Agentic AI is Delivering a Whole New World in Financial Services

What Is Agentic AI?

Agentic AI systems combine natural language processing with decision-making engines. Vinesh Sukumar, vice president of generative AI and machine learning product management at Qualcomm, says that agentic AI’s autonomy can be particularly impactful in finance.

“In a perfect world, agentic AI improves decision-making and takes independent actions to achieve specific goals,” he says.

Ray Smith, vice president of AI agents for Microsoft, says agentic solutions are easier to build because they can connect to multiple systems, make sense of both the structured and unstructured data, and follow a business process through natural language instructions.

“This allows business users with few IT skills to simply describe the solution and connect to the required knowledge sources and systems to complete a business process,” he says.

What Are the Benefits of Agentic AI in Finance?

Agentic AI systems offer a transformative blend of efficiency, cost reduction and speed. “The biggest benefit is time savings,” says Five9 CTO Jonathan Rosenberg. “Whether assisting with customer purchases or managing operational workflows, agentic AI is lightning-fast and highly efficient.”

Financial services institutions (FSIs) stand to gain from increased sales, improved lead qualification and enhanced customer satisfaction. These systems can also personalize interactions by remembering previous engagements, fostering long-term brand loyalty.

Sukumar adds that agentic AI also has the potential to redefine user experience, particularly in consumer applications. “Imagine an agent arranging a movie night with friends, booking tickets and finding dinner reservations — all autonomously,” he says. “That’s the future we’re building.”

For FSIs, the potential is even greater: AI can analyze data, streamline processes and optimize resources, empowering employees to focus on strategic tasks.

Click the banner below to subscribe to our newsletter for the latest financial services IT insights.

Agentic AI has diverse applications across industries. FSIs are increasingly using in call centers it to interact with customers; help identify and mitigate financial fraud; and assist with decision-making, such as whether to issue a loan or credit. Many businesses, including financial services, are using agentic AI for simpler tasks such as document review and, increasingly, for higher-level tasks such as making investment decisions and optimizing portfolios.

By connecting disparate systems and analyzing data in real time, these agents can improve efficiency and reduce costs, Smith says: “Agents are always on, can bridge multiple systems, and can both reason and understand the data.”

On the customer service side, Sukumar says, AI agents can streamline insurance navigation to help determine what’s covered under a homeowners or car insurance policy, and can quickly help a customer make an appointment for an in-person consultation with a financial adviser. “From checking coverage to booking appointments, these systems can eliminate the frustration of lengthy customer service calls,” he says.

UP NEXT: Some companies are building artificial intelligence centers of excellence.

What Are the Challenges of Agentic AI in Finance?

For all its promise, FSIs should approach agentic AI with caution, prioritizing ethical considerations and solving security challenges to ensure responsible and secure implementation.

“Who is accountable for decisions made by autonomous systems?” Sukumar says, noting the importance of transparency and trust.

Agentic AI’s use in underwriting is a good example: FSIs must ensure that their models are free of the kind of bias humans are plagued with, and that humans are carefully reviewing the models’ recommendations before any final decision is issued. Beware of scenarios where AI agents make erroneous decisions, such as issuing unwarranted refunds or executing unauthorized transactions.

As malicious actors exploit AI systems to manipulate data or disrupt operations, IT teams also face heightened security risks. “Strong governance and robust data protection measures are critical to mitigating these risks,” Rosenberg says. 

Indeed, a strategic approach is critical to harness the potential of agentic AI. Sukumar recommends focusing on data accessibility and infrastructure.

For starters, modernize back-end systems and implement secure application programming interfaces. Organizations should also be sure that knowledge bases such as policies, procedures and training materials are accurate and up to date.

Experts say it’s important to start small and iterate. “Early wins and proven ROI can help align stakeholders and build confidence,” Rosenberg says. 

He also recommends using generative AI analytics to identify promising use cases for agentic AI. Qualcomm’s approach has included building tools and software that enable continuous optimization. 

“If an AI agent makes a wrong decision, it should be able to self-correct,” Sukumar says.

This process of executing tasks and self-correcting is crucial for refining systems and ensuring they meet organizational needs.

“We’re working with a lot of businesses to understand how we can scale agentic AI and optimize it for their workflows,” he says. “We are going to learn a lot.”