Podcast Audio Transcript
Mimi: Hello listeners, this is Mimi; thank you for tuning in to yet another exciting podcast from Infosys BPM. Today, we will take a deep dive into Agentic AI and its role in lending function within financial services industry. While some of the banks are yet to adopt traditional AI, Agentic AI is now evolving as the game changer in AI development which is set to revolutionize financial services even further.
Joining me today is Megha Kochhar, Industry Principal, leading the digital transformation for Financial Services at Infosys BPM. She will help us understand how this autonomous form of AI is shaping lending. Welcome to the show, Megha! How are you today?
Megha: Thanks for having me on the podcast, Mimi. I’m doing good. Excited to be here.
Mimi: Great to hear that. We’re always on the lookout for new technologies and how they are shaping up our industry.
Megha, let’s start with the basics for those familiar with AI but new to the concept of Agentic AI. How does it differ from traditional AI, and why is it particularly significant in lending?
Megha: Great question Mimi. Traditional AI works within predefined parameters. For instance, it analyzes data, provides insights, and supports decision-making. Agentic AI, on the other hand, is autonomous. It doesn’t just analyze and predict; it acts, learns, and adapts in real-time without waiting for agent instructions.
For lending, this means, an Agentic AI based solution can dynamically optimize loan portfolios, manage risk autonomously, and create personalized financial solutions in real time. Exciting, isn’t it.
Mimi: Oh yes, definitely. I think this will be a game-changer.
Can we take a specific example like credit scoring, where traditional AI has already had a big impact. What more can Agentic AI bring to the table?
Megha: Traditionally, AI leverages past data like credit score, liabilities, income to assess credit history of a borrower, but now through Agentic AI, credit scoring will be assessed on the basis of continuously learning and updating models in real-time.
For example, if a borrower’s behavior indicates improving financial habits—like consistent savings or reduced credit card usage—an Agentic AI can independently adjust credit score or loan terms without manual intervention. This level of adaptability helps lenders add customers which traditionally would have been rejected while keeping borrower needs in mind.
Another example is to detect rising default risks in specific regions or industries during an economic downturn. Agentic AI will have ability to adjust lending policies—lowering exposure in high-risk areas while maintaining operations in stable ones. It can also flag at-risk loans and proactively recommend or implement restructuring plans, reducing potential losses.
Mimi: Very impressive.
Now, coming to its actual adoption in the banking industry, have any of the banks implemented or experimenting with such ideas?
Megha:Certainly, as I understand, JP Morgan has been incorporating AI and machine learning into credit scoring, fraud detection, and customer service. Their AI systems are designed to make more nuanced, data-driven decisions, to improve loan approval and risk management.
Another example in the area of AI driven fraud detection system is Danske Bank. According to AI Business, Danske Bank has improved their false positives by 60% and fraud detection capabilities by more than 50%.
So, I would say yes, banks are definitely adopting this technology as it evolves.
Mimi: Megha, that’s a big leap from how banks operate today! Sounds like a game-changer.
Do you foresee any changes in time taken to fund the loan at the origination process?
Megha: In fact, yes!
Agentic AI can pre-approve loan applications based on real-time borrower data, cross-verify documents using advanced NLP at the origination systems, and even interact directly with borrowers to clarify missing information. So, no more back and forth requesting documents or information. This directly impacts how fast the borrowers will be able to get their funds.
The point being, Agentic AI learns from previous applications to continuously improve its speed and accuracy. This enhances borrower experience.
Mimi: Okay. This sounds like Agentic AI is a perfect replacement for human agents. So, the important question is: Will lending operations run on their own without any agent requirement?
Megha: No, absolutely not, Mimi. What we are aiming with Agentic AI is to improve the straight-through rate and minimize exceptions. Areas where human in the loop will still be very much essential are – On the edge cases involving unique circumstances; auditing to ensure fairness and compliance adherence; complex lending scenarios like large scale corporate loans; and managing high value customers.
Mimi: It’s fascinating really. Let’s move to personalization. AI already helps lenders create tailored products, but how does Agentic AI push personalization to the next level?
Megha: One of the areas where banks are really looking for solutions is in the areas of payment default, recoveries and collections.
Let’s say a borrower runs a seasonal business—Agentic AI could proactively adjust monthly payments based on peak and off-peak income periods. It can even monitor borrower behavior post-disbursement, such as early repayments or financial stress signals, and modify terms to manage the relationship with borrower.
Mimi: I see. Do you think that this level of autonomy may impact fairness and create a bias, especially when it operates independently?
Megha: You raise a very critical point, Mimi. Service providers and financial institutions must ensure that fairness is designed into the system at multiple levels.
First, robust training must be ensured to identify and eliminate biases in data. Take Zest AI, a company that provides AI-powered underwriting solutions. Their AI tests for biases in the data like whether women or minority applicants are disproportionately rejected.
Second, continuous monitoring ensures the system’s autonomous decisions remain aligned with ethical and regulatory standards. Lenders must implement policy engines that encode regulatory requirements into the AI’s decision-making processes. These engines ensure that actions taken by the AI—like denying a loan—are automatically cross-checked for compliance. Additionally, regular audits to simulate AI’s behavior under various regulatory scenarios. For example, testing how Agentic AI handles borderline cases, ensuring that decisions remain fair and compliant.
Mimi: That’s incredible. Megha, this has been a masterclass on Agentic AI and its potential in lending, thanks for sharing examples to help us understand better.
Megha: Thank you—it will be exciting to see how Agentic AI redefines financial services for everyone – borrowers and banks alike.
Mimi: Thanks for listening, everyone! If you found this episode valuable, please subscribe and share it within your network.
Also, if you have any queries, do reach out to us through the email address on the podcast description. To contact us, visit us at InfosysBPM.com. Watch this space for more exciting podcasts coming up. Once again, thank you for tuning in, stay safe and sharp. Have a great day!
