AI ROI Remains an Industry Blindspot

Financial institutions find it difficult to gauge the profitability of their AI investments due to limited insight into AI resource utilization, as per a Sibos 2026 panel discussion in Miami. A key factor in this challenge is AI tokens, which are fundamental units of text utilized by large language models and serve as the primary metric for measuring client usage.

Approximately two-thirds of respondents in the poll revealed they had limited or no visibility into AI token usage within their organization, despite being aware of their overall AI spending. Only 13% of participants reported having comprehensive visibility and the ability to actively manage costs at the token/model level.

Melissa Tuozzolo, HSBC's global head of client services, recounted some horror stories from clients and partners who allowed unrestricted use of AI tools, resulting in substantial bills. Isabel Schmidt, BNY's executive platform owner for payments enablement, stressed that while AI token cost is a critical factor, it's not the sole consideration in calculating AI ROI. Other expenses include data governance and cleaning, system upgrades, and the deployment of new technology.

David White, LSEG's global head of product and data, encouraged organizations to view AI as another technology to deliver value to customers. BNY's Schmidt highlighted AI's role in enhancing efficiency, effectiveness, and capacity, citing a case where an AI-powered "air traffic control" system reduced client query response times by 10 hours.

Similarly, HSBC's deployment of AI has freed up processing capacity, allowing for reallocation of resources. SMBC's AI-powered call center is projected to reduce costs by 70% compared to the older model, but SMBC's senior executive manager cautioned that not all client value-added scenarios lead to economic value.