Banks have rushed to deploy AI across their operations, but a persistent data governance problem is making it difficult for institutions to prove what those investments are actually worth.
Research from nCino’s AI in Banking Benchmark found that 91% of banks have an AI strategy in place and 71% track clear key performance indicators (KPIs). Yet only 21% measure whether AI is contributing to increased revenue. The findings highlight a growing disconnect between AI adoption and the ability to demonstrate measurable commercial value.
The problem is not a lack of measurement altogether. Banks are already tracking how AI is being used and whether it improves individual processes. The difficulty comes when they try to connect those improvements to outcomes such as revenue growth, cross-selling and cost reduction.
nCino divides AI measurement into three layers: Activity, Capability and Outcome. Activity measures what AI does, including workflow automation and employee productivity. Capability assesses how AI improves existing processes, such as fraud detection or loan origination. Outcome measures the resulting business impact.
The first two layers are considerably more established. Employee productivity and workflow automation are tracked by 45% of banks, while 29% measure cross-selling and 26% track cost reduction. Only 21% measure increased revenue.
For RegTech and compliance teams, this creates a familiar data governance challenge. AI activity can be captured within individual systems, but measuring its wider impact requires information to move across business functions and data environments.
Siloed data remains a major obstacle. The nCino research found that 52% of banking leaders identify siloed data as their biggest data governance challenge. Deloitte’s State of AI in the Financial Services Industry survey of more than 570 financial services leaders found that 84% of firms have not yet redesigned their workflows around AI.
The same Deloitte research found that only 18% of firms are currently generating revenue from AI, despite 75% expecting to do so. The figures suggest that expectations around AI are running ahead of the infrastructure needed to measure its financial impact.
A bank could, for example, use AI to accelerate loan origination, improve fraud detection or help relationship managers identify new opportunities. However, the resulting revenue or cost saving may be recorded in another system, making it difficult to establish a direct link between the AI deployment and the final outcome.
This makes AI ROI partly a data lineage issue. Banks need to know where data originated, how it moved through different systems and how an AI-enabled action contributed to a final business result.
The benchmark identifies several requirements for reaching this level of measurement. Banks need to distinguish between generative, predictive and agentic AI deployments, measure productivity at the role level and create integrated data foundations capable of connecting previously isolated systems.
This shift could also have implications for AI governance. As banks use increasingly autonomous systems across regulated processes, being able to demonstrate how an AI deployment contributed to an outcome will become increasingly important for both internal oversight and regulatory accountability.
There is already significant appetite for a more connected approach. Some 94% of banks surveyed believe a fully integrated, end-to-end AI solution would deliver more value than AI deployed across isolated use cases.
For banks, the next stage of AI adoption may therefore be less about proving that the technology works and more about proving what it delivers. Without connected and well-governed data, institutions may continue to have sophisticated AI dashboards without being able to answer the question that matters most: what is AI actually worth to the business?
Copyright © 2026 FinTech Global
Copyright © 2026 RegTech Analyst
