Artificial intelligence may be transforming the investment landscape, but for systematic fund managers, the rise of machine learning has brought as many risks as opportunities. Bias, bad sampling, misinterpreted signals and the sheer volume of noisy data now threaten to derail investment strategies that appear “cutting-edge” on the surface.

For Matus Mrazik, Investment Manager: Systematic Equities at Jupiter Asset Management and Portfolio Manager of the Old Mutual Global Equity Fund – winner of the 2025 INN8 Invest Diamond Award for Best Global Equity General – the lesson is clear: the real competitive edge does not lie in AI for AI’s sake, but in disciplined research, rigorous testing and human oversight. “The job is quite simple and yet super difficult: we try to come up with the best possible predictions for what’s going to happen next,” he says. And prediction, he notes, becomes exponentially harder when the data itself becomes distorted.

Where AI helps, and where it hurts

In the investing world, AI has become shorthand for innovation. But Mrazik is quick to challenge that assumption. “More data isn’t always better – unless you truly understand the noise, it can be more harmful than useful.” The challenge is not access to data, but understanding the structure behind it. Financial markets generate enormous volumes of unstructured information  such as analyst notes, earnings calls, company commentary, sentiment. Without careful filtering, algorithms simply amplify the noise.

That’s why Jupiter’s team leans heavily on statistical learning, a discipline that emphasises hypothesis-driven research and validation across different regimes and environments. “Machine learning cannot be treated as a black box,” Mrazik explains. It requires framing clear economic hypotheses, testing them on out-of-sample data, and repeatedly validating that they continue to hold under shifting market conditions. “We don’t try to predict macro events because it’s impossible; instead, we listen to what the market is telling us through observable variables.”

Why human oversight still matters

With AI tools advancing at extraordinary speed, maintaining intellectual property has become an unexpected challenge. “With all the coding tools now, you can write the code immensely quickly, which makes protecting intellectual property a real challenge,” says Mrazik. This is one of the reasons Jupiter keeps human oversight at the centre of its research process. “We try not to fully rely on AI because we believe there still needs to be human oversight.”

And while models can pick up patterns at scale, Mrazik reminds us that human interpretation remains essential. “Models learn, but insights belong to people – at least for now.”

Building the next generation of systematic investing

Behind the scenes, Jupiter’s Systematic Equities team engages in deeply technical work. One example is the use of NLP (Natural Language Processing) to convert vast volumes of qualitative content including analyst reports, management commentary and sentiment signals into structured, investable data.

But none of this is implemented automatically. Research from the team’s collaborations with academics across AI, behavioural finance and econometrics is added to the investment framework only when it reliably enhances alpha generation or risk-adjusted returns. “Our job is to outperform the benchmark and we are fully accountable for every single decision we make,” Mrazik explains. 

That accountability drives the team to differentiate itself deliberately. “We’re not only in the business of being better; we’re in the business of being different, because if everybody’s the same, the market breaks.”

It’s a reminder that systematic investing is not a race to adopt the newest algorithms — it’s an arms race in which clarity, discipline and conviction matter more than speed. “Success is probabilistic; we know how much we don’t know, and that’s what keeps us pushing to stay ahead in this arms race.”

The bottom line for advisers

For financial advisers, Jupiter’s experience offers several practical lessons:

• AI is powerful, but not a shortcut. Its outputs are only as good as the hypotheses behind them.
• Noise is the new risk. More data can worsen outcomes unless interpreted carefully.
• Human judgement remains irreplaceable. Oversight, accountability and intellectual clarity matter more than ever.
• Differentiation is a necessity, not a luxury. In a world where algorithms converge, unique insight becomes the real source of alpha.

As AI reshapes investing, systematic managers like Jupiter are proving that true innovation doesn’t replace human expertise, it just sharpens it.