Why It Matters
AI can make a small team far more productive, and it can also quietly make you a worse marketer. Ben Feys of PrettyBoy shares what worked and what didn’t. Automating the email workflow made the team more productive, cut agency costs, and let them send far more email. Automating big chunks of creative led to accepting “fine” ads, and Ben called that watching his brain atrophy. His rule is to use AI for tasks you already know how to do, so you can spot a wrong formula or a weak ad. PrettyBoy now tracks fully AI-made ads against human-made ones to see which perform. Will Sartorius learned the same lesson the hard way, when “Claude native” hires produced thoughtless work. With AI creative starting to look the same in the feed, the human work of strategy, research, and judgment may be the real moat.
Who This Is For
Built for Founders members who are small-team operators, agency owners, and creative leads using AI daily and wondering where it’s helping versus hurting. Especially useful if you’re building a company brain, automating creative production, hiring juniors into AI-heavy workflows, or sensing that your output is getting faster but samier.
This blog is pulled from an AI DTC WTF podcast episode with Ben Feys of PrettyBoy and Will Sartorius.
Ben Feys runs PrettyBoy with his co-founder, Kevin. It's a two-person team, they use AI constantly, and Ben is a Foxwell Founders member who has been sharing a lot of what they've built. He's also one of the most clear-eyed people I've talked to about what AI costs you when you lean on it too hard.
That's what this episode ended up being about. AI can make a small team a lot more productive. It can also make you worse at the job.
Start with Work You Can Measure
Ben's first real build was PrettyBoy's email workflow. He walked Claude through the whole process: analyze the account like a top retention marketer, plan the content, check the claims (they sell skincare, so no wild promises), write the briefs, and load them into ClickUp. Then he asked how much of it could be automated. Claude said all of it, right up to loading the campaign into Klaviyo and hitting send.
With two people, Ben has a simple rule: anything they build has to make them more productive, beat the alternative, or save money. The email build did all of that. They rescoped their agency so its strategist spends time on ideas and account management instead of writing briefs. The contract got cheaper, and they now send far more email.
Will pointed out how easy it is to skip this step. You end up in what we keep calling AI psychosis, building for the sake of building, feeling productive because you have 20 Claude chats open. So he asked Ben what hasn't been worth it.
Where It Makes You Worse
Ben had two answers.
The first is the idea that you'll never pay for software again because you can vibe code it yourself. (The average marketer is not rebuilding Triple Whale or Northbeam.) You'll jerry-rig something that feels productive, then spend months patching it when you could have just paid for the tool and gone back to selling.
The second one is sneakier: automating big chunks of the creative process. You get a pile of output and start accepting ads that are fine. You know it isn't the best ad you could have made for that concept, but it looks nice and volume matters, so in it goes. Ben called this watching your brain atrophy in real time, which is a sentence I'd like to forget and can't.
It's fun to throw a prompt into Manus and get 40 ads back in 10 minutes. The question is whether you're critiquing those 40 as hard as you'd critique your own. Usually you're not. A good designer working from a well-researched brief, with an AI pass or two, often beats it.
So PrettyBoy now tracks ads made entirely through the AI workflow separately from ads made by humans, to see which perform. I love this. It's the opposite of vibes.
Skip the noise. Get the Download.
Performance marketing insights that are real, useful, and worth your time.
Use the Company Brain as a Memory, Not a Replacement
I asked Ben about company brains, since everyone is building one. My take is that a brain lets you see much more of the business at once. Our data lives in Klaviyo, Northbeam, Motion and a dozen other places, and the brain pulls it together.
Ben agrees, with a caveat. He thinks brains are a net positive mostly for two groups: tiny teams like his trying to compete with much bigger companies, and senior leaders who already have deep knowledge and want to hand it down to their teams.
The good uses at PrettyBoy are practical. When they're making an ad, the brain pulls the best customer review, word for word. Ben can ask which messaging angles for their top LTV persona are underrepresented in ad spend, and it finds the gaps across their retention data and ad tagging. He doesn't need to become a data analyst to use the data.
