Key Takeaways
- • Northern Trust CEO Michael O'Grady said on an Oct. 21 earnings call that the executive team is "obsessed" with managing AI costs, and that this skill will be a key factor in the search to replace CFO Dave Fox, who retires at the end of Q1 2027.
- • Future Standard CTO Hari Moorthy, speaking at FT Live's Future of Asset Management North America conference on Oct. 19, said accuracy and cost sit at "the opposite ends of the spectrum today," with costs expected to level off over the next three to five years.
- • Prudential Financial's chief data and AI officer Bob Bastian argued that asset managers must redesign work processes, warning that without process reengineering AI usage is "a waste of time and, honestly, tokens."
- • Asset managers are increasingly adopting multi-model strategies, mixing frontier models from leading technology firms with cheaper, customized open-source models, as J.P. Morgan Chase CFO Jeremy Barnum described "minimizing lock-in and maximizing optionality" on an Oct. 13 earnings call.
Summary
Asset management executives entered 2026 expecting AI-driven productivity gains to appear in their financial results. Recent commentary, however, has shifted toward the cost of AI, which leaders now treat as an essential management task rather than a technical detail. Firms have broadly embraced AI tools, though reluctance remains about allowing AI systems to make final investment decisions. Goldman Sachs, for example, has implemented a three-pillar approach to AI across its asset management division and in July hired a Google veteran to lead the engineering team for asset and wealth management.
Cost control is becoming a board-level and C-suite concern as the price of top-tier models rises. According to the report, Northern Trust CEO Michael O'Grady said on an Oct. 21 earnings call that managing AI costs, including token consumption, is a component the firm must incorporate into its planning. He also linked the skill to the ongoing search for a successor to CFO Dave Fox, who is set to retire at the end of the first quarter of 2027. The concern arrives as capital requirements for AI development keep growing, with Anthropic expected to make its initial public offering later in November, and some investors question whether long-term demand will justify data center spending.
A central tension is the trade-off between accuracy and cost. Newer, more capable models handle harder problems but cost more to run, forcing business leaders to decide when the additional computing power is worth paying for. Hari Moorthy said cost scales directly with usage and that using multiple models causes costs to "explode," adding that a leveling off is expected over the next three to five years. Prudential Financial's Bob Bastian, speaking at the same FT Live conference, said reducing AI costs depends on matching tasks to systems suited to them rather than simply replicating human requests, and on redesigning work processes.
Many asset managers are avoiding reliance on a single model provider. Northern Trust, which names Microsoft as a key technology partner, uses multiple models to route tasks, including in cybersecurity, to manage costs, and O'Grady stressed the need to use both frontier models and newer open models. Bastian summarized the approach as "right tool, right job," noting that the optimal mix will change frequently.
Cost control is becoming a board-level and C-suite concern as the price of top-tier models rises. According to the report, Northern Trust CEO Michael O'Grady said on an Oct. 21 earnings call that managing AI costs, including token consumption, is a component the firm must incorporate into its planning. He also linked the skill to the ongoing search for a successor to CFO Dave Fox, who is set to retire at the end of the first quarter of 2027. The concern arrives as capital requirements for AI development keep growing, with Anthropic expected to make its initial public offering later in November, and some investors question whether long-term demand will justify data center spending.
A central tension is the trade-off between accuracy and cost. Newer, more capable models handle harder problems but cost more to run, forcing business leaders to decide when the additional computing power is worth paying for. Hari Moorthy said cost scales directly with usage and that using multiple models causes costs to "explode," adding that a leveling off is expected over the next three to five years. Prudential Financial's Bob Bastian, speaking at the same FT Live conference, said reducing AI costs depends on matching tasks to systems suited to them rather than simply replicating human requests, and on redesigning work processes.
Many asset managers are avoiding reliance on a single model provider. Northern Trust, which names Microsoft as a key technology partner, uses multiple models to route tasks, including in cybersecurity, to manage costs, and O'Grady stressed the need to use both frontier models and newer open models. Bastian summarized the approach as "right tool, right job," noting that the optimal mix will change frequently.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.