印度催收仍以人工为主，AI 初创公司 DPDzero 正在改变这一现状

When a borrower misses a loan payment, sending automated reminders is only the beginning of the collection process for the lender. Internal collection teams have to decide whom to contact and which channel to use. In some cases, the lender may have to decide on restructuring the loan or even pursue settlement.

Bengaluru-based DPDzero is building the technology and operations layer behind those decisions. Founded in 2022 by Ranjith BR and Ananth Shroff, it works with banks, non-banking financial companies (NBFCs), and fintechs, combining data models, automated communication, voice AI, and human agents.

The company says it now engages about 2 million borrowers a month, up from 1.8 million in August 2025, when it raised $7 million in Series A funding. Earlier this year, it received a follow-on investment from SMBC Asia Rising Fund.

Two founders, one payments rail

The founders' entrepreneurial journey began at fintech infrastructure firm Setu, which was later acquired by Pine Labs.

There, Ananth built products and partnerships, working with NPCI on Setu's Bharat Bill Payment System (BBPS). Ranjith, who had spent seven years at the data analytics firm Mammoth before joining Setu, was leading the engineering.

The BBPS work put Ranjith close to lenders and to the repayment side of lending, especially during the pandemic, when in-person collection stopped, and there was an urgent need to build digital repayment infrastructure. While the rails for repayment—UPI and BBPS among them—were largely in place, what was missing was a way to reach borrowers at scale and help lenders recover delinquent money.

"This is the time to start something on our own," Ranjith recalls thinking.

DPDzero was founded in March 2022. It later received backing from Setu co-founder Nikhil Kumar in its seed round.

From Software to Collections

The founders first thought the opportunity was in developing software for collections. Working with lenders changed their minds.

"We realised that outsourcing is a big part of this industry, so we decided to offer collections as a service instead of software as a service," the co-founder and CTO tells AIM.

That was the company's biggest pivot, and a steep one. DPDzero had to take responsibility for the collection operation itself, and neither founder had a background in collections operations. That’s a challenge the company is still living through.

The product began small, with just one engineer and a dialer layer to power voice operations. Today, DPDzero builds much of the software it needs in-house, from sales tools to engineering workflow.

The business model followed the pivot. DPDzero works on what it calls an outcome-as-a-service model. It measures recovery rates, flow-rate reduction, and portfolio resolution, and lenders can track performance through a reporting dashboard.

In effect, DPDzero gets paid only for what it recovers for the lender, across delinquency stages. That shifts the risk to DPDzero.

The Strategy

DPDzero takes on portfolios from lenders and studies them for two to three months, building models around borrower behaviour. A borrower with a clean payment history may need only a text message. One with repeated failures may get an automated voice call or be contacted by a human agent.

"It's more of a parallel communication strategy across multiple channels that we engage the customer with," Ranjith says.

The raw material is metadata: the language a borrower speaks, where they live, and how they have paid in the past. Lenders can share whether a borrower bounced payments in the previous three or six months, rather than monitoring every individual payment record. "Data is both an asset and a liability," he remarks.

Borrowers are then sorted into cohorts and labels. On the monitoring side, Ranjith points to Third Eye, DPDzero’s AI-native analytics tool, which gives operations managers and analysts a live view of portfolios over periods that can stretch from 30 to 90 days. It slices performance by loan product, delinquency bucket, geography, and channel to surface gaps and suggest the next best action.

Where AI Fits

Voice is the most visible use of AI, with Tara, its multilingual voice AI agent, built specifically for collections. It handles a borrower conversation end-to-end, understanding intent, managing objections, capturing promise-to-pay commitments, and sharing payment links.

For escalations, it hands the call to a human along with the full conversation context. Ranjith says it keeps learning from live collection outcomes.

The company built the voice stack using a mix of open-source models and models with data residency in India, and it avoids other third-party AI systems. Hindi is the company's strongest language, according to Ranjith. DPDzero also supports Kannada, Tamil, Telugu, Malayalam, Bengali, Gujarati, and Marathi.

Ranjith is candid that results are a mixed bag. The AI voice agent is sometimes unable to comprehend dialects, but in tier-1 and some tier-2 markets, he says it can perform as well as a human agent.

"It's a unified voice stack. It does three things: AI voice, simple IVR broadcasting, and then it also allows agents to speak," the co-founder notes.

AI also runs through the operation itself, from matching borrowers to agents by language and priority to helping engineers write and ship code.

Compliance, With a Machine Watching

Collections is a sensitive business in India. RBI's fair practices rules make lenders responsible for the conduct of the agents they outsource to. Since 2022, RBI has barred recovery agents from intimidation and from calling borrowers before 8 am or after 7 pm. In August 2026, it notified a consolidated framework for commercial banks, effective January 2027, that tightens those rules further, including prior intimation before making an in-person visit.

Ranjith says the company's reference point is RBI's guidance on loan recovery and recovery agents for lending institutions. DPDzero documents guidelines internally, and the rules are built into its tech stack.

The monitoring piece is Roger, the AI compliance officer. With close to 1,000 employees, most of them speaking to borrowers, Ranjith says conversations can get heated, and the company needs to flag those calls quickly so agents can be retrained. He says Roger watches calls, messages, and workflows for potential breaches of regulatory or lender-approved guardrails, raises alerts and recommends corrective action.

The company also uses API integrations to give lenders visibility into collection activity.

Shaking a Legacy Market

DPDzero's customers include ICICI Bank, HDFC Bank, Axis Bank, L&T Finance, and Moneyview. It considers digital debt collection platform Credgenics and Spocto as competitors, along with a growing crowd of voice AI companies. He sees the bigger rival in the legacy collections services.

"When we started, there were very few, but collections have become a hot market," he says.

According to IMARC Group, India’s debt collection software market is poised to grow from $220 million in 2026 to $484 million by 2034, driven by rising digital adoption, AI-driven automation, and regulatory compliance.

The company has close to 1,000 employees, about 900 in operations, with three offices in Bengaluru, two in Mumbai and a smaller satellite office in Kolkata.

The startup has raised a total of about $10.75 million in venture capital deployed toward an in-house field collection agent network, starting in Bengaluru and expanding to Maharashtra and Uttar Pradesh. It did not disclose its revenue figures.

For now, DPDzero is staying home. Going international is a question of ‘when’ rather than ‘if’, Ranjith says, but the domestic market still offers enough room. "We want every lender to think of us as the default option," he affirms.