Lessons for Finance from AI in Biotech
An analysis of AI's use at Recursion Pharmaceuticals, a biotech firm, offers parallels for the financial services industry. The key takeaway is that AI's primary value comes from delivering concrete operational efficiencies, such as compressing timelines, rather than from hype-driven narratives. The firm's success is attributed to its deep integration of AI with proprietary datasets and a focus on measurable outcomes, a lesson for deploying AI in areas like fraud prevention.
- Recursion was founded in 2013 by Chris Gibson, who developed the company's core technology as part of his M.D./Ph.D. work at the University of Utah, alongside co-founders Dean Li and Blake Borgeson. The company's AI platform, the Recursion Operating System (OS), uses automated, high-throughput cellular imaging and machine learning to create vast "Maps of Biology," analyzing up to 2.2 million experiments weekly. - A key operational efficiency is Recursion's dramatic reduction in the pre-clinical timeline; they have moved from identifying a new drug target to beginning Investigational New Drug-enabling studies in under 18 months, compared to an industry standard of 42 months. This is powered by their in-house supercomputer, BioHive-2, built in collaboration with NVIDIA, which processes one of the world's largest proprietary biological and chemical datasets (over 65 petabytes). - In financial services, similar AI-driven efficiency is seen in fraud detection, where machine learning models analyze user behavior and transaction data in real-time to identify anomalies. This approach has allowed some banks to increase the detection of suspicious transactions by 95% and reduce false positives by 70%, preventing significant financial losses. - The concept of using a proprietary dataset to train AI models is mirrored in the evolution of digital identity for fraud prevention. By combining behavioral biometrics—like typing rhythm and mouse movements—with device data and transaction history, firms create a unique, dynamic profile of the user, moving beyond static credentials to prevent account takeovers. - While the FedNow service, launched in July 2023, has seen rapid adoption by over 1,300 institutions, The Clearing House's RTP network, live since 2017, still dominates in volume, processing 343 million payments in 2024 compared to FedNow's 1.51 million. For product leaders, the key distinction lies in their transaction limits and primary use cases: RTP's $10 million cap is suited for large corporate B2B payments, whereas FedNow's recently increased $1 million limit is geared toward smaller-value transactions. - Institutional adoption of stablecoins for payments is gaining traction, with 13% of financial institutions and corporations now using them. A primary driver is cost savings in cross-border B2B payments, where 41% of current users report savings of at least 10%. Major payment networks like Visa and Mastercard are actively piloting stablecoin-compatible systems to enable instant, programmable settlements.
