I’ve seen a lot of AI teams. A lot. As a founder with a couple of exits and an angel investor in over 200 companies, including some names you’d definitely recognize, I get pitched constantly. Everyone’s got a world-changing AI idea. But you know what’s funny? Most of them are focused on the wrong things.
They obsess over having the most PhDs from Stanford, the biggest GPU cluster, or the most complex model architecture. Those things are nice to have, sure. But they aren’t what make or break a team.
We just spent six months digging into the data from over 500 AI teams—some wild successes, some quiet failures, and a lot in between. We wanted to find the signal in the noise. The one thing that really, truly predicts success.
The answer we found was so simple, I almost didn't believe it. It’s not about the tech. It’s not about the funding. It’s not even about the idea itself.
So what is it? Let's get into it.
The Great AI Team-Building Myths
Before I tell you what does work, let's talk about what doesn't. When I talk to founders, I hear the same assumptions over and over. They’re chasing ghosts, spending millions on things that feel important but have almost no correlation with actual success. Here are the top three myths I see every single day.
Myth #1: You Need a Team of Elite AI Researchers
This is the big one. Everyone is trying to hire the same 100 people from OpenAI, DeepMind, or Anthropic. I get it. Having a star researcher on your team looks great on a pitch deck. It feels like a shortcut to credibility.
But here’s the hard truth: most companies don’t need to invent new model architectures. They need to apply existing ones to a specific business problem. The skills required for that are completely different. You need product-minded engineers, not just pure researchers. I’ve seen teams with a handful of brilliant, but product-agnostic, PhDs spin their wheels for a year, producing nothing of value. Meanwhile, a scrappy team of full-stack engineers who know how to use an API can ship a feature in a week.
One of my portfolio companies, a small startup in the legal tech space, learned this the hard way. They spent their entire seed round hiring two top-tier ML researchers. The researchers wanted to build a foundational model for legal text from scratch. It was a fascinating academic problem. But the company ran out of money before they had a single customer. They were trying to win a Nobel Prize when they should have been trying to solve a customer’s pain point.
Myth #2: More Data and Bigger GPUs Are the Answer
“If we just had more data…”
“If we just had a bigger GPU cluster…”
I can’t tell you how many times I’ve heard this. Founders love to blame their tools. It’s an easy out. It’s a way to avoid the harder questions about product-market fit and execution.
Of course, data and compute are important. But they are not the bottleneck for 99% of teams. The real bottleneck is the ability to ask the right questions of the data you already have. It’s about creativity and product sense, not just brute force.
I remember when we were building RemoteTeam. We didn’t have massive datasets. We had a few hundred companies using our platform. But we were obsessed with their problems. We talked to them constantly. We used the qualitative data from those conversations to guide our product development. That was infinitely more valuable than some terabyte-sized dataset of anonymous user behavior.
Myth #3: A “Chief AI Officer” Will Solve Everything
This is a newer, and particularly dangerous, trend. Big companies, feeling the pressure to “do something” about AI, are rushing to hire a Chief AI Officer (CAIO). They throw a massive salary at some big name, give them a fancy title, and expect magic to happen.
It rarely does. Why? Because AI isn’t a separate function you can silo away in a corner of the org chart. It’s a horizontal capability that needs to be woven into the fabric of the entire company. A CAIO without a deep connection to the product and engineering teams is just a figurehead. They can create beautiful slide decks about the “future of AI,” but they can’t actually ship anything.
The most successful companies I’ve seen don’t have a CAIO. They have product managers, engineers, and designers who are all fluent in the language of AI. They have a CEO who understands how to create a culture of experimentation and learning. AI is a team sport, not a solo performance.
The One Factor That Matters
So if it’s not the researchers, the data, or the fancy titles, what is it? What’s the single most important factor?
It’s the speed of iteration.
That’s it. The single biggest predictor of success for an AI team is how quickly they can go from an idea, to a prototype, to real user feedback, and back to an improved idea. The teams that win are the ones that can cycle through this loop the fastest.
Why is this so critical in AI? Because nobody knows what’s going to work. We’re in the very early innings of a new technological paradigm. The assumptions you have today will be wrong tomorrow. The only way to navigate this uncertainty is to experiment constantly. You have to be willing to be wrong, and you have to be set up to find out you’re wrong as quickly and cheaply as possible.
The Anatomy of a High-Iteration Team
When I look at the AI teams in my portfolio that are absolutely crushing it, they all share a few common traits that enable this rapid iteration.
