The new AI head at Man Group, the largest publicly listed hedge fund, breaks down his plans to change how investors work

• Man Group is the largest publicly listed hedge fund with $161.2 billion in assets under management.
• Last month it launched a new data and machine learning group focused on generative AI.
• Tim Mace, who heads the department, outlines new capabilities his team is developing.

Man Group, a London-based investment firm with over $161 billion in assets, established a new data and machine learning team last month. The newly formed task force is laying the groundwork for its AI agenda, which has the potential to change the way its investors work for years to come.

Tim Mace, the new department’s leader, told Insider that the seven team members are tasked with developing and executing the firm’s generative AI agenda.

Man’s AI and machine learning firepower will be concentrated in order to reduce duplicative efforts across disparate trading desks and teams. Mace’s team will essentially function as an internal service provider, augmenting existing investing tools with AI and developing new capabilities from the ground up.

“The challenge actually for us right now is to make sure we are clear on what everyone else is doing in the business so that we can understand how best to bring these capabilities to them, which is something I’ve been trying to do in the data world for the last few years,” he said.

At Man, generative AI takes many forms, but they all point to what could be a so-called “alpha assistant” to whom investors could delegate their grunt work. In June, the hedge fund implemented ManGPT, a version of OpenAI’s ChatGPT that anyone in the company can use to generate ideas and summarize information. According to Mace, roughly 40% of the organization now uses that tool on a daily basis.

The company’s software engineers are generating code using generative AI, which could help Man’s 500 technologists be more efficient. Mace is also looking into using AI to lower the barrier to entry for complex tools that were previously only accessible via APIs and code, allowing less-tech-savvy investors to interact with complex systems through written language.

“If you ever get to a point where you can literally type ‘I would like you to do this, can you do this for me?’ “It knows about the tools we’ve already built behind the scenes, and you don’t have to worry about it, which makes your life a lot easier,” Mace explained.

Man Group’s AI building blocks

Mace now spends the majority of his time developing models that can search through a massive text database, such as large documents, and generate detailed, up-to-date summaries. Many people in the business, from legal teams to discretionary portfolio managers and data science teams, work with large document sets. Training a model to understand a person’s written question, sift through a large amount of information that is constantly growing and updating with new data, and return a written answer could have far-reaching implications, according to Mace.

Mace refers to these capabilities — chat, code generation, document search, and generation — as “building blocks” that, when combined, have the potential to completely transform how Man’s investors work.

The moonshot is a “alpha assistant,” a tool that front-office teams could use to search documents for a specific idea and generate a strategy based on the results. You could even have the assistant generate code and run a backtest against historical data to determine the viability of the strategy.

“All these components come together in a much bigger, much more capable system, and that’s where I think we’re headed,” he said.

Mace, who joined Man four years ago to build a data platform for the firm’s investors, believes AI has “a lot of synergy with data.” Quantitative funds make market bets using mathematical and computer-based modeling. As a result, Mace believes that combining Man’s data and machine learning efforts to investigate generative AI use cases was a natural next step.

Other AI techniques in investing, such as stock timing, have traditionally been used to make predictions or recommendations on future events based on historical data. Based on prompts, generative AI generates new human-like content such as images, text, or even code snippets.

“Generative AI is very different from how we use other AI techniques, which is really exciting,” says the researcher. It opens up all these new avenues that we hadn’t considered before.”