Alternate Data: Interview with Suryadip Ghoshal, Co-Founder & Chief AI Officer at Think360.ai

Alternate Data in CX: Can AI Unlock Responsible Credit Growth Beyond Metros?

Imagine this scenario.

A small business owner in a tier-3 town walks into a bank branch. He runs a thriving local supply store. Payments flow through UPI daily. His inventory turns quickly. Customers trust him.

Yet the loan officer hesitates.

There is no credit card history. No formal loan. No bureau score.

From a traditional underwriting perspective, the system sees a blank financial profile. From a real-world perspective, the borrower runs a stable business with reliable income.

This disconnect is becoming increasingly visible across India.

Recent RBI Quarterly BSR data shows a significant shift in credit expansion. Rural, semi-urban, and urban centres are driving faster credit growth than metropolitan branches. Their share of total credit has risen from 36.9% in December 2020 to 40.4% in December 2025.

The opportunity is enormous.

But so is the challenge.

Millions of borrowers across India remain “new-to-credit.” Traditional underwriting models struggle to assess their repayment ability. The result is a paradox: growing credit demand, but limited access due to insufficient data.

This is where alternate data underwriting enters the picture.

By analysing non-traditional financial signals—digital payments, cash flows, transaction patterns, business activity, and behavioural indicators—AI-driven platforms can generate a 360-degree borrower profile in minutes.

Instead of asking “What is your credit score?”, lenders can now ask a deeper question:

“What does your real financial behaviour reveal about your reliability?”

To unpack how this shift is reshaping financial services, CXQuest.com speaks with Suryadip Ghoshal, Co-Founder & Chief AI Officer, Think360.ai who works at the intersection of AI, credit analytics, and digital lending infrastructure.

In this conversation, we explore how alternate data models can expand responsible lending, reduce risk, and transform the customer experience of credit access across emerging markets.

Realization about Traditional Credit Assessment Models

Q1. What CX moment convinced you that traditional credit assessment models were no longer enough?

SG: Let me tell you about Irfan.

When we started out in 2014, Irfan was our office boy. Punctual, reliable, never gave us a reason to worry. He’d been with us a couple of years when he decided he wanted to buy a scooty, nothing fancy, just something to make the commute easier. He went to a dealership, applied for a small two-wheeler loan, and got turned down flat.

The reason? No credit score.

I remember sitting with that for a while. Here was someone I knew personally. I’d seen him show up every single day. His salary hit his account like clockwork. He had no debt, no bad habits, no red flags of any kind. And yet, to the lending system, he simply didn’t exist.

What bothered me wasn’t the lender’s decision; technically, they’d followed the rulebook. What bothered me was that the rulebook had no chapter for Irfan. No way to account for someone who was financially responsible but had never borrowed before. The system wasn’t broken. It was just built for a different India.

That’s really where the idea behind Algo360 started taking shape- not in a boardroom, but in that moment of watching someone get turned away for all the wrong reasons.

Biggest Opportunity and Biggest Risk

Q2. Why do “new-to-credit” customers represent both the biggest opportunity and the biggest risk for lenders?

SG: The opportunity isn’t just large-  it multiplies.

India has hundreds of millions of financially active adults who’ve never had a formal credit product. They earn, they spend, they pay their bills. But to a traditional lender, they’re ghosts. No bureau record means no loan, and no loan means no bureau record. It’s a loop that keeps an enormous population locked out.

But here’s the thing most lenders underestimate: getting the first product right doesn’t just win you a customer. It wins you a relationship. Look at what Bajaj Finance did. They started with small-ticket EMI cards for consumer purchases, a phone, a television, and a piece of furniture. That first touchpoint became a personal loan. Then, a two-wheeler loan. Then more. The CLTV on a well-underwritten NTC customer is extraordinary because trust, once established early, compounds.

The risk, though, is real. When there’s no credit history to lean on, it’s genuinely hard to tell the difference between someone who’s creditworthy but new to the system and someone who’s high-risk for good reason. That distinction, which traditional tools simply can’t make, is what keeps cautious lenders on the sidelines.

