A systemically important NBFC with an AUM of approximately ₹8,400 Cr was facing mounting pressure on its asset quality. Its collections function — reliant on manual dialing teams and field agents — was too slow, too expensive, and unable to scale to the language diversity of its borrower base across five states.
The NBFC's collections challenges were structural, not operational. No amount of additional headcount would solve the underlying speed, language, and data quality gaps.
Borrower accounts crossing 30 DPD were being reached by a collections agent on average 72 hours after the bucket migration. Early contact within 24 hours of delinquency onset is the highest predictor of recovery — and the team was missing that window on 78% of cases.
Manual dialing teams were connecting on only 13 in every 100 first-touch calls. Borrowers in informal employment are frequently unreachable during business hours, requiring time-aware multi-attempt strategies that human teams cannot execute consistently.
62% of the borrower base in Telangana, Gujarat, and Maharashtra preferred to communicate in their regional language. The collections team had no Telugu or Gujarati speakers, leading to a 31% lower PTP (promise to pay) rate among non-Hindi borrowers.
An over-reliance on field agents for accounts that were contactable by phone was driving CPRC (cost per recovery contact) to ₹840 — more than 4× the sustainable level for the Two-Wheeler and Personal Loan ticket sizes involved.
Without automated time-window controls, DND scrubbing, and call recording governance, the NBFC faced RBI Fair Practices Code audit risk. Manual compliance controls depended on individual agent adherence — a significant regulatory exposure.
Fewer than 40% of collections calls had structured dispositions logged. This made bucket-level strategy decisioning, roll-rate modelling, and agent coaching effectively impossible. The data problem compounded every other problem.
ByteVox Reach was deployed as the primary pre-delinquency and early-delinquency contact layer — replacing manual dialers for EMI reminders (0 DPD), first-contact collections (1-30 DPD), and resolution-pathway identification (31-60 DPD).
Finone LMS flags accounts T-3 pre-due and on DPD crossing. Auto-push to Reach queue via webhook within 60 seconds.
5-language AI agent calls within 15 mins. Script varies by bucket: reminder tone for 0DPD, negotiation mode for 30DPD.
AI captures promise-to-pay date, partial payment intent, hardship signals. All structured and logged to CRM in real time.
Dispute cases routed to specialist agents. High-risk accounts flagged for field escalation. Payment link sent via SMS/WhatsApp.
100% call recordings, disposition codes, PTP timestamps, and consent logs — RBI FPC compliant with DPDP data handling.
The deployment followed a risk-managed phase structure to protect portfolio quality during the transition and validate AI performance before full handover.
LMS and Salesforce integration completed. Five-language voice agents configured and stress-tested with phonetic accuracy review by native speakers. RBI FPC compliance review and DND pipeline set up. Parallel pilot on 10% of Two-Wheeler bucket — ByteVox vs. manual team on identical account cohorts. ByteVox connect rate: 41% vs. manual: 13%.
ByteVox Reach took primary responsibility for 0-30 DPD outreach across Two-Wheeler and Personal Loan books. Field collections team repositioned to 60+ DPD only. Weekly metric reviews drove script and cadence optimisations. PTP capture rate improved from 18% to 39% through conversational script refinement in weeks 6 and 7.
MSME lending book onboarded with a distinct negotiation script handling working capital cycles, payment flexibility, and escalation triggers. Reach now handling 100% of first-contact across all three verticals. WhatsApp payment links driving 22% of resolutions digitally. Collections team restructured to 18 specialist negotiators from 68 manual dialers.
Performance measured across the full 90-day window against the prior-quarter baseline. All results have been independently verified by the client's collections analytics function.
"In collections, every hour matters. ByteVox got us in front of borrowers in 15 minutes instead of 72 hours — and it was speaking to them in Telugu. That is simply not something we could have built ourselves in any reasonable timeframe."
Data from this engagement showed that accounts contacted within 4 hours of DPD crossing had a 67% PTP rate vs. 28% for accounts contacted after 48 hours. The ROI of speed-to-contact dwarfs any script optimisation. Automation's core value in collections is immediacy.
Telugu and Gujarati-speaking borrowers — previously unreachable in their preferred language — showed a 41% higher payment resolution rate when contacted in their language vs. Hindi. For NBFCs with geographic diversity in their portfolio, this is a structurally significant finding.
Moving from 38% to 100% disposition data completeness enabled the analytics team to build the NBFC's first roll-rate prediction model within 45 days of deployment. The model subsequently identified 12% of the book as high-roll-risk, enabling proactive interventions that were previously impossible.
Automated DND scrubbing, quiet-hour enforcement, and per-call consent capture eliminated 100% of previously identified RBI FPC audit risks. The client's legal and compliance team independently assessed the ByteVox audit trail as fully satisfactory for regulatory submission — without any additional documentation.
Book a 30-minute tailored demo for your collections or EMI outreach workflows.
Schedule a Demo
