Bank of New York Mellon Skips Tokenmaxxing Frenzy, Focuses on AI Outcomes

The recent frenzy surrounding tokenmaxxing has left some tech companies scrambling to climb leaderboards and burn through AI tokens. However, at Bank of New York Mellon (BNY), leadership took a different approach – one that prioritizes meaningful outcomes over token-counting exercises.

According to Dermot McDonogh, BNY’s Chief Financial Officer, the bank never bought into the idea that AI success is measured by how many tokens are spent or prompts are fired. Instead, BNY focused on building an internal platform that leverages Large Language Models (LLMs) in a way that’s both efficient and scalable.

McDonogh noted that token costs are “modest within modest” relative to BNY’s engineering budget. The bank’s leadership viewed tokenmaxxing as a distraction from more pressing concerns, focusing instead on outcomes rather than inputs. By doing so, BNY has managed to scale its AI efforts without getting bogged down in cost-per-query optimization.

BNY’s decision to eschew tokenmaxxing was not taken lightly. According to McDonogh, the bank had an early and deliberate AI strategy that involved building partnerships with hyperscalers and model providers, as well as fostering a culture of innovation within its walls. This approach has allowed BNY to harness the power of AI without fixating on cost-per-query metrics.

Internally, systems route tasks to the appropriate models, ensuring efficiency and eliminating the need for employees to optimize prompts manually. McDonogh’s focus is squarely on outcomes – a refreshing change from the prevailing narrative in the tech industry.

The impact of BNY’s approach is already showing up in financial metrics. 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 over the same period. McDonogh frames these gains less as cost savings and more as capacity creation – a subtle but important distinction.

BNY’s use of AI is not limited to a few high-profile projects. Rather, it’s embedded across its operations, with applications ranging from client onboarding to restricted-party payment screening. In the first quarter of 2026, more than 40% of BNY’s code was authored by AI, rising to roughly 50% more recently.

BNY’s approach serves as a much-needed counterpoint to the prevailing narrative around tokenmaxxing. By prioritizing outcomes over inputs and focusing on capacity creation rather than cost savings, the bank has managed to scale its AI efforts without getting bogged down in token-counting exercises.

As the industry continues to grapple with the implications of AI on productivity, BNY’s quiet rebellion against tokenmaxxing offers a compelling alternative. By measuring success not by how many tokens are burned but by how effectively AI changes what an organization can do, BNY is setting a new standard for AI success – one that prioritizes meaningful outcomes over fleeting metrics.

The question now is: will others follow suit? As the industry continues to navigate the complex landscape of AI adoption, BNY’s approach serves as a reminder that true innovation requires more than just throwing money at the problem. It demands a willingness to rethink the very metrics we use to measure success – and a commitment to prioritizing outcomes over inputs.