Will AI Replace Growth Marketers?
No, AI won't replace growth marketers. But it's already absorbing the parts of the job that were always execution, and that changes who's valuable. The growth marketers at risk are the ones who measure themselves by output. The ones who'll thrive own the judgment: deciding what's worth testing and why, then turning experiments into learning the whole company uses.
If you work on or around a growth team, you've probably seen the story by now.
For about 10 months, a big chunk of Anthropic's performance marketing ran on 1 non-technical person and Claude. Which is wild, because the company's valued at $380 billion.
It went mega-viral, and it split the growth community into two camps.
Camp one: this is a glimpse of the future, and AI is about to replace whole marketing and growth departments.
Camp two: cool story, but not a big deal. Early-stage teams have always run lean. This is just the next version of the tools we've always used.
After coaching 90+ growth leaders over the past 5 years (directors, VPs, heads of growth) and being a 2x head of growth myself, here's where I land.
The time and resources it takes to do world-class growth work dropped off a cliff. Stuff that used to take 30 minutes takes 30 seconds. Work that used to need an engineer doesn't. Work that used to need a designer doesn't. The data request that used to mean a ticket and a 2-week wait now happens in a couple of prompts.
When the cost of execution drops like that, the things that make you valuable change too. Some skills get commoditized. Others get a lot more valuable.
Which growth marketers are most at risk (and which will thrive)
Heads of growth ask me a version of this on almost every call right now: who on my team is actually safe, and who isn't? Here's the pattern I'm seeing.
Most at risk
- Measure their value by output: campaigns launched, pages tested, creatives shipped
- Kept AI at arm's length, waiting for someone else to validate the playbook
- Specialized in one channel (the Google Ads person, the lifecycle person)
- Walk into the room with a tactic instead of a business problem
- Run the team like a shipping factory and hoard the learnings
Pulling ahead
- Point to a few bets that mattered, and can tell you why they picked them
- Already in the lab building, getting reps while others wait
- Cover more surface area: design, data pull, code, and the read
- Start from the business problem, then work back to the tactic
- Run the team like a learning function and spread what they find
Those 5 contrasts map to 5 shifts happening right now. Here's each one.
Shift 1: Judgment over output
Growth marketers are obsessed with moving conversion rates. The way you move them is by trying a lot of stuff, because most experiments won't work. Maybe 1 out of every 3 is a winner, so you take a lot of shots.
So we prioritize output. The campaigns, the landing pages, the creatives, the CRO tweaks. Over time that volume becomes the thing we point to. It ends up in our self-evals: how many campaigns we launched, how many pages we tested. To push it harder, a lot of teams set goals around the number of experiments shipped.
Like most OKRs, people gamed it. They shipped a pile of tiny, low-stakes tests to hit the number. Activity went up. Business impact didn't.
I did this too. Scaling my first team, I didn't always know the highest-impact work, so I prioritized speed and volume. Button copy. Color changes. Headline swaps. I danced around the big, scary bets because I didn't want to spend the political capital to fight for the resources.
Here's what changed. Anyone getting savvy with these tools can ship hundreds of experiments a quarter now. Volume stopped being a differentiator.
The growth marketer who still points to volume as proof of their value is the one at risk. Their resume reads like everyone else's. The one who pulls ahead points to a small number of bets that actually mattered, and can tell you why they picked them and what it did for the business. The most valuable skill in growth right now is figuring out what's worth testing, and why. The question moves from "could we test this" to "should we test this."
Shift 2: Curiosity over caution
A lot of the growth marketers I talk to spent the last 18 months keeping AI at arm's length, for understandable reasons. The hype was way ahead of what the tools could do. We couldn't keep up with the pace of new tools and still hit the goals in front of us. And a lot of our community was nervous the rhetoric was right and AI was coming for our jobs.
That was me for a while too. The first wave of AI marketing tools wasn't impressive. Fast, but low quality. So I dabbled, but I wasn't sold.
Same with most of my clients. They had top-down goals to use more AI but weren't sure where to apply it. A lot of them just got to the same place a different way, without finding real leverage. That's changed. I'm watching some clients pull ahead of their peers with the same budget and the same constraints as everyone else. The difference is they're more curious.
