Financial institutions are rapidly embracing agentic artificial intelligence tools that can make autonomous decisions without human oversight, with more than half of new AI use cases launched this year leveraging generative capabilities. According to Evident Insights research, the number of employees working on agentic AI at major banks has grown 13-fold in the past year, signalling a fundamental shift towards more sophisticated automation in financial services.
Agentic AI talent surge drives adoption
The 50 largest banks tracked by Evident’s AI Index now employ approximately 340 specialists working on agentic AI systems, representing a dramatic increase from negligible levels just 12 months ago. Five institutions dominate this emerging field: JPMorganChase, Wells Fargo, Citigroup, Capital One, and UBS collectively employ nearly half of all agentic AI specialists in the banking sector.
These specialists are developing multi-step autonomous systems that can execute complex workflows without human intervention. Unlike traditional AI tools that require constant oversight, agentic systems can plan, execute, and adapt their approach based on changing circumstances whilst working towards broader organisational goals.
Leading use cases demonstrate practical applications
Wells Fargo has deployed agentic loan underwriting agents built on LangChain’s framework, capable of re-underwriting loans by retrieving archived documents, extracting relevant information, matching data to internal systems, and performing calculations to determine loan eligibility. The system maintains human oversight for final review but dramatically improves processing efficiency.
BNY has developed a multi-agent sales platform called Eliza using Microsoft’s Autogen framework, employing 13 distinct agents that collaborate to recommend financial products to clients. The system analyses client questions and profiles to determine optimal product recommendations, though it currently maintains human involvement in the final presentation process.
JPMorgan leads research breakthrough
JPMorganChase’s research team has published significant advances in Legal Agentic Workflow (LAW), a suite of tools designed to assist legal teams with contract analysis. The system achieved over 95% accuracy in determining contract termination dates, compared to just 3% accuracy for OpenAI’s GPT-3.5-Turbo performing the same task.
The bank’s research demonstrates how agentic systems can handle complex reasoning tasks that traditional language models struggle with, particularly in extracting specific information from lengthy legal documents where context and interpretation are crucial.
Enterprise-wide AI adoption accelerates
Bank of America reports that over 90% of its employees now use Erica, the bank’s AI assistant, whilst allocating $4 billion to AI and new technology initiatives in 2025. JPMorganChase has rolled out its LLM Suite to 200,000 employees, representing two-thirds of its workforce, demonstrating the scale of AI integration across major financial institutions.
NatWest has deployed generative AI tools to 99% of its workforce, whilst Standard Chartered’s SC GPT platform now reaches 70,000 employees worldwide. These implementations represent unprecedented adoption rates for enterprise AI systems in the financial sector.
Market dynamics and competitive positioning
The rapid deployment of agentic AI reflects banks’ recognition that autonomous decision-making systems represent a critical competitive advantage. Capital One’s Chat Concierge, one of the few publicly available agentic banking tools, uses four distinct agents to guide customers through car-buying processes, demonstrating the practical application of multi-agent systems in customer service.
Citigroup’s CTO David Griffiths has stated that “the agentic model is becoming much more viable,” reflecting growing confidence in these systems’ readiness for production environments. The technology promises transformative productivity and revenue gains, though implementation remains largely theoretical rather than proven at scale.
Technology infrastructure evolution
The deployment of agentic AI requires sophisticated technical infrastructure capable of managing multiple autonomous agents simultaneously. Banks are investing in frameworks like Microsoft’s Autogen and LangGraph to create robust guardrails and interaction protocols between different AI agents within single workflows.
This infrastructure evolution represents a significant shift from simple AI tools to complex autonomous systems that can operate across multiple business functions simultaneously.
REFH – Newshub, August 14, 2025
