Fintech firms embrace AI for credit, payments, customer insights, more

AI models help financial institutions in understanding delinquencies and speed up disbursements

India's decade-old fintech sector is putting artificial intelligence (AI) at the heart of its work, using the technology for purposes as varied as credit assessment and understanding complex data.

 

Gandhinagar-based Infibeam Avenues, a key player in the payment aggregation space through its CCAvenue brand, is just one fintech company investing heavily in AI. The company reported a 27 per cent yearly growth in consolidated revenue, reaching Rs 944.5 crore in the second quarter of FY25. Its expenses increased by Rs 900.5 crore, driven by “forward-looking investments” in AI-powered innovations, according to Vishal Mehta, chairman and managing director of Infibeam Avenues.

 

“In AI, the first place we are investing in is video intelligence. We are creating systems where the [AI] model can understand video, the context within that frame, scene identification and activity within an input. Once that is done you can enable payments on top of it,” said Mehta. 

 

‘Uberisation of payments’

 

He calls the model “Uberisation of payments”. It will be like the cashier-less technology at Amazon Go stores: Companies like Infibeam will enable customers to make payments using video intelligence technology that tracks and monitors items added to their shopping carts.

 

Infibeam’s Phronetic.AI division has secured $1 million contracts from hospitals and gas stations to deploy similar services. 

 

“For example, at gas stations we have the dispensation data and if I know the vehicle number and there is a payment instrument tied to it, it becomes seamless to conduct a transaction without having the driver to undergo any hassle,” said Mehta. 

 

Chennai-based fintech Kaleidofin assists financial institutions with its credit assessment model that is tailored for the informal sector and uses data models built on AI and machine learning (ML) rails across 30 million plus data points. 

 

“We have invested in data pipelines, in big data systems and as it becomes more scalable, we are able to ingest alternate data sources. We don’t just work on one data set or a customer segment, but we are able to cross cut data across multiple segments and see if there are similarities in these spaces,” said Natasha Jethanandani, co-founder and chief technology officer (CTO) of Kaleidofin. 

 

The company’s Kaleidofin Inclusive (Ki) score model offers credit assessment in the informal and under-banked sector. The AI/ML-based model uses demographic, geographical, credit history, savings and payments data to provide lending institutions with a probability of a delinquency for customers before extending credit.  

 

AI models help financial institutions get a view of delinquencies, streamline rejections, and speed up disbursements.

 

“Tractor financing is a product that we participated in. The number of days to issue a disbursement has gone down from 15 days to less than two days. Rejection rates have come down to 15-25 per cent, from around 45-50 per cent,  while still ensuring strong risk management, since the right type of credit is reaching the right people,” said Jethanandani.

 

Data points for the finance sector can include sanitation, drinking water, education, household income in an area and weather.

 

More than $3.5 billion worth of loans have been underwritten based on Ki, according to Kaleidofin. 

 

Generative artificial intelligence (GenAI), which can create new content, such as text, images, videos, or music, is helping in complex data. 

 

Help with data

 

“We will also see developer experience becoming richer. Developers are kind of big users in our customer base. A lot of configurations, data, understanding and setup will become simple and natural similar to the way we converse,” said Khilan Haria,  senior-vice president and head of payments, product at Razorpay, which offers payment gateway services for online merchants. 

 

For small customers, AI tools enable them to perform tasks such creating a payment link, automatically generate code snippets and assist merchants to integrate with payments platforms. 

 

“Let's say if you have built a system on top of React, Native and Python. You can auto generate a code with our concierge. You can understand who are the customers who have made payments from certain geographies across a specific timeline, and get actionable insights on particulars such as refunds, for example,” said Haria, referring to developer use cases.  

 

Fintech has to be careful in using GenAI, some in the industry have said. Rahul Chari, co-founder and CTO of PhonePe, believes that while one should be excited about GenAI, they also have to be very cautious.

 

“The reason I say that we have to be cautious when it comes to GenAI is because of the need to have explainability in, say, underwriting. This can lead to biases and it can go against financial inclusion,” said Chari in an earlier conversation with Business Standard. 

 

He believes that GenAI can be used as part of the development life cycle to increase productivity. Testing, documentation and observability are some of other areas where the technology can be used. “We have been envisioning using GenAI in insurance. For instance, the finer details of policy documents sometimes run into 15, 20 or 30 pages…can we make this conversational. Wherein the customer can ask questions and they get conversational questions and answers,” he said.