Five AI Questions We're Hearing from Financial Services Leaders

Last year’s AI question—does it work?—has evolved into whether AI can earn trust, with governance and compliance now cited by half of banking executives as key limitations in Grant Thornton’s 2026 Banking Insights AI Impact Survey. Only 18% expressed confidence in passing an independent audit of AI controls, underscoring a critical trust gap. Research by Economist Enterprise similarly found fewer than half of organizations mandate formal governance frameworks for autonomous systems, signaling systemic vulnerabilities in oversight.

The practical challenge lies in embedding governed data and AI into daily financial decisions, such as liquidity management and balance sheet reconciliation. Databricks will demonstrate a solution at Sibos Miami that identifies GL-to-Risk data breaks inflating risk-weighted assets, releasing trapped regulatory capital while maintaining audit trails. The approach emphasizes human oversight and regulator-friendly documentation, addressing a persistent pain point in financial reporting.

For anti-money laundering teams, AI’s value hinges on traceable, auditable workflows that distinguish real risks from noise. Databricks will showcase a live agent processing AML alerts using bank-specific data, drafting Suspicious Activity Reports while logging every step against KYC/AML evidence. The session on October 1 will explore how institutions can prove AI’s actions, not just its capabilities, in financial crime operations.

Client engagement and transaction banking also face AI integration hurdles, from scattered data to real-time decision delays. Examples include RBC Brewin Dolphin’s generative AI saving 4,700 hours annually on client-review prep, and Coinbase’s sub-100ms P99 latency fraud detection benchmark. Cost discipline remains a growing concern, with one CFO noting token expenses as trivial in early 2026 but warning of escalating challenges as adoption scales.