Man Group uses AI to systematize the creation of trading strategies. Agents analyze academic papers for ideas, build code, run backtests, and construct signals. Over 15 models created this way are now trading client assets, proving the viability of automating research itself.
When asked about constraints like compute or data, Man Group's CTO identified the biggest hurdle as managing organizational change. The challenge is implementing AI's vast potential quickly while maintaining safety, risk management, and regulatory compliance in a fiduciary environment.
Man Group finds more value in meticulously pre-processing data than in using the latest frontier models for quant research. Adding descriptive, plain-English metadata that explains the context of the data (e.g., "each row is a person buying something") is key to unlocking meaningful insights.
Facing an 86x increase in token usage, Man Group chose to educate employees on model costs and capabilities rather than building an automated router. This transparent approach empowered users to find creative efficiencies, fostering a culture of responsible AI use instead of imposing top-down restrictions.
To maintain explainability and meet regulatory standards, Man Group's system requires AI agents to first write a clear, English-language investment hypothesis before writing code for a new trading model. This prevents the creation of "black box" strategies and ensures every trade is defensible.
The firm's AI spending is increasingly driven by autonomous agents executing entire workflows, not by individual employees in a chat window. This fundamentally changes corporate budgeting, creating a new challenge of allocating costs to cross-departmental processes rather than to specific people or teams.
AI's ability to process unstructured data (e.g., complex contracts, verbal trade info) is allowing Man Group to apply systematic trading to previously inaccessible markets like crypto and securitized credit. It helps standardize pricing and connectivity where no clean data feeds exist.
The investment firm now expects every new hire, from operations to the front office, to "up the bar" on AI proficiency. The ideal candidate is a long-term, strategic thinker who can orchestrate AI agents, rather than a specialist focused on granular, in-the-weeds execution.