One client, "Alex," head of growth at a B2C startup, told me AI used to feel like the boogeyman. Something he didn't fully understand and wanted to keep at a distance. Then he leaned in. Hooked up his MCPs, started playing with Claude Code, built his first couple of agents. Here's roughly what he told me on a recent call:
"Dude, I'm going crazy with AI. I'm having so much fun. Claude Code has been the ultimate cheat code. I've hooked all my disconnected tools and data into one hive mind. It feels like I have the world at my fingertips. It's even improved my mental health, because for a long time this felt like the boogeyman, and leaning in changed my outlook on the future."
A senior growth leader telling me his mental health at work got better because he stopped fighting AI and started building with it.
The growth marketer waiting for someone else to figure out the playbook is at risk today. Every day you wait is a day of reps you're not getting. The one pulling ahead is already in the lab, building, connecting things, learning by doing even though it's messy. Curiosity is becoming one of the most defensible skills you have.
If you're not sure where to start, pick 1 part of your week you do every time, and go build a version of it in Claude or whatever tool you like. Don't aim for polished. The skill you're building is thinking in workflows and prompts, and you only get it from reps.
Shift 3: Surface area over specialty
The real story here is how much one growth marketer can do on their own now. (This is basically the Anthropic story.)
Another client, "Francis," head of growth at a big consumer AI startup, lost his dedicated growth engineer to core product work. So he figured out if he could do it himself. He connected his MCPs to Snowflake and the company's codebase, all wired into Claude Code. Now when he wants to test something, Claude knows which data to pull, runs the analysis, calculates significance, and writes the code to ship the experiment live.
Last month he ran a test that drove a 17% lift in revenue from the test cohort. He shipped it solo. Claude did what used to be 3 people's jobs. And the part worth highlighting: the hypothesis, the design, the read on the data all came from Francis. Claude helped him ship it.
I see the same thing in my own work. A few years ago, a custom dashboard meant a ticket to the data analyst, a 2-week wait, and a couple of meetings about formatting. Now it's 3 minutes and a couple of prompts. I've got 10x more dashboards than I did 2 years ago, because I can do it myself.
The growth marketer who was valuable because they specialized in one thing is in a tougher spot now. The one getting more valuable absorbs the adjacent work, because the surface area one person can cover just expanded. One person owns the experiment design, the data pull, the code, and the read.
I'm even seeing a few directors and VPs with no direct reports who set the strategy and own most of the execution. Strategy is the real work. Execution is the part AI helps with. (I've watched a few people flip this, handing the strategy to the AI while they do the execution. In my experience that hasn't worked nearly as well, yet.)
Shift 4: Business problem first
The people most at risk walk into every room with a tactic instead of a business problem. Their starting point is "what should we ship?" or "I can do more with AI now, so what should we do?" The better starting point is "what does the business actually need here?"
When execution was the hard part, you could be a campaign-first thinker and still look productive, because everything took forever to build. The slow execution cycle covered for the missing strategy. Now anyone can ship in 20 minutes, so that cover is gone.
Back to Alex. His team executes like crazy, thinking in sprint cycles and feature releases, rarely stepping back to ask what the most important work even is and how it ties to the company's goals. That's the muscle to build.
The growth marketer who fully understands the business problem, then works backward into the projects, is worth a lot more than they were a year ago. The one who walks in with a tactic looking for a reason to ship it is now competing with a $20-a-month subscription. The value is knowing which campaign to run to move the business.
Shift 5: Learning over shipping
The first 4 shifts are about you as an individual. This one's about how you run the team. And it's what separates the growth leaders the rest of the org can't live without from the ones who could be replaced tomorrow.
AI changed how we ship, and in the process it exposed what a growth team actually produces. We care about wins and growth rate, obviously. But the way we get there is by learning what works and what doesn't. The learning is the real product.
