WORLD - Banks are entering a new phase of artificial intelligence adoption in which systems are no longer limited to automating routine tasks but are increasingly capable of making decisions, coordinating workflows, and initiating actions across platforms with limited human intervention.
This emerging model, widely referred to as agentic AI, differs from traditional automation and earlier generative AI tools. Instead of responding only to prompts, agentic systems interpret objectives, access internal databases, evaluate incoming data, and determine subsequent operational steps.
Large-Scale Economic Potential
The shift is drawing significant attention across financial services. According to McKinsey, generative AI and related technologies could deliver between $200 billion and $340 billion annually in value to the banking sector through productivity improvements, revenue expansion, and stronger risk management.
However, industry adoption remains uneven. A recent Digital Banking Report, sponsored by OpenText, found that while 96% of financial institutions report some level of engagement with agentic AI initiatives, only 19% have progressed to production deployment.
Pilots Without Production
The gap highlights a persistent challenge: experimentation is widespread, but operational integration is limited. Many banks continue to run isolated pilot programs while struggling with structural barriers such as fragmented data systems, inconsistent governance frameworks, and unclear accountability structures.
Industry analysts note that the technology is advancing faster than institutional readiness. Deloitte has argued that AI generates the most value when workflows are redesigned around its capabilities rather than simply layered onto existing legacy processes. Yet only a small share of institutions report being able to integrate content, communication, and transactional data in real time.
Governance concerns are also emerging as a constraint. The same Digital Banking Report found that 52% of institutions lack confidence in their ability to manage governance systems required for compliant agentic AI deployment.
Early Movers in Banking
Despite these challenges, large banks are already scaling early use cases. Bank of America reported that its virtual assistant Erica has handled more than 3 billion client interactions, processing roughly 2 million customer requests daily.
Meanwhile, JPMorgan Chase has deployed its internal LLM Suite across approximately 250,000 employees to support research, customer service, and advisory workflows.
Consulting estimates suggest the productivity upside could be substantial if institutions successfully restructure operations. Accenture has projected potential productivity gains of up to 30% in banking, contingent on redesigning workflows rather than simply embedding AI tools into existing systems.
Workforce Readiness Gap
Workforce readiness remains another pressure point. Employees are increasingly expected to supervise AI-driven processes rather than directly execute them, requiring new skills in oversight, interpretation, and risk management.
However, only 14% of institutions rank employee enablement as a top AI priority, according to the Digital Banking Report. This suggests a widening gap between technological deployment and human readiness.
Risk, Regulation and Trust
Risk and compliance considerations continue to weigh on expansion. Issues related to data privacy, model transparency, cybersecurity, and accountability are becoming more complex as AI systems take on more autonomous functions.
As agentic AI expands into decision-making roles, regulators and internal risk teams are increasingly focused on explainability and control mechanisms to ensure outcomes remain auditable and compliant.
From Pilots to Operational Scale
Industry observers say banks that move beyond pilots will need to align business, technology, and risk functions from the outset, prioritizing data quality and governance before scaling use cases.
Rather than expanding experimental tools across departments, the emphasis is shifting toward building structured, enterprise-wide systems that can support AI-driven decision-making safely and consistently.
While adoption is accelerating, the sector remains in an early phase of transformation. Analysts say the competitive advantage will not come from early experimentation, but from the ability to operationalize AI at scale and restructure institutions around it.
