Banking Executives Cite Governance as Major Bottleneck for AI Performance

At the Sibos conference, the focus has shifted from whether AI works to whether organizations can trust it. [1]

According to Grant Thornton's 2026 Banking Insights AI Impact Survey, half of banking executives reported that compliance and governance are currently limiting AI performance. [1] Furthermore, only 18 percent of those executives felt confident they could pass an independent audit of their AI controls. [1] Similarly, a 2026 report by Databricks and Economist Enterprise revealed that fewer than half of organizations formally require a governance framework for autonomous agentic systems. [1]

Financial services leaders are looking to use AI to move beyond automated reporting cycles and start investigating the drivers of financial changes. [1]

For example, treasury officers want AI to provide real liquidity positions across all currencies and entities without requiring a days-long spreadsheet process. [1] At Sibos Miami, Databricks will present on using AI to find data breaks that inflate risk-weighted assets, aiming to release trapped regulatory capital while maintaining a traceable audit trail and a human maker-checker process. [1]

The survey data highlights a critical maturity bottleneck in enterprise financial AI: institutions are deploying capabilities faster than their compliance frameworks can audit them. While automated reporting is straightforward, integrating governed AI into daily balance sheet and liquidity decisions remains a challenge. The emphasis on human-in-the-loop maker-checker controls and auditable reconciliation trails reflects a necessary pivot toward regulatory survival rather than pure efficiency. This mirrors broader industry struggles to operationalize autonomous agentic systems safely under existing risk frameworks. What the sources don't address: how financial regulators are adapting their own auditing methodologies to evaluate these new AI-driven reconciliation trails in practice.