← all case studiesNBFC · Collections
NBFC · Mumbai, Maharashtra

40% higher EMI recovery. Zero RBI violations.

From 55% recovery to 77% — automated, compliant, in three languages.

RBI CompliantHindiMarathiCollectionsBFSI
40%higher recovery ratevs previous manual process
₹400cost per recoverydown from ₹1,600 with humans
2M+calls per monthfully automated
0RBI violationssince day one of deployment

The challenge

A Mumbai-based NBFC with a ₹400 crore book was struggling with EMI recovery rates stuck below 55%. Their 120-agent collections floor was expensive — fully loaded cost of ₹1,600 per recovered account — and couldn't scale during peak cycle weeks without quality dropping. Calls happened almost entirely in Hindi, leaving a large Marathi-speaking borrower base feeling alienated and less likely to comply. The bigger risk: sporadic call-window violations were creating real regulatory exposure.

The solution

Agni was deployed as the first-touch collections layer. Every delinquent account receives an Agni call before a human agent is ever involved.


The system enforces RBI rules automatically: calls only between 8AM–7PM, maximum two per day per number, with every call opened by a DPDP-compliant consent disclosure in the borrower's detected language. It switches fluently between Hindi, Marathi, and Hinglish mid-call based on borrower response.


For accounts showing payment intent, Agni sends a UPI payment link during the live call and tracks completion in real time. Every call is recorded and stored for 2 years on India-only servers.


Human agents now handle only the 23% of accounts that escalate — with full call context handed over automatically.

We used to dread RBI audits. Now we welcome them. Every call is documented, timed within the window, and opens with consent. We couldn't do this at scale with humans.

Head of Collections, NBFC (Mumbai)

The results

Within 90 days, EMI recovery rate climbed from 55% to 77%. Cost per recovered account dropped from ₹1,600 to ₹400 — a 75% reduction. The collections team went from 120 agents to 25 senior specialists handling genuine escalations only.

Monthly headcount savings: ₹48 lakh. Zero RBI compliance incidents since deployment.

The Marathi-language cohort — previously the worst-performing segment — became the second-best performing segment within 60 days.

Ready to see results like these?

Agni deploys in under a week. No dev team needed. Start free — no credit card required.

Explore the API.

Build voice AI into your product. Full REST API, webhooks, and India-language SDKs.

more case studies

← back to all case studies