Alternate data closes that gap. Not by lowering the bar, but by finally being able to see the full picture of how someone actually manages money. Income patterns, payment behaviour, and spending discipline, these signals exist for almost everyone. We just haven’t always had the tools to read them properly.

Alternate Data in Digital Underwriting 

Q3. What exactly qualifies as alternate data in digital underwriting models today?

SG: Most people hear “alternate data” and think of maybe three or four signals- UPI transactions, utility bills, that sort of thing. The actual universe is much larger, and it keeps growing.

The two sources we find most powerful are SMS data and Account Aggregator flows. Your SMS inbox is, in a way, a running financial diary, salary credits, EMI reminders, BNPL alerts, and payment confirmations. It’s all there, and AI can read it in seconds. Account Aggregator takes it a step further by letting borrowers share consented, real-time bank data across institutions. Together, these two give you a remarkably complete view of someone’s financial behaviour; how they earn, what they owe, where they spend, and how consistently they honour their commitments.

Beyond these, the data ecosystem in India is genuinely impressive right now. For a salaried individual, EPFO records confirm employment history and employer details. Payroll data, ITR filings, Form 26AS, Form 16; all of these give you income verification without a single physical document. For a self-employed borrower or small business owner, GST transaction data, Udyam registration, and MCA filings tell you about business vintage, cash flows, and financial health in ways that a personal bank statement alone can’t.

Then there’s identity and address telco records, utility bills across LPG, PNG, water, electricity, government IDs like DL, Voter card, and Passport. And on the spending side, POS and payment gateway data for merchants, plus ecommerce transaction and delivery histories that reveal consumption patterns and address stability.

None of these signals is a magic bullet on its own. But stack enough of them together, and you get something far more useful than a bureau score, a live, behavioural portrait of how someone actually manages their financial life today, not two years ago.

Fragmented Financial Signals into a Reliable Borrower 

Q4. How does AI turn fragmented financial signals into a reliable borrower profile within minutes?

SG: Start with what a borrower’s financial data actually looks like before any processing happens. Hundreds of SMS messages. Thousands of transactions across accounts. Salary credits mixed in with grocery spends, EMI deductions sitting next to cab rides and subscription renewals. It’s completely unlabelled, spread across sources, and on its own, mostly unreadable.

No human underwriter could make sense of that in any reasonable timeframe. A well-built AI system does it in the time it takes to pour a cup of tea.

With Algo360, what we built was essentially a translation engine. It takes that mess of raw signals and converts them into structured lending intelligence, income estimation from transaction flows, liability detection from payment patterns, employment indicators, spending behaviour broken down by category, BNPL obligations, and early warning signs that sometimes appear in spending data well before a borrower is in visible distress.

The outputs run through two frameworks we developed. The Data Quality Score – DQS – tells the lender how much usable signal is actually available for this particular borrower and how much weight to give it. The Lending Risk Score then translates everything into a behavioural risk assessment that can sit alongside a bureau score, or substitute for one entirely when there’s no bureau history to draw on.

The borrower doesn’t see any of this. They apply. The system works. They get an answer. That invisibility is, honestly, the point – the best underwriting experience is one the customer never notices.

Behavioural or Transaction Signals

Q5. What behavioural or transaction signals often reveal more about repayment ability than traditional credit scores?

SG: A bureau score is essentially a memory. It tells you what someone did with credit before, whether they repaid on time, whether they defaulted, and how many accounts they’ve held. That’s not worthless. But it’s backwards-looking by design.

Behavioural signals work differently. They tell you what someone is doing with money right now.

Consistent income inflows over six to twelve months are about as direct a measure of repayment capacity as you can get. You don’t need a salary slip or an employment letter; if the credits are hitting the account regularly, the income is real. That’s a stronger signal than a lot of what ends up in a formal application.

Spending behaviour is more nuanced but equally telling. Someone whose discretionary spend tracks sensibly against their income, not blowing 80% of their salary on food delivery and fashion, but not living like an ascetic either, is usually demonstrating a level of financial judgment that matters for credit. The opposite pattern, where lifestyle spending keeps climbing regardless of income, is a warning sign that won’t show up anywhere in a bureau record.

