Banking AI Adoption Fails to Improve Earnings

The banking industry has achieved near-universal adoption of artificial intelligence tools, yet the share of organizations reporting any measurable impact on earnings has remained unchanged over the last four years [1]. While the technology has successfully accelerated individual tasks, the failure to address inefficiencies between departments has left the cost to originate loans flat and pull-through rates in decline [1].

The Efficiency Gap

The rapid integration of AI in banking has mirrored previous waves of digitization, where paper documents were converted to PDFs and moved behind portals [1]. While these iterations improved the interface for individual tasks, they failed to bridge the operational gaps between departments. For example, while mortgage closing times have decreased by approximately two weeks since 2021, the actual cost to originate those loans at banks and credit unions has remained flat [1].

This disconnect is also visible in customer-facing processes, such as retirement rollovers. Despite an application process that now takes five minutes, the underlying transfer often still relies on physical paper checks sent via mail [1]. This friction results in roughly one-third of job changers opting to cash out their retirement funds rather than complete the transfer [1]. Analysts suggest that because individual departments have dedicated budgets and owners, but the "space between" departments does not, digitalization efforts have consistently focused on document processing rather than end-to-end workflow optimization [1].

The Human Intelligence Deficit

Beyond operational silos, the industry faces a growing concern regarding the balance between artificial intelligence and human expertise. Recent data indicates that 80% of banks performed no derivative hedging to offset higher interest rates in 2023, highlighting a potential blind spot in basic financial literacy [3]. Industry observers argue that banks seeking a return on their AI investments must pair these tools with a renewed commitment to human intelligence, ranging from entry-level hiring to boardroom strategy [3]. Without this dual focus, the rush to automate may continue to yield faster task completion without improving the bottom line [1].

What to watch

• End-to-end metrics: Monitor whether banks shift from measuring task-level speed to tracking end-to-end process efficiency across departmental boundaries [1].
• Human capital investment: Watch for changes in bank hiring and training strategies that prioritize financial literacy and hedging expertise alongside technical AI implementation [3].
• Cost-to-originate trends: Track whether the cost to originate loans finally begins to decouple from the current flat trend as banks attempt to address inter-departmental friction [1].

The current banking AI gap suggests that technology alone cannot solve structural inefficiencies that lack clear ownership. Until banks assign accountability for the entire span of a process, the industry is likely to see continued speed gains that fail to materialize on the income statement [1].