BNY AI metrics focus on outcomes over token counts, CFO says

BNY AI metrics are built around business results and workflow performance, CFO Dermot McDonogh told Fortune, rather than the amount of AI tokens the bank consumes. The distinction matters as companies try to prove that heavy spending on generative AI is improving productivity, not just raising usage numbers.

Fortune’s Jeremy Kahn has reported that “tokenmaxxing” became a prestige metric at some large technology companies, with engineers encouraged to rise on leaderboards by using more AI tokens. Kahn wrote that the practice distorted incentives and highlighted the distance between AI investment and measurable productivity gains.

McDonogh told Fortune that BNY did not adopt that framing. He said token costs are “modest within modest” compared with the bank’s wider engineering budget, and that executives considered the topic but saw it as a distraction from more useful measures.

What AI metrics does BNY track?

BNY tracks AI across core work areas including innovation, prospecting, onboarding, transactions and streamlining, according to McDonogh. In this context, the bank is measuring whether AI changes how work gets done, rather than counting prompts, tokens or deployed agents as stand-alone signs of progress.

McDonogh said the bank began building its AI program early after ChatGPT emerged. BNY has spent several years developing an internal platform that can work across large language models, while also forming relationships with hyperscalers and model providers, he told Fortune.

The bank’s internal systems send tasks to models suited to the job, which McDonogh said helps control efficiency without asking employees to manually fine-tune every prompt. He told Fortune he could not say how many prompts the bank used in a recent week because his attention is on outcomes.

BNY’s reported adoption numbers show where AI has entered routine work. McDonogh said more than 40% of the bank’s code was written by AI in the first quarter of 2026, and that the share has since moved to about 50%.

AI also helps draft about half of annual account plans, supports 25% of client onboarding and reviews about 70% of restricted-party payment screening, according to McDonogh. Those figures give the bank a more concrete view of AI’s role than token consumption alone would provide.

How is BNY tying AI to productivity?

McDonogh pointed to employee productivity measures as evidence that the bank is getting more capacity from its workforce. Fortune reported that BNY’s revenue per employee rose from $338,000 in 2022 to $401,000 in 2025, while pre-tax income per employee increased from $99,000 to $143,000.

McDonogh described the effect as capacity creation rather than a headcount-cutting exercise. He said BNY has not reduced its footprint but has been able to do more with the workforce it has.

The bank is also expanding an internal platform called Eliza, which McDonogh described as a firm-wide context layer that improves as it takes in more data and use cases. Employee access is structured through three proficiency levels, ending with a “pioneer” designation that requires training and testing.

Within the finance function, McDonogh said AI is already affecting regulatory reporting, balance sheet analytics and predictive modeling. He also said AI is helping with earnings preparation by pulling together analyst expectations and preparing for questions from investors.

This story draws on original reporting from Fortune.