How to Thrive as a New Team Member Using AI

It was 2 AM in Los Angeles. Leo Pohlmann and I had spent the entire day at Six Flags Magic Mountain, rode twenty-some rides until we nearly threw up, left the park at 10 PM, and then talked for three hours. Our Uber to the airport was seven hours out. Instead of sleeping, we hit record.

What came out is the most honest version of a conversation most teams never have out loud: a 19-year-old walking through every mistake he made in his first few months, and me explaining why those mistakes are the predictable ones.

AI is only as smart as you are

Leo’s first real failure was subtle, because on the surface it looked like productivity.

He was given a video and asked to cut it and write an article. So he handed it to Claude with roughly one sentence of instruction: cut this in Descript and write a good article for it.

The video came out fine. Video editing tolerates thin instructions. The article was bad, and it stayed bad even after a few rounds of improvement.

The reason wasn’t the model. Leo hadn’t watched the training that explained the SEO Tree or the frameworks the article was supposed to follow. He couldn’t give Claude context he didn’t have. AI is only as smart as you are.

This compounds. When I later gave him a harder task, building an audit page for different companies, he failed at it, got my feedback, and couldn’t understand the feedback either. So he passed that to Claude too. There’s a ceiling here, and it isn’t the tool’s.

In Germany there’s a saying Leo brought up: a good horse only jumps as high as it must. Most people coming out of school have been trained to clear exactly the required height. That habit is fatal in a field where nobody is measuring the bar for you.

The two ingredients I look for

I want to reframe this, because it isn’t a confession of failure. It’s a positive example. Months earlier, Leo had already shown me the only two things that reliably predict whether someone makes it.

The first is putting in the time. Not saying you will. Actually doing it, including when it’s inconvenient, including at 2 AM after a theme park. Most of you won’t.

The second is being coachable. Willing to listen, willing to learn, willing to accept critical feedback without treating it as an attack.

When those two are there, I know that learning new tools or new frameworks or adapting to whatever ships next year will not be a problem. When they’re missing, no amount of tooling helps.

Business does not run where you think it runs

Most of what new team members get wrong isn’t technical. It’s that they’ve never seen how a business actually operates.

Real work happens over email, in project management systems like Basecamp, with time tracked in TimeCamp. It happens on LinkedIn and Facebook, not TikTok and Instagram. A lot of young adults think email is for old people and attachments, and try to run professional relationships through Instagram DMs and Snapchat. That is not how real business is done. The moment a client hands you off to their operations person, their marketing person, and their video person, it collapses.

Two habits matter more than they sound.

Inbox zero means every email gets answered that day and the count actually reaches zero. Not skimmed, not filed, not deleted. When Leo started, his inbox was a disaster. Messages from clients, from teammates, from me, including the ones he assumed weren’t relevant to him. They were.

Time tracking matters because you do not know where your time goes. I’ve watched this play out for thirty years with a lot of friends of mine. You think you spent five minutes browsing. It was an hour and a half. Until you measure it, you’re guessing.

Level one work is the work AI takes first

Here’s the part that should change how you think about getting paid.

Editing a podcast is level one. It’s mechanical, it follows an SOP, and if the host asks the right questions it takes three or four minutes of actual effort. That’s why the best ones have almost no processing. Look at Joe Rogan. Anyone can do this. There are millions of people on Fiverr and Upwork who do exactly that, and Fiverr got its name because tasks were five dollars.

So when you ask to be paid per task, two things are true. A task is worth less by definition, and agents are already doing tasks for pennies. Competing there is like being an Uber driver competing with self-driving cars. The self-driving car will drive for nothing.

Leo learned this the hard way. He’d declined a flat weekly arrangement I offered because he wanted to be paid per task so he wouldn’t get lazy. Then he moved up to work that had real value, and suddenly there was nothing to invoice. Meetings, studying, learning, relationships. None of it fit the per-task model, even though that’s where the value was.

Level two and three look different. Turning videos into articles, updating SEO across sites, linking content together. That requires understanding how the client makes money, who they serve, what their whole body of content looks like, and where this one piece fits into it.

I flew to New York City and back just to interview Nathaniel Stevens, who sold his company for $341 million. Leo edited it. The cut was okay, but Nathaniel wanted changes and so did I, because Leo had operated the tool without understanding who Nathaniel is, why I flew out there, or what Nathaniel gets out of it. These people are smart and they have real business reasons. If we don’t know what those are, we’re the ones missing something. Once Leo actually understood Nathaniel’s companies, and saw the parallels with Justin Sonnenreich, whom he’d met over steak in Cologne, he could see how the content connects. That’s the jump.

