AI in regulated industries: BofA and S&P on the huge gains to be had, and the risks of rushing in

At the Fortune AIQ Summit in New York, Bank of America’s CTO Hari Gopalkrishnan warned against rushing to AI as a solution, stating that simpler tools and process inventories are often better alternatives. Gopalkrishnan explained that the bank reviews every AI project through 16 risk pillars, including privacy, bias, workforce impact, and intellectual property.

His bank started using AI for fraud models over a decade ago and now employs Erica, a virtual assistant that handles over 3.2 billion client interactions, saving the bank the equivalent of 11,000 people answering calls. Additionally, the bank has spent approximately $400 million on AI and expects that budget to double next year.

The spending is carefully managed, with a model-agnostic approach and an orchestration layer that uses approved open-weight models for simple tasks and proprietary models for more complex reasoning.

S&P Global’s Chief Client Officer Sally Moore discussed how her 160-year-old company is repositioning itself aggressively while being careful. S&P is the world's largest credit rating agency and one of the largest index providers, with its financial data feeding regulated workflows. The company invests heavily in data, recognizing it as the currency within AI.

Moore credited the 2018 acquisition of Kensho, an AI company, as a significant advantage for S&P. S&P has since integrated Kensho into its business, splitting its Market Intelligence into two units: Kensho Data & Platforms and Enterprise Solutions. The company has also created a Chief Client Office, led by Moore, to work more closely with clients on AI.

S&P serves 60,000 clients at various stages of AI adoption, providing assistance with tasks ranging from bankers preparing for meetings to specialized agents built for specific jobs. Moore emphasized the importance of using the right tool for the job, acknowledging that AI is not always the best solution.