I thought I knew how to manage a team. I’d done it for years, built two companies from the ground up, and sold them. RemoteTeam was acquired by Gusto, and MovieLaLa found a home at Gfycat. I figured, how different could managing a team of AI engineers be? The answer, I learned the hard way, is very different.
Managing an AI team is tough. Managing one remotely is a whole other level of difficulty. The usual playbook? It doesn’t just fail; it backfires spectacularly. After five years of trial and error, of banging my head against the wall and wanting to throw my laptop out the window, I’ve come away with some hard-won lessons. They’re not what you’d expect. They’re counterintuitive. But they work.
1. Stop Hiring for Speed. Hire for Clarity.
The typical startup mantra is "move fast and break things." You want the 10x engineer who can churn out code at a blistering pace. I used to hire that way. For my first AI venture, I brought on a guy who was a coding machine. His GitHub was a thing of beauty, a torrent of commits. But we were a remote team, and his work was a black box. His code was complex, his documentation was an afterthought, and his explanations on Slack were cryptic. It took two other engineers the better part of a week just to understand and integrate one of his modules. We weren’t moving fast; we were stuck in quicksand.
That’s when it hit me. In a remote AI team, the most valuable skill isn’t the speed of coding; it’s the clarity of communication. Now, I hire for writing skills. Seriously. During interviews, I ask candidates to explain a complex machine learning concept to a non-technical person. I have them document a piece of code they’ve written. If they can’t write clearly, they’re not a fit. A brilliant idea is useless if you can’t explain it to your team. Clear writing is clear thinking, and in a distributed team, it’s the bedrock of everything we do.
2. Your Best Engineers Need the Most Management
Conventional wisdom says to give your senior engineers autonomy. Leave them alone, and they’ll produce magic. I tried that. I had a principal AI researcher, a true genius in her field. I gave her a high-level goal and told her to run with it. For weeks, I heard nothing. When I’d check in, I’d get a quick "making progress." I trusted her. That was a mistake.
When she finally emerged, she had built something incredible. It was also completely wrong for our business needs. We had wasted two months of runway on a research project that, while academically fascinating, was a commercial dead-end. The problem wasn’t her; it was me. I had abdicated my responsibility as a manager.
AI projects are not like building a CRUD app. The path is not linear. It’s a winding road of experimentation, dead ends, and unexpected discoveries. Your best people, the ones deep in the weeds of research and development, can easily lose sight of the bigger picture. They need more management, not less. Not micromanagement, but strategic guidance. I now have weekly one-on-ones with my most senior engineers where we don’t just talk about progress; we talk about direction. We question assumptions. We align their brilliant work with the boring, practical needs of the business. It’s the only way to make sure the magic they create actually moves the needle.
3. Kill the Daily Stand-up. Embrace the Weekly Write-up.
I used to be a stickler for the daily stand-up. 10 AM, every day, on Zoom. "What did you do yesterday? What are you doing today? Any blockers?" It seemed like a good way to keep everyone in sync. But for an AI team, it was a disaster. It broke their concentration. Deep work, the kind required for solving complex AI problems, can’t happen in 90-minute increments between meetings. The stand-up became a performative ritual, a recitation of tasks that didn’t capture the real, messy work of research and development.
So I killed it. I replaced the daily stand-up with a weekly write-up. Every Friday, each team member writes a short, clear summary of their week. What they worked on, what they learned, what problems they’re wrestling with, and what they plan to do next week. It’s all done asynchronously, in a shared document. The benefits were immediate and profound.
Engineers got their focus back. They could structure their days for long, uninterrupted blocks of deep work. The quality of communication went up. Instead of a mumbled, half-awake update on Zoom, I got thoughtful, well-articulated summaries of their work. And it created a written record of our progress, a knowledge base that became invaluable for onboarding new team members and tracking our evolution over time.
4. Unstructured Time is Your Most Productive Time
In a remote setting, it’s tempting to structure everything. To fill the calendar with meetings and status updates to simulate the "buzz" of an office. This is a trap. For an AI team, the real breakthroughs don’t happen in scheduled meetings. They happen in the spaces in between. They happen when two engineers are just chatting on Slack about a weird result from an experiment, or when someone is staring out the window, letting their mind wander.
I now actively encourage "unproductive" time. We have a dedicated Slack channel called #the-void where people can post random thoughts, interesting articles, or funny memes. We have virtual "coffee breaks" with no agenda. I even give every engineer a "curiosity budget" to spend on books, courses, or anything else that sparks their interest, whether it’s directly related to their work or not.
This isn’t about being lazy. It’s about recognizing that creativity in AI is not a linear process. It’s about fostering a culture of curiosity and serendipity. Some of our biggest breakthroughs have come from these unstructured moments. An offhand comment in #the-void led to a new approach to data augmentation that boosted our model’s accuracy by 5%. A book on cognitive science, purchased with the curiosity budget, inspired a new architecture for one of our core algorithms. You can’t schedule innovation. You can only create the conditions for it to emerge.
5. Your Gut is Lying to You. Trust the Data.
As an entrepreneur, I’ve always relied on my gut. It’s served me well through two exits and over 200 angel investments. But when it comes to managing an AI team, my gut has been wrong more times than I can count. I’d have a "good feeling" about a particular approach, a "hunch" that a certain model architecture was the way to go. I’d push the team in that direction, only to be proven spectacularly wrong by the data.
AI development is a science. It’s a world of metrics, of experiments, of rigorous validation. My gut, shaped by years of building SaaS products and marketplaces, was not calibrated for this new world. I had to learn to shut it up and listen to the data. This was humbling. It meant admitting that my intuition was not just unhelpful, but often actively harmful.
Now, we are ruthless about data. Every decision is backed by an experiment. Every new idea is framed as a hypothesis to be tested. We don’t have opinions; we have data points. This has been a cultural shift, and it hasn’t always been easy. But it has been transformative. It’s taken the ego out of our decision-making. It’s allowed us to fail fast and learn faster. And it’s the single biggest reason we’ve been able to make the progress we have.
Managing a remote AI team is a new frontier. The old maps won’t help you here. You have to be willing to throw out the playbook, to question your own assumptions, and to embrace a new way of working. It’s not easy. But if you’re willing to learn, to adapt, and to trust the process, the rewards are immense. You get to work with some of the smartest people in the world, on some of the most interesting problems of our time. And that, for me, is worth all the struggle.
Frequently Asked Questions
Can I implement all of these at once?
I'd strongly recommend against it. Pick the 2-3 items that resonate most with your current situation and focus there. Trying to do everything simultaneously is a recipe for doing nothing well.
How do I know which items apply to my situation?
Start by honestly assessing where your biggest bottleneck is right now. The items that address that specific constraint will give you the highest return on your time and energy.
Which item on this list has the highest impact?
It depends on your stage and context, but in my experience, the items near the top of the list tend to have the broadest applicability. That said, sometimes the less obvious items create the biggest breakthroughs for specific situations.