Learn Do Teach

The framework underneath all of this is LDT: learn, do, teach.

First you learn it. Not memorize it, understand it. Then you do it yourself. Then you document it well enough that another person can follow your documentation and succeed without asking you a single question.

I use a recipe as the test. If someone else can follow your recipe and produce the finished meal without you intervening and without burning it, the recipe is good. That means real specificity: ingredients, steps, 350 degrees, and exactly how many minutes in the oven. Every detail specified.

Most people skip straight to performing. It’s the difference between watching Gordon Ramsay make beef Wellington on TV and actually having the recipe. Leo admits that’s exactly what he did at the start.

The payoff compounds. Those recipes train new humans and they train agents. Leo and hundreds of other people are contributing to what amounts to a master cookbook, and we’re proud of it. Nobody in digital marketing has more of this than we do, because anyone who does, we say let’s work together. We give it away, because your opinion of a recipe doesn’t matter until you’ve actually tried it. Most people never try it.

Reputation beats selling

There’s an easier path to clients than convincing strangers.

Most people are salespeople. They want to announce that the beef Wellington is the most delicious meal you’ll ever have. We never do that. We make the meal, people smell it, they taste it, they tell their friends. It speaks for itself.

That’s what a reputation does. Leo started with none. Why would anyone trust a 19-year-old? But he built connections with people who have one. Jan Koch in Cologne. Harry Gold in Boston, a friend of mine for fifteen years. We’ve done good things for each other for years without invoices, and those favors are worth more than anything I could have charged him.

Which reframes the “unpaid” time. Spending hours with those people isn’t a distraction from billable work. It’s the most valuable use of Leo’s time, because it builds his network and his reputation, and it creates value for us because everyone can see exactly what we do.

Most people start by borrowing our reputation to land the first client, because a vouch is what makes a stranger safe to hire. Then you do good work, you publish it, and eventually people come for you instead of for me.

Multiply value, don’t multiply hours

If you want to be rich, you have to multiply the value you create. You cannot work a hundred times more hours. You can build agents that do the work a hundred times over, and you can charge based on the impact.

Who cares if the agent did 99% of the work? But for the agent to do the work, you need the understanding. And to have the understanding, you had to do the learning and perform the task yourself the first few times, often unpaid. There’s no way around that step, which is exactly why most people never clear it.

The alternative shows up in my inbox constantly. Someone generates a hundred articles, bills me $10 each, and creates $2,000 of cleanup, plus the time it takes me to train them. Can you imagine paying someone to spray graffiti all over your building, and then paying someone else to scrub it off?

What it actually looks like

Ben Hamel was making $10 an hour at Sky Zone, scanning badges for birthday parties. He watched the free Dollar a Day training, implemented it at the one location he worked at, and it worked. The other trampoline park owners noticed the birthday party bookings. He scaled it across dozens of parks and built an agency. I had never met him. He came to me years later and said he at least owed me dinner.

Nothing was gatekept. The training was already public. Some parents pay $7,500 for their kid to come into our coaching program because they’ve decided it’s better value than a college degree in entrepreneurship or AI, but you don’t have to pay us anything.

The lifestyle is also less photogenic than Instagram suggests. No Lamborghini, no luxury airplane. We worked between rides at the park. We worked in the Uber. We worked at the restaurant. We recorded this at 2:30 in the morning. Is this a beautiful podcast studio? No. We’re mechanics, and the proof of a real mechanic is grease on their fingers because they’re working on the car.

The one-minute test

If you want to work with us, the entry requirement is deliberately simple.

Watch one of our training videos that’s at least twenty minutes long. Then immediately make a one-minute video explaining what you learned. Not tomorrow. Not when your hair looks good. Not once you’ve written the perfect script. Right then, imperfect, because that’s how you demonstrate you’re the kind of person who takes action when the circumstances aren’t perfect. You’ll never have perfect circumstances.

About 99% of people disqualify themselves right there. They want to get on a call and have me convince them that entrepreneurship is worth trying, or that AI is worth learning. If you need me to hold your hand, this isn’t going to be a fit.

It takes longer to make an excuse than to make the video.