Workflow
Research Finds Two in Three US Banking Executives Believe Agentic AI Will Change How the Industry Operates

Core Insights - A significant majority (71%) of banking professionals believe that investing in AI agents is essential for maintaining competitiveness, while over half (54%) express a desire for stronger regulatory guidance [1][11] - The research indicates that 70% of respondents predict agentic AI will have a "significant" impact on the banking industry, with many viewing it as a "game changer" [3][4] Current State of Agentic AI in Banking - There is a notable divide in the adoption of agentic AI based on institution size, with larger banks leading the way; 64% of larger banks are testing or actively deploying agentic AI compared to only 38% of smaller institutions [8] - Among those already deploying agentic AI, 94% report satisfaction with their current implementations [4][6] Key Use Cases - The primary use cases for agentic AI identified by banking professionals include: - Customer Service (59%) - Employee Efficiency (39%) - Fraud/Risk Management (37%) [9] - A majority (73%) believe agentic AI will enhance customer experience, and 79% think it will improve employee experience [6] Adoption Barriers and Incentives - Major concerns hindering adoption include security and compliance (51%), concerns about autonomy (51%), integration with legacy systems (46%), and costs (42%) [17] - To facilitate adoption, over half of the respondents indicated that stronger regulatory guidance (54%) and seamless integration with existing tools (53%) would be beneficial [11] Technology Perception - Following a demonstration of SoundHound's Amelia 7 platform, 88% of respondents found the technology valuable for their business, and 98% believed it represented an improvement over current technology [12] Industry Outlook - The research underscores a strong enthusiasm for agentic AI among larger institutions, with a recognition of its potential to fundamentally change the banking industry through enhanced reasoning, action, and task performance capabilities [13]