a standard operating expectation. If you’re still doing one-to-one, you’re already behind.

The Unchanged Five

1. The Story is still everything

AI can generate content at industrial scale. It cannot craft a narrative that makes someone feel something. The market narrative, the “why now, why us, why you” … that is still a deeply human activity. A vibe engine can simulate the shape of a story. It can’t understand why one word lands and another doesn’t. It’s never stood in front of an audience and seen how words make people nod, or frown, or smile.

2. Presenting to a room full of people

The stage hasn’t changed. Standing up in front of a customer, a room full of people at an event, an EBC … that still requires a live human who can read the energy in the room, adjust in real time, and make people believe something. No agent is doing that for you, no AI generated Speilburg level generated video will convey that. Practice your content. Know your collateral to the point of obsession. Pivot in the moment.

3. Stakeholder navigation

PMM is the “connective tissue” between Product, Sales, and Marketing. It was true ten years ago and it is painfully true today. The politics, the competing priorities, the art of getting three different leaders to agree on a single narrative … that is entirely human work. High EQ is not a feature you can prompt for. Data helps, humans close.

4. Real-world context and instinct

AI knows what is on the internet. It does not know what your biggest customer said in that candid moment after the QBR, or what the sales rep has been hearing in the field but not logging anywhere, or why a particular message lands differently in EMEA versus North America. That context lives in humans, not models. It always will. Back channels are not systems of record.

5. The final sign-off

Ultimately, the PMM remains the last line of defense. The human who looks at the AI’s output and says “yes, that’s true and emotionally resonant” or “nope, that’s technically accurate but it will land wrong.” That judgment … the ability to feel whether something is right … is the one thing that cannot be delegated. Not yet. And maybe not ever … maybe.

The PMMs who thrive in the next decade won’t be the ones who resisted AI, or the ones who handed everything over to it. They’ll be the ones who figured out exactly where each belongs.

AI for the 80. Humans for the 20 that actually matters.

Adam