They are stacked with product-minded engineers. These are people who can not only build, but can also think critically about the user. They can take a vague customer problem and translate it into a concrete technical solution. They’re not waiting for a perfect spec. They’re partners in the creative process.
They are obsessed with the customer. They don’t hide behind their keyboards. They are on the phone with users every single week. They have a visceral understanding of the customer’s pain. This is so important. You can’t solve a problem you don’t feel in your bones.
They have a bias for shipping. They would rather ship a V1 with 80% of the features than wait six months for a “perfect” V2. They understand that value is only created when something is in the hands of users. Perfectionism is the enemy of progress in a fast-moving field like AI.
They use the right tools for the job. They don’t build everything from scratch. They are masters of leveraging existing APIs, open-source models, and third-party tools. They know that the goal is to solve the customer’s problem, not to win a science fair. If a simple API call can do the job, they use it. They save their precious engineering resources for the parts of the problem that are truly unique to their business.
I saw this firsthand with a company I invested in last year. They were building an AI-powered tool for sales teams. The founding team was two engineers. They spent the first month not writing a single line of code. Instead, they did 100 interviews with sales reps. They shadowed them on calls. They lived and breathed their workflow.
By the end of that month, they had a deep, almost painful, understanding of the problem. Then, and only then, did they start building. And because they were so clear on the problem, they were able to build an incredibly simple, yet powerful, solution in just a few weeks. They used off-the-shelf APIs for transcription and summarization. They focused all their energy on the user experience and the workflow integration.
They launched to a small group of beta customers and the feedback was electric. They were iterating on the product daily based on user feedback. Within six months, they had a product that was being used by some of the biggest sales teams in the world. They didn’t have a team of PhDs. They didn’t have a massive GPU cluster. They had a deep understanding of their customer and a relentless focus on iteration.
How to Lead a High-Iteration AI Team
So how do you, as a leader, create an environment where this kind of rapid iteration can happen? It’s not about just telling your team to “move faster.” It’s about creating the psychological safety and the operational structure that allows them to do so.
1. Define the Problem, Not the Solution
This is the most important job of a leader in an AI-driven company. Your job is to be relentlessly clear about what problem you are solving and for whom. But you must resist the urge to dictate how the team solves it. Give them the context, the customer insights, and the business goals. Then get out of the way and let them figure out the technical implementation.
2. Celebrate Smart Failures
If you want your team to experiment, you have to make it safe for them to fail. Not all failures are created equal, of course. A failure due to sloppiness or carelessness is not okay. But a failure that comes from a well-designed experiment that disproves a hypothesis? That’s a smart failure. That’s learning. And you need to celebrate that learning just as much as you celebrate the wins. If every experiment has to be a success, nobody will take any risks.
3. Shrink the Batch Size
Don’t ask for a six-month plan. Ask for a one-week plan. What’s the smallest possible thing we can build and ship to learn something new? The bigger the project, the more momentum it gathers, and the harder it is to change course. Small batch sizes allow you to stay nimble and react to new information.
When I was running MovieLaLa, we had a rule: every Friday, we had to ship something. It didn’t have to be a huge feature. It could be a small bug fix, a tweak to the UI, or a simple A/B test. But we had to ship something. That weekly cadence forced us to break down big ideas into small, manageable chunks. It kept us honest and it kept us moving.
It’s a Culture, Not a Strategy
Building a high-iteration AI team isn’t about a specific strategy or a project management methodology. It’s about culture. It’s about creating a team of curious, customer-obsessed builders who are empowered to take risks, learn quickly, and ship constantly.
Forget about hiring that one rockstar researcher. Forget about building the biggest data moat. Focus on building a team that can out-learn and out-iterate everyone else.
Because in the world of AI, the fastest learners win. It’s that simple.
I’ve seen it happen over and over again. The team that looks best on paper is rarely the one that ends up on top. The winner is the scrappy team in the corner that’s just a little bit faster, a little bit closer to the customer, and a little bit more willing to be wrong.
So, take a hard look at your team. Are you set up for learning, or are you set up for looking good in a board meeting? The answer to that question will tell you everything you need to know about your chances of success.
Frequently Asked Questions
What experience informs this perspective?
This perspective comes from over a decade of building companies in Silicon Valley, two successful exits (RemoteTeam to Gusto, MovieLaLa to Gfycat), and investing in 200+ startups including Anthropic, OpenAI, and Scale AI. I write about what I've lived.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.
How can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.