I just got off the phone with the 50th AI risk manager on my list. Fifty conversations with some of the sharpest minds at top hedge funds and investment banks. I started this project expecting to find a secret sauce, a dozen different complex strategies for managing AI risk in finance. What I found was something far more surprising: they all said the same thing.
Not literally, of course. But when you cut through the jargon and the corporate-speak, the core message was identical. It wasn't about some groundbreaking new algorithm or a complex mathematical formula. It was about a return to first principles, a set of simple, powerful ideas that anyone in the AI space should take to heart.
I’ve built and sold two companies, RemoteTeam and MovieLaLa, and now I spend my days as an angel investor in over 200 companies, including some of the biggest names in AI like Anthropic, OpenAI, and Scale AI. I’ve seen firsthand how easy it is to get seduced by the complexity of AI. We love to talk about neural networks with billions of parameters and models that can write poetry. But in the world of high-stakes finance, where a single error can cost millions, the conversation is refreshingly different.
Here’s what the top 50 AI risk managers in the world are actually focused on.
1. Your Model is Only as Good as Your Data
This might sound obvious, but it’s the point that every single person I spoke to hammered home. Before they even think about the model, they are obsessed with the data. Garbage in, garbage out. It’s a cliché for a reason.
One manager at a multi-billion dollar fund told me a story that stuck with me. They had a team of PhDs build an incredibly sophisticated trading model. It backtested beautifully. They were ready to deploy it and print money. But in the final review, a junior analyst noticed something odd in the data. A single data source had been mislabeled for a period of three months. It was a tiny error, but it was enough to completely invalidate the model's predictions. They had to scrap the whole thing and start over.
That story is not unique. I’ve seen it happen time and time again. We get so excited about the power of our models that we forget to check the foundation they’re built on. The best AI teams I’ve seen are the ones that have a healthy paranoia about their data. They are constantly cleaning it, testing it, and trying to break it. They know that the real competitive advantage isn’t in the model, it’s in the data.
So, before you go and hire a team of expensive data scientists to build you a state-of-the-art AI model, ask yourself this: is your data clean? Is it accurate? Is it complete? If the answer is no, then you’re just building a house of cards.
2. The Human is the Most Important Part of the Loop
There’s a myth in Silicon Valley that AI is going to automate everything and make humans obsolete. The people actually managing billions of dollars with AI know that’s nonsense. For them, the human is the most important part of the system.
Every single risk manager I talked to had a story about a time their model went off the rails. A flash crash, a sudden political event, a global pandemic—these are the kinds of black swan events that no model can predict. And in those moments, you need a human to step in and take control.
One of my portfolio companies, a fintech startup that uses AI to provide investment advice, learned this the hard way. They had a beautiful, automated system that was delivering great returns for their clients. Then COVID-19 hit. The market went into a freefall, and their model, which had never seen anything like it, started making some very strange decisions. Luckily, they had a team of human advisors who were able to step in, override the model, and protect their clients’ assets.
That’s why the best AI risk management teams are not just made up of quants and engineers. They also have traders, economists, and psychologists. They understand that the markets are not just a collection of numbers, they are a reflection of human behavior. And you need humans to understand humans.
3. Simple is Almost Always Better
I’m a tech guy. I love complex, elegant solutions. But in the world of finance, complexity is a liability. The more complex a model is, the harder it is to understand. And if you can’t understand a model, you can’t trust it.
This was another point that came up in almost every conversation. The risk managers I spoke to were deeply skeptical of black-box models. They want to be able to see how a model is making its decisions. They want to be able to explain it to their bosses, to their clients, and to the regulators.
One of the most successful hedge funds in the world, a name you would all recognize, is famous for its use of simple, linear models. They’ve been outperforming the market for decades with models that you could probably build in an Excel spreadsheet. Why? Because they understand them. They know their limitations. And they know when to trust them and when not to.
I’m not saying that there’s no place for complex models in finance. But I am saying that you should always start with the simplest model that can get the job done. And you should only add complexity if you have a very good reason to do so.
4. The World is Not a Static Place
The final principle that everyone agreed on is the need for continuous monitoring and adaptation. The market is a living, breathing thing. It’s constantly changing. And if your model isn’t changing with it, it’s going to become obsolete very quickly.
The best AI teams have a process for constantly monitoring their models. They are looking for signs of drift, for changes in the data, for anything that might indicate that the model is no longer performing as expected. And when they find something, they are not afraid to go back to the drawing board.
This is not a one-time thing. It’s a continuous process of iteration and improvement. It’s about having the humility to know that you’re never going to have a perfect model. And it’s about having the discipline to constantly make it better.
My Takeaway
So, what’s my big takeaway from these 50 conversations? It’s that the secret to AI risk management is that there is no secret. It’s about a relentless focus on the fundamentals. It’s about data quality, human oversight, simplicity, and continuous improvement.
These are not the sexiest ideas. They’re not going to get you on the cover of a magazine. But they are the ideas that are going to keep you from blowing up your fund, your company, or your career.
As I continue to invest in the next generation of AI companies, these are the principles that I’m going to be looking for. I’m not interested in the hype. I’m interested in the discipline. Because in the world of AI, that’s the only thing that matters.
Frequently Asked Questions
Do all experts agree with this view?
No, and that's fine. The best ideas in business are often contrarian. I share my perspective based on my experience and data, but I encourage you to seek out opposing viewpoints and form your own conclusions.
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.
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.