GenAI & Value Investing: Fuel and Oxygen Are There. The Spark Is Missing

There are these two big ideas — and yet they’re drifting past each other like strangers.

On one side, we have the rise of AI. Deep learning, large models, transformer architectures. Researchers are building things that were impossible just a few years ago. Machines can now learn patterns from vast data, make intelligent predictions, generate new ideas. The power is real.

On the other side, we have value investing — not as a formula, but as a philosophy. The core idea that when you buy a stock, you’re buying a business. Not a symbol on a screen. A real business with customers, cash flows, capital allocation decisions, and an economic future. And the goal is not to guess the next price move, but to understand the value of what you’re owning.

Now here’s the tragedy.

These two worlds — AI and value investing — should have already met. They should have fused. But they haven’t. And every time I look around, I see missed opportunity.

Because most of the AI efforts in finance today are aimed at predicting prices. That’s where all the modeling energy goes — price time series, short-term trading signals, alternative data scraped and fed into ever more complex architectures. But this leads nowhere. It’s a zero-sum game. Even if you win, someone else loses. No value is created for the world. It’s cleverness chasing cleverness. At some point it’s just noise, refined.

At the same time, many value investors — even the thoughtful ones — are stuck in systems that look increasingly outdated. Regression-based backtests. Multi-criteria screeners. Mechanical filters like “Simple Way,” “Net Nets,” “Magic Formula.” These worked in the past because the world was simpler. But technology has moved forward. The tools exist now to go deeper — to learn, to simulate, to adapt — and yet they are not being used where they matter most.

And that’s what saddens me. Because I honestly believe that the best of both worlds is not only possible — it’s needed.

To my value investor friends, I want to say: there is something much more powerful today than a single screener or a factor sort. Machines can now learn complex relationships, adapt to regimes, understand context. You don’t have to throw out your philosophy. You just need to give it better tools.

To my AI friends, I want to say: look again at what investing really is. Not a prediction task for the stock, but for the business. Not just another dataset. The essence is understanding businesses — their cash flows, their cycles, their capital structures. That’s where value is created. And that’s what most AI models in finance are totally missing.

And yet — it could be different.

Advanced AI can already handle a lot of the nuances we care about. It can recognize cyclicality. It can flag inflated earnings. It can understand what happened to companies in 2008. It can learn what kind of stocks tend to underperform over time, and what characteristics the winners share. These are not technical limitations anymore. These are directional problems.

We are simply pointing the models at the wrong target.

What if, instead of predicting prices, we trained models to predict future business performance? Real operating numbers. Earnings quality. Returns on capital. Stability across cycles. What if we used machines to learn how real businesses evolve, not how prices flicker?

That’s where this could become something transformational.

And yes, I’ve spent years digging into these questions. I’ve used combined criteria systems like Portfolio123. I’ve watched strategies go through boom and bust cycles. I’ve studied the characteristics of losing stocks — not just the winners — because often the failures teach more. And I’ve come to believe this:

Long-term investing success is not about raw intelligence. It’s about temperament. About judgment. About holding on when it’s hard. If we want machines to be truly useful in investing, we don’t just need them to compute — we need them to learn that mindset. That’s not easy. But it’s possible.

The mechanical approach has huge advantages: no bias, no fatigue, no emotion. But it can also miss subtle traps — exceptional items in earnings, changes in capital allocation, unsustainable booms. That’s where human experience still matters.

So maybe it’s not machine vs human. It’s machine with human — each correcting the blind spots of the other. That’s the vision I keep coming back to.

In fact, I keep picturing it like this: the fuel is there. The oxygen is there. But the spark that brings them together — that’s what’s missing.

And this article’s mission, really, is just that, to bring this spark.

If we can bring together the philosophy of true investing with the capabilities of modern AI, we’ll stop building empty towers of prediction and start building something that actually understands value.

That’s not just a better model. That’s a better future for the world’s resources management and capital allocation and a lucrative non-zero-sum game for those who pursue it.