Most advice about growth marketing career AI skills has it backwards. People are racing to learn the tools — which prompt, which model, which plugin — while ignoring the things that actually compound. The tools change every quarter, and over time I've come to believe that what you build on top of them is what lasts.
I've spent most of my career scaling marketing functions through real inflection points, including taking a company's growth through a period of rapid expansion and into the public markets. The marketers who got more valuable through that stretch were never the ones who knew the most features. They were the ones who knew what to point those features at.
The growth marketing career AI skills that compound
AI collapsed the cost of execution. Writing the email, building the variant, pulling the query, drafting the brief — all of it is cheap and fast now. So execution stopped being the bottleneck, and everything upstream of it became the job. The most durable way I've found to think about this is that growth is a system, not a function — a connected machine spanning acquisition, lifecycle, and retention rather than a single team or channel. That framing rewards the T-shaped operator who can go deep in one area while reasoning across the whole system, and it points at a specific set of skills, the ones I'd bet a career on.
- Judgment. Knowing which experiment is worth running, which channel deserves budget, which result is real and which is noise. AI will happily generate a hundred mediocre ideas; choosing the one that matters is still on you.
- Systems thinking. Seeing the whole funnel as a connected machine — acquisition feeding lifecycle feeding retention — rather than a pile of disconnected tactics. AI optimizes locally, which means you have to optimize the system.
- Directing AI, not being replaced by it. Treating models like a team of fast, literal, tireless junior operators. You set the goal, the constraints, and the definition of done, and most of the leverage lives in that direction.
- Data literacy. Reading a result and knowing whether to trust it — significance, sample size, attribution, confounders. AI will hand you a number with total confidence, so you need to know when it's wrong.
- Writing and clear thinking. They're really the same skill. If you can't write the problem clearly, you can't brief a person, an AI, or yourself.
None of these are new, and that's exactly the point. AI didn't invent a new skill tree. It raised the price of the skills that were always scarce and crashed the price of the ones that weren't.
Prompting is not a skill. Problem framing is.
There's a whole cottage industry around "prompt engineering," and I think most of it is noise. Prompting is a syntax, and syntax gets easier with every model release — the models are actively being trained to need less of it.
What doesn't get easier is framing the problem: knowing what you're actually trying to solve, what good looks like, which constraints are non-negotiable, and what tradeoffs you'll accept. That's the hard part, and it's the part AI can't do for you, because it's the part that requires knowing the business.
This is also why I'm skeptical of the hype around AI marketing. The bottleneck was never the writing or the building. It was deciding what to build and knowing whether it worked — and that's exactly the part that gets harder, not easier, once execution is free.
What juniors should stop optimizing for
If you're early in your career, this is where I'll be direct. Some of the things that used to make a junior marketer useful are now worth close to nothing, and clinging to them is a trap.
- Raw production speed. Being the person who could crank out 40 ad variants overnight was a real edge once. It isn't anymore, because the machine does that for free.
- Tool memorization. Knowing every menu in every platform. Interfaces are dissolving into natural language, so learn what the tool is *for*, not where the buttons are.
- Polish without a point. A beautiful deck that doesn't change a decision. AI makes things look finished instantly, which means "looks done" is no longer evidence that anyone actually thought.
- Waiting to be told the exact steps. The marketers who pull ahead are the ones who can take an ambiguous goal and run with it. Following instructions is the one thing AI does better than you.
The time you reclaim is better spent on the things that compound. Learn how the business actually makes money, sit with the data until you can argue with it, and build the muscle for running experiments that hold up instead of ones that only look impressive in a readout.
The marketer who can frame a problem, direct an AI to attack it, and read the result honestly is worth far more than one who can only execute fast — and that gap is widening, not closing.
I write more about how I think about growth, experimentation, and building AI-native teams on my site, Sebastian Behar. If you want a wider feed of how the growth and PM roles are evolving, Lenny's Newsletter is worth following. My broader view is that the era of cheap execution doesn't make growth marketers less valuable. It makes the *right* growth marketers far more valuable, and it quietly retires the ones who confused activity with impact. Worth deciding early which of those you intend to be.
