Leveraging the power of AI in the lending sector

Katerina Folkman is Head Of Analytics at Clix Capital, a lending firm that uses tech to make loans simpler, faster, and more accessible. Here she talked to Thomas George and Sunil Rajguru about the company and their plans for the future.



Can you tell us something about Clix Capital and the tools that you use?

You have NBFCs (Non-Banking Financial Companies) who have large balance sheets and are not necessarily FinTechs. They always follow traditional underwriting and they might compete with big banks and for the same type of customers. Then there are small FinTechs that use truly advanced analytics and AI (Artificial Intelligence) that don't have large balance sheets. So Clix is on the NBFC side, trying to incorporate the best of the FinTech world by building internal and quite large analytical teams. We are developing our own internal algorithms and models to identify low risk high potential customers outside of usual assessments. Everything goes through our internal decision engine called DELPHI. We built it in-house over the last couple of years. That gives us a lot of flexibility. So we can have different targeting for different segments for different types of the customer. We can tweak it anytime.

Could you share your perception of AI and thoughts about maturity of the AI ecosystem in India. Is the AI hype real? Do you see AI as a game changer for your business in the next 5 years?

Currently most common application of AI in financial services in India is related to image recognition. What is commonly referred to as “AI” is in-fact, unstructured data analysis e.g. photo and video data mining for identity verification. Some firms also mention their chatbot activities as belonging to AI, even though they are usually rule-based decision engines.

What we at Clix envision as Artificial Intelligence, is a self-learning, continuously self-improving ensemble of models, taking feedback from decisions made in the past, to sharpen today’s algorithms. We are connecting our internal ML/DL decisions engines into such feedback loops, running with minimal manual intervention. The use cases will include loan underwriting and customer lifetime value management e.g. cross-sell and retention. We also experiment with deep learning models for video analytics, building Clix Visage, the tool able to recognize human emotions and sentiment from the short customer videos.

At what stage of AI adoption has your organization reached? Have you piloted with AI, done function-specific deployments, or large enterprise-wide implementations? Please walk us through some examples of the functions, BUs and geographies where you have deployed it. If you have not deployed AI, do you plan to deploy in next 2 years? Where do you plan to deploy?

Some of our AI activities are in the pilot/experiment stages and some already executed. e.g. Clix Visage for video analytics, DELPHI 2.0 self-learning decision engine, and reinforced learning models for Cross-Sell recommender engines. Some of these ideas are being further scaled up in this financial year, based on the results so far.

Can you please elaborate on your enterprise vision and strategy for AI, if it’s there? What are your primary focus areas for AI? For instance, do you focus on better customer experience, front office functioning, or better process automation, enhance the back-office, or, do you adopt AI for fraud management and detection, risk management, governance, etc.

We believe AI is critical for our unique value proposition, allowing us to differentiate our approach from traditional NBFCs and banks. We apply AI for underwriting, cross-sell, fraud management and risk management as well.

For example, AI allows a very granular look into each customer applying to the loan, assessing individual willingness and ability to pay. Instead of traditional segments of “salaried Cat A” approved applicants, Clix is able to find “granular pockets” of “low risk – high potential” customers outside of typical profiles. This helps us to compete, and be the first lender to create offers for such customers. We are now further sharpening this work based on internal behaviour data of our own customers. We are also working on developing alternative digital channels to reach different customers online. AI-based sharp underwriting will be critical to deliver on this vision.



What business objectives did you have when you decided to embark on your AI journey?

Key business objective is growth and scale, and be where other lenders cannot be, in terms of customer segments. This means finding our “Next 5 MM customers” through deep analytics-driven understanding of customer behaviour and motives. And reaching out to these customers just in time of their need.

How do you/plan to enable scalability of your AI projects to other geographies and/or functions? What would be the critical success factors?

Our AI projects are running across Clix products and geographies. Now we’re working on applying AI beyond underwriting and customer value management, for example for customer servicing. We have launched an AI enabled chatbot Maya for customer experience to this effect. The critical success factor is mining and integrating alternative data sources including digital customer footprint. Another element of success is right partnerships with alternative data providers who can also serve as distribution channels, delivering Clix offers “just in the time of need” for their digital customers.