For most of the last decade, credit management in India meant one thing: seeing your score. Free checks went from a novelty to a standard feature of dozens of apps, and crores of borrowers now know their number.

Utkrishta Kumar, founder and CEO of Oolka, believes that phase is largely complete.

"Visibility was the right first step, and it is largely done," he says. "Almost anyone who wants to see their score can see it. What they still cannot do is find out what is actually wrong in their credit profile, or get anyone to fix it. That is the next phase, and it is a much harder problem than showing a number."

From insight to action

Kumar spent the years before Oolka as chief business officer at Meesho, working closely with mass-aspirational users across the country. The pattern he took from that work shaped the company. 

Oolka, headquartered in Bengaluru and now used by more than one crore Indians, was built around that gap. Its AI, Dhruva, reads a user's Experian credit report in plain language and identifies the specific entries holding back their credit profile. It then acts on them: filing disputes with the credit bureau, drafting NOC, closure requests to lenders, building repayment plans, and negotiating penalty waivers and revised repayment terms.

Checking the report and getting a consultation is free. Users pay only when Dhruva takes action on their behalf.

"Advice was never the scarce thing in this market," Kumar says. "Every app will tell you to pay on time and keep your utilisation low. Very few will write the dispute, send it to the bureau, and keep following up until the account is corrected. This is what people actually needed, and for a long time it was too expensive to do for someone with a small balance."

Why an AI that acts has to remember

Acting on a credit report, Kumar argues, requires something most AI products do not have: memory.

He points to Hariom, a user who came to Oolka in December with a credit score of 598. He owed money to six lenders and had taken seventy small loans from one of them alone, one after another. Recovery agents were calling, and he was arranging his daughter's wedding. His first message was simple: "Mujhe 3 mahine mein credit score 750 chahiye… beti ki shaadi karani hai… kya aap madad kar sakte hain?"

Over the next six months he exchanged 2,695 messages with Dhruva. It drafted dispute and closure emails one lender at a time, and Hariom sent every one himself. By June, his home loan had been approved and the older, costlier loan was closed. He went on to invite Dhruva to his daughter's wedding, and promised to send the card on the same chat.

"A case like that runs for months," Kumar says. "Lenders reply by email. The bureau updates. Payments clear. None of that happens inside a chat window. If the system only remembers the conversation, then by month four it is advising him to close an account he already closed in February."

Dhruva instead maintains what Oolka's engineers call a knowledge graph: a running understanding of each user's financial life, built between conversations. It covers what has happened on every account, what Dhruva itself drafted and sent, and what each lender said back. Every remembered fact traces back to the moment it was learned.

"A bigger context window is a longer transcript. It is still a transcript," Kumar says. "Chat history remembers the conversation. Dhruva remembers the relationship."

Memory with limits

Kumar is direct about the risk that comes with that capability. "A memory that judges people is dangerous," he says. "So Dhruva reasons from evidence, what actually happened, not from labels about what kind of person someone is. Nothing reaches the live assistant without being checked against real history first. We would rather remember less and be right."

In a regulated sector, he adds, that traceability is also practical. If a lender or a user asks why the system said something months earlier, there is a dated record behind the answer.

The scale of the demand

Oolka's data suggests the demand runs well beyond the metros. According to the company's Financial Freedom Report, published in August, its users are spread across 19,065 of India's roughly 19,100 pincodes and all 36 states and union territories. They have exchanged 11.6 crore messages with Dhruva, in all 22 scheduled languages and more than 200 further languages and dialects.

The outcomes are beginning to show. More than 14,000 users now see their credit score rise in a typical week, three times the June figure.

What comes next

Today, Dhruva remembers what has happened across a user's entire history with Oolka. The next layer, Kumar says, is an AI being built to understand what is true about a user's financial situation and act on it.

"The first phase of credit management gave people a number," he says. "The next phase has to give them an outcome."

He still thinks about the reply Hariom sent when Dhruva reminded him that it was an AI: "Kaash aap sach mein insaan hote toh is dharti par swarg hota sir." If only you were really human, there would be heaven on this earth.

"What Hariom needed was someone to stay with him until the work was done," Kumar says. "That is what financial dignity looks like, and we want every Indian to have it."

Oolka is an AI credit assistant used by more than one crore Indians. More at oolka.in.