Traditionally, credit decisions were made using fixed rules. Lenders checked income, credit score, and repayment history, and if a borrower met the criteria, the loan was approved else rejected. This process works, but it leaves many people out and often misses important details about real financial behaviour. Today, artificial intelligence is changing this process. Instead of relying solely on records, lenders can now understand how people manage their finances in real life.
From rules to behaviour
This change in the underwriting process from rule-based evaluation to behaviour-based prediction is the primary use case for artificial intelligence. AI-led underwriting models can study patterns in bank transactions, spending habits, and cash flows.
For example, a person may not have a long credit history but may show stable monthly earnings and responsible spending patterns. Traditional models might reject this person, but AI might categorise this applicant as a stable customer with the ability and intent to pay. is becoming one of the most important factors in modern underwriting. Instead of looking at static data, lenders now look at how money moves in and out of an account. Regular stable income, consistent savings, and controlled expenses indicate a stable borrower profile.
Role of alternative data
Another important change is the use of alternative data alongside traditional information. These include utility payments, digital transactions, and even business activity for small enterprises. This helps lenders include people who are unserved and underserved in the organised credit system. While AI processes large amounts of data and identifies dynamic behaviour patterns, it assists in optimising the customer selection process. It also helps lenders to offer differential ROI based on their individual profiles.
Thus, inclusion will be one of the biggest benefits. This is especially important in India, where a large share of the population is in the informal sector and self-employed. However, lenders must be careful. AI models need to be monitored to ensure bias does not creep in. If the data used is biased, the outcomes can also be biased; hence, transparency and explainability are critical. Proper governance and regular monitoring are essential to ensure that data and rule engines remain ethical.
RBI’s AI roadmap
Based on the “Framework for Responsible and Ethical Enablement of Artificial Intelligence” (FREE-AI Report) released by the Reserve Bank of India (RBI) on August 13, 2025, the central bank has provided a comprehensive roadmap for AI adoption in credit underwriting.
The report highlights that AI is transforming underwriting by using alternative data for “new-to-credit” (NTC) customers but urges a balanced, “responsible innovation” approach to avoid risks. The key findings on AI in Underwriting (2025 Report) mention that roughly 20.8% of surveyed entities (mostly large banks & NBFCs) were using or developing AI systems, with 13.7% of these applications focusing on credit underwriting. These institutions are primarily using AI for document verification, credit risk assessment (to assess repayment ability), &automated decision-making.
The report also noted that AI models are increasingly using non-traditional data, colloquially called alternative data, such as GST filings, telecom usage, e-commerce behaviour, and utility payments, to thin-file or unserved borrowers. That said, high implementation costs, talent gaps, a lack of high-quality data, and legal uncertainties hinder wider adoption, particularly for smaller lending institutions.
Principles for responsible AI
The Core Principles for Responsible AI (The ‘Seven Sutras’), as proposed by the FREE-AI committee, must be incorporated into AI underwriting models:
- Transparency, secured and trusted AI systems.
- Monitoring to ensure that final authority and responsibility for decision-making remain with humans.
- Responsible and socially aware use of AI
- Ensuring fairness through unbiased outcomes and preventing the exclusion of borrowers.
- Accountability by the entity deploying the AI-led process.
- Explainability of the Models for decisions.
- Safety, Resilience, and Sustainability of models and processes.
The road ahead
In the future, underwriting will become more dynamic. Credit assessment will continue throughout the lifecycle. This means lenders can adjust limits and offer support based on real-time behaviour. The rise of intelligent credit is not just about technology. It is about creating a more accurate, inclusive & responsive system. Lenders who adopt these models will definitely be better prepared to serve a wider range of customers and manage risk more effectively.
Disclaimer: The information provided in this article is for informational purposes only and does not constitute financial, legal, or professional advice. While every effort has been made to ensure accuracy, readers should verify details independently and consult relevant professionals before making financial decisions. The views expressed are based on current industry trends and regulatory frameworks, which may change over time. Neither the author nor the publisher is responsible for any decisions based on this content.
Sachin Seth, Regional Managing Director, CRIF India & South Asia