One of the World's Largest Hedge Funds on Its 86x Growth in Token Spending
Man Group executives discuss their implementation of generative AI across asset management operations, revealing an 86x increase in token consumption since January and demonstrating how AI agents are systematizing quant research, augmenting PM workflows, and identifying novel trading signals across multiple data modalities.
Summary
Gary Collier (CTO) and Tushara Fernando (Head of Data and AI) from Man Group describe a comprehensive AI integration strategy across the firm's full spectrum of asset management operations—from systematic/quant trading to discretionary fundamental investing. They explain that the shift from their previous "Head of Machine Learning" title to "Head of Data and AI" reflects the expansion beyond traditional ML techniques to include generative AI for both prediction and enablement of human creativity.
The hosts note that token consumption has grown 86 times since January, driven by both improved model capabilities and the emergence of agentic workflows that can autonomously handle complex tasks. Collier and Fernando describe three concrete applications: First, discretionary PMs now use AI agents to synthesize unstructured data from podcasts, broker research, and alternative data sources to surface material investment insights without consuming their limited research hours. Second, they've been developing AI systems for over a year that can autonomously ideate quant trading hypotheses, write code, validate strategies, and present them for human approval—with 15-20 models having successfully completed this pipeline. Third, they're using AI to structure and semantically tag proprietary data across market data, alternative data, and institutional knowledge so that AI can connect disparate datasets and identify cross-market insights.
A critical insight emerges around data infrastructure: Tushara emphasizes that frontier model capability matters less than proprietary data quality and semantic structure. They've invested heavily in tagging, metadata enrichment, and building unified semantic layers so AI can navigate between different data sources. Rather than implementing automated token routers, they've chosen education-based governance, making token spend and model tradeoffs transparent to 1,700+ employees and finding that this approach has driven creative efficiencies (like intercepting tool calls outside the agentic loop).
On organizational impact, they see both superstar effects (people orchestrating multiple agents at strategic levels) and democratization (non-technical staff using AI for routine tasks). They acknowledge ongoing tensions around whether subject matter experts will voluntarily encode their proprietary 10% alpha into AI playbooks versus keeping edge cases private. They've solved this partially by focusing on shared workflows (backtesting approaches, report reading) while allowing individuals to retain sensitive strategy particulars. They also note that bottlenecks are increasingly organizational and regulatory rather than computational—ensuring safe, compliant deployment of agentic systems across a regulated investment firm remains the constraint.
About this episode
<p>We've gone through a number of a technological revolutions in investing, whether it was the dawn of the high frequency trading era or the introduction of robotraders. When it comes to AI, the big question that remains in the investment context is whether or not the technology will be implemented like those past tech innovations — meaning it will be integrated into the flow of the business without upending everything as we know it — or if AI will transform the very nature of investing. Right now, AI's use in investing is a mixed bag: People are excited about its potential, but several firms are still trying to figure out its value. Today, we speak with Man Group's CTO Gary Collier and Head of Data and AI Tushara Fernando about how one of the largest publicly-traded hedge funds in the world is actually implementing AI into its work. We speak with them about empowering their quants with AI tools, the challenge of integrating AI safely, and the creative ways their staff is thinking about token spending, which is up 86-fold this year.</p> <p><br />Only Bloomberg.com subscribers can get the Odd Lots newsletter in their inbox — plus unlimited access to the site and app. Sign up here: <a href="https://www.bloomberg.com/subscriptions/oddlots?in_source=oddlotspodcast">bloomberg.com/subscriptions/oddlots</a></p><p>See <a href="https://omnystudio.com/listener">omnystudio.com/listener</a> for privacy information.</p>
Key Insights
- Man Group's token consumption increased 86x since January 2024, driven by agentic workflows enabling multi-hour tasks to be completed by AI, fundamentally changing how teams think about task allocation.
- Discretionary PMs now use AI agents to asynchronously synthesize podcasts, research reports, and alternative data into investment theses, allowing portfolio managers to access insights on-demand rather than spending hours on manual research.
