Financial institutions are increasingly deploying agentic AI instead of general AI models, but approaching the tech cautiously. 

Some 33% of enterprise applications in financial services will feature agentic AI deployment by 2028, compared to less than 1% in 2024, with the agentic AI market expected to grow to $196 billion by 2034 with a CAGR of 43%, according to a recent report from data analytics company Market.us the report stated. 

Agentic AI consists of models capable of autonomous decision making and action while general generative AI models generate original content, Malcolm deMayo, global vice president of financial services at Nvidia, told BAN. 

Agentic AI solutions are developed for “a precise task, trained on specific data sets for a well-defined area of expertise,” Dimitrios Papanastasiou, global head of gen AI solutions at Moody’s Analytics, told Bank Automation News. Financial institutions are pivoting toward agentic AI rather than general AI models due to agentic AI’s ability to become an expert in its domain through reinforced learning, reduced hallucinations and boosted efficiencies, he added. 

FIs including JPMorgan, BNY, CommBank, Wells Fargo and Capital One are deploying agentic AI, deMayo said. Finding efficiency is a primary driver behind the its usage. 

Top use cases for agentic AI are: 

• Automated intelligent document processing;  

• Credit decisioning;  

• Personalized product recommendations; and 

• Repetitive tasks like unfreezing credit cards, transferring money, transcribing meetings and updating CRM systems. 

Maturing AI strategies 

Banks are maturing their AI strategies as they lean into the technology, so they are often looking for specific solutions from specialists that can fit their overall roadmap and architecture, Sairam Rangachari, chief product officer at core provider Temenos, told BAN.  

Agentic AI “agents can be highly specialized, modular and collaborating with each other as part of a broader agentic toolchain and broader architecture,” he said.  

After more than two years of AI experimentation, banks are under pressure to reduce costs and accelerate revenue generation opportunities, Rangachari said.  

“Customer expectations are … forcing banks to think beyond reactive chatbots on how to proactively help customers and to personalize at scale,” he said. 

READ MORE: ROI of AI: Is the juice worth the squeeze? 

Nvidia’s deMayo said: “Banks are favoring agentic AI for repetitive, mundane tasks and end-to-end journeys. The pivot for some use cases to agentic AI is a by-product of banks’ desire to continuously lower the cost of delivering financial services without sacrificing always-on customer service.” 

As AI is being deployed at scale within FIs, highly specialized tools that need limited intervention and have low error rates are in demand, Rangachari said. Agentic AI is the next frontier of AI deployment within financial services, he added. 

Cost and organizational structure 

Agentic AI models perform specific tasks while general AI models can do a bit of everything, John Kain, head of financial services market development at Amazon Web Services, told BAN.  

So, for banks, the question boils down to whether they should deploy:  

• Multiple AI models throughout operations vertically; or  

• One AI model horizontally. 

“With agentic AI models, the cost might be comparatively higher due to the need for multiple models. But agentic AI models can provide you much higher efficiency and accuracy, which might be difficult to measure in monetary terms,” Kain said. 

Agentic AI models are also able to communicate with each other, even if they are in different silos, and this simplifies operations, Rangachari said. 

Institutions with large operations and multiple business lines are shifting to agentic AI frameworks, compared to smaller institutions that can make do with one horizontal general AI integration. 

Choosing between agentic and generalized AI models, such as LLMs, comes down to need, Nickolaus Lachman, head of AI and data science for Morristown, N.J. -based Valley Bank, told BAN.  

Lachman said: “Agentic AI can automate tasks or processes to make them faster and reduce errors, whereas LLMs can be used to improve employee productivity in several areas, such as:  

• Generating first drafts of documents;  

• Summarizing complex content; and 

• Improving access to knowledge bases.”  

Agentic AI in action 

Financial institutions are deploying agentic AI in private banking, payments, wealth management, customer servicing and lending operations, Bentzi Aviv, global head of fintech solutions at Chesterfield, Mo-based fintech Amdocs, told BAN. 

FIs that have rolled out agentic AI solutions include: 

• Mastercard, which launched Agent Pay in April; 

• Moody’s Analytics, which has tools for knowledge finding and credit report generation in beta testing; 

• PayPal, which has a tool that simplifies and tracks payments; 

• Oracle, which uses agentic AI to fight fraud; and 

• JPMorgan Chase, which simplifies back-office tasks like email responses and treasury management. 

Banks ranging in size from the $1.7 trillion Citi to $61 billion Valley Bank are exploring the use of agentic AI across their operations. 

“We are looking to improve the quality of customer interactions and make them more efficient, specifically in voice and email,” Valley’s Lachman said. “We are also exploring the use of agentic AI to make AML/compliance processes more efficient, e.g. in speeding up transaction investigations.” 

Citi is exploring agentic AI for customer servicing and documentation processes, Terry O’Neil, head of connected commerce and strategic growth for Citi Retail Services. 

No sector in financial services will be untouched by the impact of agentic AI, Dean Leavitt, CEO of fintech Boost Payment Solutions, told BAN.  The NYC based company provides B2B payments and aids in cross border payments, according to their website. The company processed $23 billion in transactions in 2024, up 91% YoY, according to the company’s Feb 4 release.

“Agentic AI has the potential to transform payments by automating transaction processing and reconciliation, significantly reducing manual intervention and error rates,” he said. “Agentic AI can accelerate the path from opportunity to production — what we call ‘speed to spend’ — by streamlining decision-making and execution.”

Register here for “Seamless integration: The new frontier in embedded payments,” a free Bank Automation News webinar set for Tuesday, July 22, at 11 a.m. ET.