The bad use looks like: "Write me 20 briefs this week. Five iterations, five net new, five videos." You're overwhelmed and you'd rather ship volume than fall behind. Everything starts looking the same, and the muscle memory slowly goes.
That's the line for me too. I use it as a memory bank. What did I tell that member last month, and how did I phrase it? That's a great use. Letting it replace my judgment isn’t.
Know how to do it before you hand it to AI
Matthew Gattozzi recently wrote a newsletter arguing that AI is making us worse marketers because we've stopped thinking things through. I agree. Data is only as good as the plan you point it at. The human part of the job is the strategy: why we're doing this, how we'll judge it and what the data needs to tell us. Skip that and you get a lot of output and a lot of overwhelm. (His other line, "nobody wants your PDFs," got a big laugh from Ben, who admitted he once proudly sent Kevin a seven-page Claude PDF before realizing most of it was filler.)
Ben's rule is that AI works best when you already know how to do the task and just want it done faster. He's had AI build complex financial and marketing models. If you can't tell it what inputs to use, check its Excel formulas and spot check a few numbers yourself, you'll end up acting on wrong information.
He knows what the answer should look like, so AI can handle the math. That's also why he thinks people who learned this the hard way get the most out of AI – they already know what good looks like.
Will Learned This the Expensive Way
Will, our resident former investment banker, had a confession here. He has a very analytical formula for creative: 60% iterations, 20% reskins and 20% net new personas, built on tagging ads by persona, angle and emotion and mining Reddit for gaps. He turned it into a workflow and figured he could hire junior strategists to run it.
He even put "Claude native" in a job posting. The applicants were very good at Claude. The work was, in his word, atrocious, because it wasn't thoughtful. Now new hires learn creative strategy 101 first, and nobody touches Claude until they can write a brief on their own.
His bigger point is that the market is full of people who think AI makes them a lot more powerful than it does. A bit harsh, maybe, but I think a lot of us need to hear it.
The Skills That Still Matter
So what should a marketer actually get good at? Ben's list was refreshingly old school.
Be someone who wants to understand how things work, not someone who wants a faster result. Ben's habit is to build something, then step back and ask whether he understands what he did.
Use AI as a teacher. Ask it what the great advertisers did, how Ogilvy and the long-form copy legends sold on emotion, and use it as a sounding board while you write.
Get serious about research. Anyone can open the ad library. Can you explain exactly why a winning ad works? If you iterate on an ad without knowing why it won, you're basically retesting from scratch every time.
Give AI your best human work to learn from. Ben fed Claude a 115-page PDF of every email they wrote before AI, and that's what made the output sound like them. Kevin has made close to a thousand ads over four years, winners and losers. AI helps him because he already has a process. Someone typing "write me 10 video briefs" with no background won't get the same results.
I'd add that there's never been a better time to write long-form creative by hand. Use AI to learn from your history, then trust your own sense of what will land.
The Human Work Might Be the Moat
I asked Ben what running a business looks like a year from now. His answer is that it's never been easier to start something, and it's going to be harder to do it well enough to last.
His bet is that the brands that stand out will deliberately resist making everything easier. Big, well-funded brands can afford to flood the feed with AI content. A smaller brand copying that playbook without the budget is going to have a rough time. PrettyBoy has AI ads that have been top spenders for six months, so Ben isn't anti-AI. But when he scrolls, the AI ads all start to feel the same. When everyone zags, it might be time to zig, and for them that could mean going all in on human-made content and copy.
In other words, stick to the plan. AI makes it really easy to get pulled in 10 directions. Before you chase the next thing, ask whether it helps you hit the goal you already set.
Ben put it best at the end: lean into AI as hard as you can, because it isn't going away. Don't resist the calculator. But know what it's good for, keep getting better at the hard parts yourself, and be just as excited about those.
Watch the full episode here:
Note: This blog was written by a human, but AI helped draft the outline and synthesize ideas.
