AI has arrived in our lives — whether you like it or not. We are all knowingly or unknowingly participants in a real-life version of Turing’s imitation game — machine intelligence that is indistinguishable from human behaviour. Generative AI (ChatGPT) is
a well-known example.
But there are lesser-known AI capabilities powering the world of finance. In fact, finance has been one of the earliest test sites of AI, especially since it involves the processing of large data sets.
Investment banks first adopted AI/ML decades ago to derive and predict trading patterns. They have been also using natural language processing to read tens of thousands of pages of unstructured data in securities filings and corporate actions to figure out
where a company might be headed. When it comes to lending, the use of AI has not been as extensive as in the world of high-frequency trading, yet significant.
Let me try and put a face to AI in lending:
Meet Poorna, a small-scale farmer, who has never had access to formal credit. However, she has been consistently buying her farming supplies and selling her produce via a dedicated e-commerce platform. Soon enough she begins to see pre-approved loan offers
within her customer journey. She is taken through a few steps where she is required to take a selfie, add a few details such as PAN and Aadhaar numbers, and the money gets credited to her bank account.
A woman who was denied credit for decades could now access credit, without having to approach a bank, or even step out for that matter. This is the power of AI-driven credit-decisioning models. Having said that, there are several technology pieces involved
in enabling such seamless delivery of credit — and not all can be called AI/ML.
There are only elements of AI involved, the rest are simply the power of credit infrastructure. Although no human intervention was required to disburse the loan to Poorna, decisions such as ‘Should Poorna be given a loan?’, ‘If yes, how much?’, ‘What should
be the interest and tenure of the loan?’ are all rules-based decisions that have been automated using a business rules engine.
These decisions that are outcomes of pre-defined rules cannot necessarily be considered AI. To qualify as AI, a system must exhibit some level of learning and adapting. For this reason, decision-making systems, automation, and statistics are not AI. However,
an AI/ML model may have helped predict the likelihood of Poorna defaulting. Similarly, a trained AI/ML model may have helped with face authentication after her selfie was registered. There may be an AI/ML-trained model that monitors transactions in real-time
and sends the bank early warning signals for delinquency.
Within the lending lifecycle, AI can add value in five main areas: customer acquisition, credit decisioning, monitoring & collections, deepening relationships, and customer service. While AI definitely adds significant value, the notion that AI can do it
all without a healthy dose of real-world innovation on part of the financial services players is misguided. The space is only just beginning to heat up and there’s an immense opportunity for first-mover innovators to build solutions that truly solve the last-mile
credit access and affordability problems.