I've been playing around with Claude a lot lately — partly trying to make my life as a marketer easier by 'automating the low-value work', as all AI tools now promise; partly hoping to test the limits of the most trendy and mighty AI that has caused near-apocalyptic panic about humans being replaced by 2028.

No doubt it has genuinely elevated what's possible. It feels like your fresh-out-of-school intern suddenly gained years of experience and the proactiveness to match overnight. But here's what I didn't expect: for every project, I wasn't spending less time working. I was just spending it differently — not on outputs, but on foundations.

And this made me see my role as a PMM with a new lens.

New technology, same old challenges

For AI to do anything useful, I had to answer the same questions every time:

  • What's the product?
  • What does it actually solve?
  • Who is the ICP?
  • What are the real differentiators?

The difference in output quality between sharp, consistent answers and vague ones is not marginal. It's the difference between "we elevate how companies integrate AI" and a headline that actually sounds like you and makes someone stop scrolling.

If you're a PMM reading this, you might wonder: aren't these typical PMM tasks — what we're paid to do? Precisely. My biggest takeaway is that the 'garbage in, garbage out' rule applies to AI strategy in marketing too. All that speed and scale is only as good as the inputs you give it. And for GTM, those inputs are nothing but positioning - the oldest PMM task there is.

Positioning is the decision about where you play and how you win. Without it, AI will hallucinate. More powerful AI will just hallucinate faster, at scale, straight into your customer's inbox.

So, for companies that want to automate outreach sequences, generate content at scale, or qualify leads autonomously: "what can agents do?" is the wrong starting point. The right question is "do my agents have the right foundations to work from?" And that makes PMMs' work more strategically important than ever.

Getting the story right is a team sport - of AI and humans

It doesn't mean you can't leverage AI until the foundations are perfectly done. It's a collaboration, not a handoff. But your team needs to take on the thinking and judgment AI genuinely can't.

Humans make the calls that matter. Which customer problem are you actually prioritising? Which market are you choosing to win? Which differentiators are real versus wishful thinking? These aren't questions you can prompt your way out of. They need people who've done this multiple times, followed enough sales calls to know when a room is losing interest, and why. Trust PMMs' experiences, and get your product-marketing team aligned on this first — it's adding extra time, occasionally uncomfortable, but completely non-negotiable.

Then bring AI in as your sharpest thinking partner. Feed it your positioning thesis and ask it to steelman the counterargument. Give it 50 customer interviews and ask it to cosplay a prospect to find underaddressed pains. Point it at your top three competitors' messaging and ask where your story dangerously overlaps. AI is genuinely good at surfacing what you're too close to see — but you have to know what you're asking it to look for.

With solid positioning, the rest follows naturally.

Before spinning up your next agent workflow, answer these three honestly:

  • Can everyone on your GTM team describe your positioning the same way without looking it up?
  • Does the content your agents produce explain why you're different — or just what you do?
  • Is your ICP specific enough to draw a line? If the answer to "who is this for" covers basically any company that does X, that's a category, not a customer profile.

Most teams know which of these they can't answer cleanly - and why. Fix those first.

Let PMMs do what they were always supposed to do

I started vigorously using AI partly because of all the noise about which roles it would eliminate — call it 'know your enemy' research. My verdict: product marketing is not one of them, if PMMs are actually doing what this function is for.

AI raises the stakes on getting the story right. The more powerful the automation you want to build on top, the more solid that base needs to be — you don't build a skyscraper on a cottage foundation.

With the heavy-lifting work — research, analysis, content execution — handed off to AI, PMMs finally have the room to do what actually moves the needle. Companies that figure this out first won't just have a better story — they'll have a real competitive edge.

Get the foundations right first. Then let the machines run.