How the experts use AI to maximise returns – and what steps you can take too

The debate between AI and human fund managers doesn’t have to be an either/or proposition – and in fact, it shouldn’t be.

As Minotaur Capital’s Co-founder and Portfolio Manager Thomas Rice points out, AI can’t replace human judgement.

“AI is a powerful tool that can enhance what you do, but the investor still needs to decide what creates outperformance, what their edge is, and how markets work. That entire investment philosophy remains crucial,” Rice says.

You can see this demonstrated in Chris Conway’s year-long experiment where he pitted Gemini against Plato Asset Management’s Dr Don Hamson and Wilson Asset Management’s Tobias Yao. The fundies trounced their AI competition.

Equities

Fundie vs AI: One year on, who is the winner?

View

Equities

Wilson AM's Tobias Yao crushed AI backing founder-led businesses (74% to 39%)

View

AI can’t do it alone – but here’s what it can do. It can make life easier, and the investment industry has been using it to enhance their portfolios for decades. Just not necessarily in the ways you expect.

I spoke to Minotaur Capital’s Thomas Rice and Macquarie Asset Management’s Benjamin Leung to understand how they use AI as part of their different investment processes. For context, Macquarie's Leung runs systematic/quant strategies, while Minotaur's Rice runs a fundamental global equities strategy. They also discuss some of the opportunities it has helped them uncover, and find out the simple ways investors can incorporate AI into their own research processes.

What is AI when it comes to investing?

AI is a broad term these days. It’s easy to forget that older and more traditional processes like quantitative screening have existed for decades and even large language models (LLMs) have been around in some format and use.

After all, your basic ETF could be called a form of AI investing, trading automatically based on set algorithms, be it a straight index replication or smart beta, with the aim to outperform using certain metrics like quality or yield.

You’d be hard-pressed to find any fund manager who doesn't lean on some form of quantitative screening to help narrow down what might be worth investigating further. LLMs have taken it further due to the large-scale ability to process and interpret.

Leung highlights that tokenisation, neural networks, and machine learning that underpin large language models have been around for a long time. He uses a complex framework of criteria and information for investment decisions.

“For us, technology and data driven decision making have always been a pervasive part of the investment process, assisting with many parts such as data collection, modelling, risk management, and execution. 

For example, we have deployed machine learning techniques to refine momentum signals, determine the relative weighting of signals, and optimise trading,” says Leung.

For fundamental investor Rice, his own complex framework of data and information is about uncovering the prospect of mispriced opportunities.

“I’ve found that often the best opportunities are in situations where the current numbers don’t look particularly attractive – which is exactly why they might be mispriced,” says Rice.

Using AI to pull together vast sources of information

Rice has built a system called Taurient to help with every aspect, from idea generation to portfolio construction.

“We’ve built what we call “idea pipelines” that process information and flag situations worth investigating. One of our core pipelines focuses on identifying change – strategic shifts, new product launches, management transitions – basically any situation where there’s a higher probability that the market might be misinterpreting something,” Rice says, adding that the pipeline processes 35,000 articles weekly.

“We’re looking for qualitative signals that might indicate future value creation, rather than just filtering on current metrics,” he says.

Leung’s framework also looks beyond current metrics.

“We look at over 70 different criteria for stock selection and another 100+ for risk management across thousands of stocks. These factors include metrics commonly used by fundamental approaches, but expand to many more,” Leung says.

“These insights are then combined in very precise ways depending on the sector, country and market cap of the security.”

Using AI is far from a static activity; you have to keep evolving the tech and criteria. Rice mentions he has made around 1,700 updates to the codebase of Taurient over the past year alone, and some have been very stock-specific.

For example, he built custom tracking for subscriber counts and views across channels for one of his largest positions, Cover Corp (TYO:5253), which manages YouTube streamers, explaining that this is a crucial forward indicator for the stock.

Unexpected opportunities that AI uncovers

The ability to analyse information at scale can help businesses identify early opportunities or movements ahead of time.

As an example, Leung points point to the sharp rotation from growth to value in late 2020, which caught many by surprise, but Macquarie were able to tap into it early.

“Our investment process detected a subtle rise in correlation between the prevailing popular approaches – growth, trend following and quality investing. In response, the process increasingly favoured valuation-led investment ideas,” Leung says, highlighting the portfolio was rewarded for its earlier rotation.

For Rice, looking for mispriced opportunities, there’s been the chance to research companies he wouldn’t have otherwise considered. He shares the examples of Chugai Pharmaceutical (TYO:4519) and CyberArk (NYSE: CYBR).

“Our system flagged a Japanese-language article in Nikkei Business discussing their work on repurposing GLP-1 therapies for obesity treatment. That led us to investigate their pipeline more deeply, and we discovered a promising GLP-1 candidate that wasn’t getting much attention,” Rice says, noting that the position in Chugai Pharmaceuticals has been one of the biggest contributors to his portfolio.

Similarly, his system flagged a TechCrunch article about CyberArk’s $1.54 billion acquisition of Venafi, focused on machine identity security – an action he argues was an inflection point for the business.

“This is a rapidly growing segment driven by cloud services and AI adoption. Without our system’s ability to filter through vast amounts of content and spot these subtle shifts, we might have missed this inflection point,” he says.

No replacement for human judgement

Both Leung and Rice caution that using AI is about enhancing your activities; you still need to apply your own lens and research. AI is only as good as your inputs and technical setup.

Leung points to the role of experts to interpret the algorithms and provide insights into why certain decisions are made, while Rice views an overarching investment philosophy and application as critical. Rice doesn't think this will change, but rather AI continues to add as a support tool within the process.

“I think the future of AI in investing is about merging [your investment philosophy] and insights with AI’s ability to process vast amounts of information. Great investment returns still come from differentiated insight – AI just helps you surface and validate those insights more efficiently,” Rice says.

How investors can incorporate AI into their own process

In today’s world, investors have more ability than ever to incorporate AI into their own research process. It doesn’t have to cost money, either.

For example, those looking to start with some quantitative screens could look at Market Index scans as a starting point for research, for a range of trading signals like low P/E.

Some trading platforms are also increasingly incorporating screening and data into their platforms. Perth-based trading platform Marketech, for example, partnered with Bridgewise for AI-powered fundamental analysis. It ranks companies on the ASX based on fundamentals described as ‘future performance indicators’.

Going further, Leung suggests investors can use generative AI tools like Gemini or ChatGPT as a research assistant.

“Investors can articulate the rationales for their investment thesis or asset allocation and ask it to challenge their assumptions and conclusions. It may offer insights that might have been overlooked,” says Leung.

He cautions that AI can sometimes generate information that isn’t entirely accurate or relevant, so investors should view it as a helpful starting point for research rather than the final decision-maker.

Rice suggests Google’s NotebookLM, which allows you to upload multiple documents and ask questions. It is a free service.

“It’s particularly good for generating summaries and understanding relationships between different pieces of information,” he says.

For those with more budget, ChatGPT Pro offers access to Deep Research.

“The key with any of these tools is being specific in your prompts. Rather than asking for “a long report on Company X”, you’ll get better results by breaking it down into focused questions about specific aspects of the business,” Rice recommends.

Regardless of what approach you take, Rice and Leung remind investors there is no substitute for your own deeper analysis – AI is about uncovering research opportunities and enhancing your reach. The ultimate stock selection and portfolio construction should come down to your investment philosophy and analysis. 

Are you using AI for your investment process? Let us know your approach in the comments below.