- Man Group has deployed 15-20 AI-generated quant trading strategies that originated from autonomous hypothesis generation, code construction, backtesting, and validation before human investment committee approval.
- Frontier model capability ranks below proprietary data quality in importance for investment applications; structured, tagged, and semantically connected data drives alpha more than the latest model versions.
- Rather than implementing automated token routers, Man Group chose education-based governance, allowing employees to understand model trade-offs and create efficiency improvements voluntarily, resulting in creative cost reductions.
- Regulatory and organizational constraints now represent the primary bottleneck for AI deployment in finance, not compute or data—safely running agents across a regulated firm requires governance infrastructure that the industry hasn't fully solved.
- Man Group hasn't yet solved internal compensation alignment around who owns AI agent workflows, but has partially addressed it by encapsulating shared processes (backtesting, reporting) in AI playbooks while allowing individuals to retain proprietary strategy details.
- The scalability of agentic systems (with task duration doubling every seven months per the Meter benchmark) is shifting organizational focus from execution to planning, requiring collaboration across teams rather than within individual silos.
Topics
Transcript
Now, a message from Meta. Meta is launching America's Workforce Academy. The program offers paid training, a job, and a path to America's future. Because the future is for everyone. Learn more at meta.com slash America's Workforce Academy. Wise is the smart way to manage the currencies you need around the globe. When you send money abroad using your bank, you could get hit with hidden fees and exchange rate markups. There's a better way. Try Wise. Wise uses the exchange rate you'd usually find on Google with no unwelcome surprises. Plus, most transfers happen in under 20 seconds, which means your money arrives in less time than you've been listening to me. It's simple and free to sign up…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Odd Lots
These Are the Sharps Actually Making Money on Prediction Markets
This OddLots episode features journalist Adam and prediction market traders Brian and Daniel discussing how skilled traders consistently profit from prediction markets through rigorous research, data analysis, and ground-level reporting rather than luck or AI. The conversation explores market efficiency, insider trading concerns, the zero-sum nature of prediction markets, and how professional traders systematically beat retail investors.
How a Major Grocery Store Chain Can Dramatically Lower the Cost of Food
Scott Patton, Chief Commercial Officer of Aldi U.S., discusses how the grocery chain dramatically reduces costs through operational efficiency, including limited SKUs, private label dominance, direct farmer partnerships, and technological innovations like digital price labels and optimized barcode placement. The conversation covers Aldi's new Times Square location, supply chain strategies, and how consumer trends like GLP-1 adoption are reshaping grocery shopping behavior.
What Dan Wang Saw on His Last Trip to China
Dan Wang, author of "Breakneck" and China expert, discusses his recent month-long trip to China with podcast hosts Joe Weisenthal and Tracy Alloway, covering observations about phone culture, influencer culture, demographic challenges, and the tension between material abundance and widespread discontent among young people.
Carmen Li's Plan to Build a Futures Market for Compute
Carmen Lee, CEO of Compute Exchange and Silicon Data, discusses her efforts to build futures markets for GPU compute on the CME, drawing parallels to oil commodity markets. She explains how GPU price indices are constructed, who the natural buyers and sellers of compute futures would be, and addresses challenges like GPU performance variance and fungibility. The conversation covers spot markets, refurbished chips, residual value calculations, and the broader question of whether AI represents a bubble.
Anjney Midha's Plan to Radically Lower the Price of Compute
Anjney Midha, founder of AMP PBC and early Anthropic investor, discusses the physical and economic inefficiencies in AI compute infrastructure, arguing that most data centers run at dangerously low utilization rates. He explains how AMP is building a software-based 'grid' to standardize and optimize compute across heterogeneous chip types, similar to how AC/DC standardization unlocked electricity distribution. He also challenges the notion that only three frontier AI labs exist, arguing instead for a 'jagged frontier' with many specialized leaders.