The numbers come first.
BNY's revenue per employee climbed from $338,000 in 2022 to $401,000 in 2025. Pre-tax income per employee rose from $99,000 to $143,000 over the same period. Those are not projections — they are reported figures, and CFO Dermot McDonogh ties them directly to the bank's disciplined approach to artificial intelligence.
Skipping the leaderboard game
As Fortune reported, 'tokenmaxxing' — the practice of measuring AI success by the volume of prompts or tokens consumed — became a status symbol at several large tech companies, where engineers were urged to climb internal leaderboards by burning more compute. Critics argued the practice skewed incentives and widened the gap between AI spending and actual productivity.
BNY never played that game. 'It's not something we spend any time talking about,' McDonogh told Fortune, adding that token costs are 'modest within modest' relative to the firm's broader engineering budget. 'I couldn't tell you how many prompts we did last week,' he said. 'I'm focused more on outcomes.'
Building the platform, not the hype
Since the emergence of ChatGPT, BNY has spent several years constructing an internal, LLM-agnostic platform — branded 'Eliza' — and forging partnerships across hyperscalers and model providers. The system routes tasks to the appropriate models automatically, removing the burden of manual prompt optimization from employees. McDonogh credited CEO-level commitment and cultural adoption as equally important as the technology itself. 'There's been a demystification,' he said. 'People don't feel insecure about AI. That's a really important cultural point.'
Employee adoption is structured in three tiers of AI proficiency, culminating in a 'pioneer' designation that requires formal training and testing. Access to more advanced models is gated by expertise — a design that reinforces both quality and accountability.
Measurable penetration across the business
The operational footprint is significant. In the first quarter of 2026, more than 40% of BNY's code was authored by AI, rising to roughly 50% more recently. About half of annual account plans are drafted with AI assistance, 25% of client onboarding is AI-supported, and roughly 70% of restricted-party payment screening is reviewed by AI. In finance specifically, McDonogh points to regulatory reporting, balance sheet analytics, predictive modeling, and earnings preparation as active use cases.
McDonogh frames the productivity gains not as headcount reduction but as capacity creation. 'We haven't reduced the footprint, but it's allowed us to do more with the footprint that we have,' he said.
What the market should take from this
Capital rewards clear rules, and BNY's AI story is built on exactly that principle: define the outcome, measure it honestly, and let the platform scale. The bank's refusal to chase token-count theatrics is a case study in how free enterprise disciplines technology adoption — not through regulation or mandate, but through the simple accountability of a P&L. When AI spending ties directly to revenue per employee and pre-tax income per head, the incentive structure is clean.
The broader lesson for corporate America is straightforward: vanity metrics are a tax on productivity. BNY chose to pay the engineering cost once, build the infrastructure right, and let the income statement be the judge. So far, the income statement agrees.
