Addressing Core Challenges in Healthcare and Financial Inclusion Using AI

As CTO of Paisabazaar, Sharma is deeply involved in the day to day application of technology solutions to enable financial inclusion, and AI plays a critical role in this. Majority of customer applications received by Paisabazaar, Sharma noted, were getting rejected on the basis of eligibility criteria or customers were choosing the wrong model. This led Sharma and his team to develop an approval-based model on ML – the first in India for widespread commercial use – which utilizes customer credit data to minimize the chances of rejection. This model, built completely inhouse with local data and local talent, is a true reflection of Aatmanirbhar AI and indigenous AI in India. In addition, Paisabazaar uses AI and ML technologies like OCR, anomaly detection for digital documentation, facial recognition and fraud prevention. As the products are geared to addressing localized problems, the customer retention and turnaround improves, enabling more customers to sign up, and then allowing more local data to flow in.

Narayanan explains that in healthcare, two major problems exist – access to and quality of healthcare. Even in urban pockets, the doctor patient ratio is highly skewed and this ratio widens as one moves to interior towns and villages. Moreoever, the quality of care solely depends on the doctors available and chances of a better outcome are higher if one gets treatment from an experienced and credible doctor. These two major challenges can be addressed using AI, and that’s what mFine is creating a niche in. mFine has been building an AI-driven medical assistant that enables better doctor patient interaction, learns and adapts to user information, provides a comprehensive EHR for doctors and a seamless, tech-first experience to the patient. This tech was developed completely in house – not entirely with local data at first, but as the machine began to work with patients, the flow of data became more localized and now, the machine is attuned to local medical needs and requirements.

Identifying The Problem Statement

Sharma agrees that there are a multitude of problem statements within the BFSI space. Within the lending space, where Paisabazaar is a market leader, there are several challenges that can be effectively addressed using AI such as underwriting, identifying possible MPAs, subprime loans and more. By harnessing facial recognition, OCR, analyzing spending behaviours of customers, decoding customer data are all very effective in addressing these challenges. Other generic challenges in any fintech that can benefit from AI include KYC, customer identification and validation.

Managing Data Security In Sensitive Sectors

Its not very often that companies make a full disclosure on their data security policies but Sharma and Narayanan were very clear about this aspect during the webinar. Sharma stated that financial data is sacrosanct and customer protection is above all else, and extensive measures are taken to ensure this data is kept secure. This is not just to maintain security of the customers, but also to establish its credibility as a trusted financial services provider. A trust deficit can be a hugely detrimental factor to a startup’s growth. Narayanan adds by saying the incoming data goes through complete anonymization, where any kind of data irrelevant to the main purpose is stripped clean, before being fed into the machine. Another specific challenge that mFine encountered in early days was procuring healthcare data – something that’s hard to come by in India anyway. So this led to the team to make a basic intelligent system, slowly build it so it funnels the data based on responsible parameters and eventually, this led to more data being introduced. This is a common challenge in ML – described as the cold start problem.

Grooming Local Talent

As the applicability of AI grows, and more companies enter the fray, we’re looking at a very real challenge pertaining to adequate leadership. Moreover, AI itself is a rapidly developing field and has nuances that require constant learning. There are simply not enough AI leaders out there. Sharma is very aware of this challenge, and he believes the challenge starts at home. One of the pillars of an AI company is to have strong technical talent. India is still nowhere close to nations like China and USA when it comes to leading AI research, and companies are finding the need to hire from a narrow pool of experienced data scientists and AI researchers. One solution, Sharma believes, would be to upskill schools in AI that would lead to a larger funnel of AI experts and bring down hiring costs as well. Narayanan believes that the leaders of AI companies ideally should be no different than a typical leader of any company but the necessary caveat is they should be technologically adept, and groom a product mindset.