True Balance scores 100% of loan applications with AI, looks beyond credit bureaus to curb defaults

Digital lending platform True Balance now runs every loan application through its AI-driven scoring systems, while retaining human review for roughly one in 10 cases. The lender is betting that combining conventional bureau scores with a wider set of financial data can help it assess borrowers with thin or no credit histories.

The company said its AI-led underwriting has helped it approve several times as many eligible applicants as before while cutting its relative default incidence by more than half over the past year. It also said its credit-risk prediction accuracy has improved more than twofold.

“Our AI-led developments over the past year have yielded significant business impact,” Debarya Dutta, chief AI officer at True Balance, told ET AI. “By maximising data efficiency, we are now approving several times as many eligible applicants as before, all while cutting our relative default incidence by more than half.”


True Balance primarily serves customers in Tier-II and Tier-III cities and has an average loan ticket size of around Rs 12,000-13,000. A large proportion of its customers are either new to credit or have limited borrowing histories, according to Dutta.

The company has more than 100 million accumulated installed users and raised over Rs 2,000 crore in debt in FY26.

Also Read: Governance, not models, is the biggest obstacle to enterprise: Red Hat’s Vincent Caldeira

India’s thin-file credit opportunity

The push comes as India's formal credit ecosystem has expanded significantly, even as the share of new-to-credit borrowers in retail loan originations has declined.

A July report by TransUnion CIBIL showed that the share of new-to-credit consumers in retail loan originations fell from 32% in the March 2017 quarter to 13% in the March 2026 quarter. Meanwhile, consumers from semi-urban and rural regions accounted for 63% of India’s credit-active population in March 2026, up from 53% nine years earlier.

This creates an opportunity for lenders to look beyond conventional credit histories and assess borrowers who may have limited or no prior borrowing records.

The lender continues to use traditional credit bureau scores, including CIBIL, Experian and CRIF, but supplements them with alternative financial data for borrowers who may fall outside the traditional credit system.

“With explicit customer consent, we build a comprehensive profile using alternate data points, such as digital toll (FASTag) activity, investment and mutual fund holdings, and bank statement data, alongside traditional credit bureau information,” Dutta said.

He added that it currently draws on around 200 billion financial transactions for customer analysis and underwriting. These include salary credits, UPI and card spending, loan repayments and other financial activity, which can reveal patterns that may not be captured by conventional credit histories.

Dutta said True Balance has been using machine learning in underwriting for several years, but has more recently begun using technologies such as embeddings to analyse customers as sequences of financial transactions.

Humans remain in the loop

Scoring every application using AI does not mean that True Balance’s underwriting process is fully automated.

Around 10% of AI-scored applications are sent for manual checks when the system requires additional validation. The models produce probability-based assessments rather than a simple approve-or-reject decision, with cases escalated when the system lacks sufficient confidence.

“We have a strict escalation policy — an ‘escalate to human’ mechanism,” Dutta said.

He argued that alternative financial signals could help lenders assess people who have regular incomes but little conventional borrowing history.

“A large proportion of citizens in India don’t have access to credit because credit bureaus don’t find them creditworthy. They haven’t taken credit cards, for example,” Dutta said.

He cited household drivers who may receive a regular income through UPI but have little or no credit history as one category that could potentially be assessed using these broader financial trails.

Business returns and wider deployment

True Balance said its investments in AI are generating positive returns by improving customer selection, raising approval rates and reducing delinquencies. AI-led operating efficiencies have also helped lower expenses in areas such as marketing, according to chief financial officer Anupam Vasdani.

Vasdani told ET AI the company’s overall operating cost ratio had fallen from 60% to 35% over the past year, although True Balance cautioned that the reduction could not be attributed to AI alone. The company has not disclosed how much it spends on AI or quantified the returns from those investments.

The technology is also being used to prioritise collection efforts, identify borrowers who need to be contacted, generate leads, support marketing activities and respond to customer queries.

Also Read: AI may accelerate anti-competitive practices in digital markets: CCI chief

AI is also becoming part of True Balance’s expansion beyond lending into insurance and credit cards, as well as its push into Indonesia and Vietnam. Dutta said AI was helping the company expand internationally, although its operations in the two Southeast Asian markets had yet to match the performance of its India business.

The lender remains cautious about deploying AI in areas where mistakes could have a direct financial impact on customers. Its planned AI-powered customer assistant is being tested internally and with controlled groups before a wider rollout.

True Balance is also developing a separate “traceability agent” intended to allow the company to track how the system arrived at a response or decision.

For digital lenders, the broader promise is to move from evaluating only what borrowers have done within the formal credit system to understanding their wider financial behaviour. The challenge will be demonstrating that these additional signals can expand access to credit without replacing one opaque scoring system with another.

Vasdani is scheduled to speak at the Global Fintech Fest on September 11 on a panel titled “Cheaper to Run, Not Cheaper to Fund: Confronting NBFC Lending’s Unsolved Capital Cost Problem”. GFF 2026 is being held from September 8 to 11 at Jio World Centre and Trident BKC in Mumbai.