Digital financial services firm Navi is deploying generative artificial intelligence (GenAI) to detect customer fraud and falsification of information, its founder and managing director Sachin Bansal told Moneycontrol in an exclusive interview.
According to him, large language models or LLMs are good at extracting data from videos and images, a capability that is being put to use by the fintech firm. For instance, if a customer looks like a 20-year-old and the age provided in the PAN card gives the age as 40 years, that's potentially a fraud signal for the company.
“This fraud detection of age or gender, like the Aadhaar card says female and person is male... And let's say the address that they have put down is a residential address, but they are at a shop somewhere, things like that... Those paradigms will be very interesting to explore. It could do things which humans can't even do or we will have to put like an army of people to do those things,” Bansal said.
Having made a fortune at the time he exited the country's largest e-commerce company, the Flipkart co-founder, invested a large chunk to start Navi, a digital and mobile-first financial services company. Given his strengths in technology, Navi was expected to use the most advanced technologies to scale up.
Bansal said that AI could impact different areas and could range from small incremental benefits to paradigm changes. Navi has deployed ChatGPT, the chatbot that first caught the world's attention, and several other free AI models. It is using Microsoft’s Copilot AI tool for coding. According to Bansal, these tools are helping augment humans and making them more productive.
“GPT continues to be the best as of now. The other area that it will apply is in data annotation work, it's able to do it much faster. And I think over time, we could even use it for (credit) underwriting as well. For example, Copilot can help in complete paradigm changes. Things that we couldn't even do with humans,” Bansal said.
Also read: Premium for good governance will be very high, says Sachin Bansal
As of now, Navi’s 80-strong data science team looks at different AI models and their applicability for Navi. While it is a horizontal function that could affect various divisions, Bansal said that a team focussing on AI can add more value to the time and effort to create specialised models for the firm.
“It is a specialised function. It is not that anybody and everybody can do it well. Anyone can do a decent job of it. Let's say if you and I can sit and maybe write decent prompts. But to do an excellent job of it, I think it requires a bit of nuance,” Bansal said.
According to him, the rise of AI has enhanced the respect for technology among the banking fraternity.
“Now you can go online and try it yourself and see the result, you can see the power of it. So I think that has changed everything quite a bit. Now there is more curiosity and respect for what tech can do in finance,” he added.
Bansal added that there are pitfalls, pointing to the controversy at ChatGPT developer OpenAI which led to turmoil in its management, or Google’s GenAI offering Gemini having several oops moments.
“We have to be very careful because new technologies are unproven and unknown. What if they react randomly in some other way that we haven't thought of, especially the AI models? I mean, we don't know whether LLM has a brain of its own, it has an agenda of its own, trying to manipulate us,” Bansal said.
He is also hopeful of AI improving high-end technology access for more people and companies at a much lower cost, which could have far-reaching implications.
“These models are getting better. The techniques—they are coming up with new research all the time. This allows them to become more compute efficient, use less computing power, and use less power (over time),” Bansal said, addressing concerns about GenAI and LLM being power-guzzling, thus adding to the climate change challenge.
Navi is not planning to invest or develop its LLM but sees itself as a big user as it matures and becomes cheaper. In fact, Bansal said that he is impressed with the capability of Open AI’s ChatGPT to handle Indian languages.
“I don't see any bias... The ability to handle Indian languages is excellent. If somebody's talking in Kashmiri, their bot will reply in Kashmiri,” Bansal said, indicating that even a language used by a relatively small segment of the country’s population is well covered by the ChatGPT.
Discover the latest Business News, Sensex, and Nifty updates. Obtain Personal Finance insights, tax queries, and expert opinions on Moneycontrol or download the Moneycontrol App to stay updated!