Utility payments are underrated. PNG, electricity, water, these are monthly commitments that people honour or don’t, completely outside the formal credit system. A borrower who’s paid their electricity bill on time for three years has a repayment track record. We just haven’t historically counted it.

Even the texture of daily spending tells you something. A stable, predictable pattern across groceries, commute, and household expenses that signals a settled life. When those patterns turn volatile, it often precedes financial stress by weeks or months. That’s early warning information that no bureau score can give you.

Alternate Data Models Improve Lending Outcomes 

Q6. What metrics prove that alternate data models actually improve lending outcomes?

SG: The honest answer is: watch three numbers, and if all three are moving in the right direction, the model is working.

First, approval rates in the thin-file segment. The question isn’t just whether more applications are getting approved, it’s whether they’re getting approved without a corresponding jump in defaults. If you’re approving more NTC borrowers and the delinquency rate stays flat or improves, that’s the model doing exactly what it’s supposed to do. It’s not relaxing standards. It’s improving visibility.

Second, how those NTC cohorts actually perform over time. This is where conviction gets built inside lending organisations. If behavioural underwriting is identifying the right borrowers, you should see them repay at rates comparable to bureau-scored customers. When that data starts coming in, and it does, for lenders who’ve committed to this properly, the conversation shifts. Alternate data stops being a pilot and becomes infrastructure.

Third, the operational metrics. Decision time, drop-off rate, time-to-disbursal. These matter enormously for CX, and they move substantially with AI-driven underwriting. The document-chase stage, where a huge proportion of applicants quietly abandon their applications, disappears almost entirely. That’s both a better customer experience and a direct improvement to conversion.

India’s digital lending market has already run this experiment at scale. The fintech platforms that now dominate personal loan origination got there on the back of exactly this approach. The results aren’t theoretical anymore; they’re in the market share numbers.

Ideal Digital Underwriting Expertise

Q7. If you had to redesign the lending journey from scratch today, what would the ideal digital underwriting experience look like?

SG: Honestly? Most people aren’t ready for how simple it could be.

No form to fill. No documents to upload. You send a WhatsApp message  or make a voice call, in whatever language you’re comfortable with, and say you need a loan. Two or three minutes later, you have an answer.

What happens in between is entirely invisible to you. The moment you give consent, a set of API calls goes out simultaneously. Account Aggregator pulls your bank flows. The SMS parser reads your financial inbox. Telco data confirms your identity and address. EPFO or GST validates whether you’re employed or running a business. The AI assembles all of it into a risk profile. A decision gets made

The conversation you’re having on WhatsApp or over the phone is just a few clarifying questions, not a form being read to you, but an actual dialogue. And you only get asked for a document if something in the data creates a flag that needs a human to look at. For the vast majority of borrowers, that never happens

The infrastructure for this –  AA, EPFO, GST, telco APIs, bureau, all of it exists and works in India today. We’re not waiting on technology.

What we’re waiting on is for more lenders to be willing to design the journey starting from the borrower’s experience, rather than starting from their own compliance checklist and working backwards. 

Credit demand in India is clearly expanding beyond metropolitan centres

But the infrastructure to evaluate borrowers must evolve just as quickly.

Alternate data underwriting offers a powerful shift. It moves lending from static credit history to dynamic financial behaviour.

For banks and fintech firms, this transformation is not just about risk modelling.

It is about customer experience.

When lenders understand borrowers better, they can approve loans faster, reduce friction, and extend financial access responsibly.

In a country where millions remain outside formal credit systems, AI-powered underwriting could become one of the most important drivers of financial inclusion and sustainable growth.

The challenge now lies in scaling these systems responsibly—balancing innovation, transparency, and trust.

Explore more insights on emerging technologies shaping customer experience in our AI in CX hub on CXQuest.com.

And if your organisation is experimenting with AI-driven decisioning, the bigger question remains:

Are your credit models designed for yesterday’s borrowers—or tomorrow’s economy?

The post Alternate Interview with Suryadip Ghoshal, Co-Founder & Chief AI Officer at Think360.ai appeared first on CX Quest.