Here's why that matters. A growth team usually works at optimization. The company already went 0 to 1. There's a working model and a few channels that convert. Your job is breaking through the conversion plateaus. Most of the time we don't actually know what'll work. We have hunches: segmentation, maybe, or the copy, or a friction point in the UX. We call them hypotheses, but a lot of the time we won't know until we try.
If you're at huge volume, you can spray and pray. The rest of us can't. So the job gets specific. What are we testing? Why? What do we expect to learn? And who else in the company benefits from that learning?
That last question is the one most teams skip. I skipped it. I was so locked onto my team's number that I treated everything else as someone else's problem. Product shipped things that contradicted what we'd learned in activation. Sales used copy on calls that we'd already proven didn't convert in paid ads. We left a pile of free wins on the table, because I ran the team like a shipping factory instead of a learning function.
Done well, it's a process with 4 steps:
1. Ideation
Build a backlog of strong ideas that could move the KPI, not whatever someone thought of in standup.
2. Prioritization
Use a framework to pick the right bet, not the loudest or most senior person's pick.
3. Experiment design
Write down what you hope to learn, how you'll know if it worked, when you'll move on if it doesn't, and who else should see the result.
4. Wins and losses tracker
Capture what worked, what didn't, and why, so it compounds over time and doesn't walk out the door when someone leaves.
We used to do all of this by hand. Now AI does most of it. The hard part is the discipline to actually follow the process once it's in place.
And once you learn something valuable, every team that touches the customer needs to know. Product needs to know what's working in activation. Sales needs to know which messaging converts. CS needs to know which signals point to expansion or churn risk. If you keep your learnings inside your team's four walls, you waste most of their value.
The growth leader who runs their team like an execution factory, head down, hoarding the learnings, gets stuck. They look productive. They are productive. But the rest of the org doesn't actually need them there. The growth leader who runs the team like a learning function is the one the org can't live without. The learnings, and how you spread them, are the product. Campaigns and wins are the byproduct.
The 5 shifts, in your pocket
One. Judgment over output. Volume was a proxy, and the proxy just got commoditized.
Two. Curiosity over caution. The leaders building right now will be running departments in 2 years.
Three. Surface area over specialty. Judgment is the work; execution is the part AI helps with.
Four. Business problem first. Walk in with a problem, not a tactic.
Five. Learning over shipping. Run your team like a learning function. The campaigns are the byproduct.
Watch the full breakdown
I broke this whole thing down on YouTube, where I post stuff like this every week for people who work on and around growth teams.
Frequently asked questions
Will AI replace marketers?
AI is replacing tasks, not people. It's taking over the execution layer across marketing: first drafts, reporting, basic analysis, shipping. The marketers at risk are the ones whose value was that execution. The ones who own strategy, judgment, and business outcomes are getting more valuable, not less.
Is growth marketing a dying career?
No. Demand for people who can actually find growth is going up. What's dying is the version of the job that's mostly manual execution. If your resume is a long list of stuff you shipped, that's the part AI commoditized. If you can point to a few bets that moved the business and explain why you picked them, you're in a better spot than ever.
What growth marketing skills are safe from AI?
Judgment about what's worth testing and why. Translating a business problem into the right experiment. Reading results in context. Spreading learnings so product, sales, and CS all get smarter. The strategic muscle is the safe part. The execution is the part AI helps with.
Can AI do growth marketing on its own?
Not yet, and not well. AI ships fast, but the hypothesis, the test design, and the read on the data still come from a human. The growth marketers winning right now use AI to cover way more surface area while they stay the one deciding what to build and why. The people who flip it, handing strategy to the AI and doing execution themselves, haven't gotten the same results.
How can growth marketers stay relevant as AI takes over execution?
Get curious and start building now. Pick 1 thing you do every week and rebuild it in Claude or whatever tool you like. Don't aim for polished, aim for reps. The skill you're after is thinking in workflows and prompts. Then practice starting from the business problem instead of the tactic.
If you're looking for more support, there are 2 ways I can help:
The Head of Growth Guide The full system I implement with my coaching clients. The playbook I wish I'd had when I was head of growth myself. 1:1 coaching If you're working through something specific and want a thinking partner who's been in the seat. I take on a small number of clients at a time.