Inside a Modern NBFC Collections Stack: What Public Data Reveals
A modern Non-Banking Financial Company (NBFC) collections stack has four visible layers: risk-based segmentation, omnichannel outreach (SMS, WhatsApp, voice), voice AI for early-bucket reminders, and smart allocation of human agents to hard accounts. All of it operates inside RBI's recovery-conduct rules — the same rules that shape every borrower contact.
This is an industry teardown built on publicly available information and RBI circulars. It is an explainer, not legal advice; refer to the official circulars and your compliance team for binding obligations.
Collections is where a lender's economics are won or lost. For an NBFC serving thin-file and first-time borrowers, recovery efficiency directly drives credit cost. Yet collections is also the most tightly conduct-regulated activity in lending — the Reserve Bank of India (RBI) holds the lender fully responsible for every contact, however it is made.
Reading public circulars, industry reporting and observable market practice, this teardown reconstructs the layers of a modern collections stack and shows where automation fits within the rulebook.
What Are the Layers of a Modern NBFC Collections Stack?
Based on publicly available information about how lenders operate, a contemporary stack functions as a sequence — each layer deciding how, when and by whom a borrower is contacted.
Layer | Function | Typical tools |
|---|---|---|
Segmentation | Score accounts by risk, Days Past Due (DPD) and willingness-to-pay | Analytics, behavioural models |
Omnichannel outreach | Reach borrowers on their preferred channel | SMS, WhatsApp, email, voice |
Voice AI | Automate early-bucket reminders and confirmations | Multilingual voice agents |
Agent allocation | Route hard/late-bucket cases to skilled humans | Dialer, workflow, field apps |
Governance | Log, monitor and audit every contact for compliance | Call analytics, QA |
The logic mirrors the query-handling funnel in banking: automation absorbs the high-volume, low-complexity work so scarce human effort concentrates where it changes outcomes. This is the backbone of the complete India collections playbook.
Why Does Segmentation Come First?
Not every overdue account needs the same treatment. A borrower who missed a payment because a salary credited late needs a gentle reminder; a genuinely stressed account needs a restructuring conversation; a high-risk account needs a skilled field agent. Segmentation decides which is which — usually by DPD bucket, exposure, past behaviour and channel responsiveness.
Getting this right is the single biggest efficiency lever. Send the wrong treatment and you either waste expensive agent time on self-curing accounts or under-serve accounts that are quietly deteriorating. Effective segmentation is why voice AI is so effective in early-bucket debt resolution — the early buckets are exactly where a polite, consistent, automated nudge resolves most cases before they harden.
Where Does Voice AI Fit — and Within Which Rules?
Voice AI sits mainly in the early buckets (0–30 and 30–60 DPD): payment reminders, due-date confirmations, promise-to-pay capture and simple query resolution — at a scale and consistency human teams cannot match. The NBFC collections use cases for voice AI and the way Indian lenders run outbound collections with voice AI both show the pattern in practice.
Crucially, this automation operates inside a strict conduct perimeter. RBI's circular RBI/2022-23/108, dated 12 August 2022 — "Outsourcing of Financial Services – Responsibilities of regulated entities employing Recovery Agents" — directs that lenders and their agents must not intimidate or harass borrowers, and must not call before 8:00 a.m. or after 7:00 p.m. for recovery of overdue loans (RBI notification; Business Standard). These build on the RBI's earlier Guidelines on Recovery Agents and the Fair Practices Code.
The regulator has been explicit that these obligations apply whether the caller is a salaried employee, a third-party agent, or an automated system. A well-built voice AI stack enforces calling-window limits, frequency caps and respectful scripting by design — turning a compliance burden into a controllable, auditable process. Reporting notes NBFCs may even incur additional cost to train agents to these standards (Business Standard, 2026) — a cost automation can help contain.
A sample early-bucket reminder stays firmly inside the rules:
"Namaste, main [NBFC] se bol raha hoon. Aapki EMI ki due date kal hai. Kya aap aaj payment kar sakte hain? Main aapko payment link bhej doon?"
How Are Human Agents Allocated in a Modern Stack?
As accounts age into later buckets or show non-response, the stack escalates from automation to people. Agent allocation matches the hardest accounts — high exposure, legal-stage, complex disputes — to the most skilled recovery staff, while automation continues handling the reminder-driven majority. This is where AI adoption in Indian lending is heading more broadly, as covered in AI adoption in NBFCs: humans for judgment, automation for volume.
How AI Helps
AI restructures the collections stack around effort-where-it-matters. Segmentation models decide treatment; voice AI handles early-bucket reminders and confirmations at scale, in the borrower's language, strictly within RBI calling windows; and analytics flag which accounts to escalate to human agents.
YuVoice is YuVerse's voice AI layer for collections — multilingual, conduct-aware voice agents that run early-bucket outreach with built-in guardrails for calling hours and frequency, and hand later-bucket cases to human teams with full context. Across the platform, voice AI already handles about 2.5 crore calls a month, giving lenders consistent, auditable, compliant contact at the volumes a growing loan book demands.
FAQ
Q1. What is a collections stack? It is the set of layers an NBFC uses to recover dues — segmentation, omnichannel outreach, voice AI, human-agent allocation and governance — working together so each account gets the right treatment at the right time.
Q2. Is it legal to use AI voice calls for collections in India? Automated calls are permitted, but they must follow RBI's recovery-conduct rules — no harassment, no calls before 8:00 a.m. or after 7:00 p.m., and full lender accountability. This is an explainer, not legal advice; confirm with your compliance team.
Q3. Where does voice AI work best in collections? In the early buckets (0–60 DPD) — reminders, due-date confirmations and promise-to-pay capture — where consistent, polite, multilingual outreach resolves most cases before they escalate.
Q4. What do RBI rules require of recovery contact? Per the RBI's August 2022 circular, lenders and agents must not intimidate or harass borrowers, must respect the 8:00 a.m.–7:00 p.m. calling window, and remain fully responsible for every contact, including automated ones.
Q5. How does segmentation improve recovery? It routes self-curing accounts to gentle automated reminders and reserves skilled human agents for genuinely stressed or high-risk cases — improving both efficiency and borrower experience.
Q6. Does automation replace human collectors? No. It handles the high-volume early-bucket work so human agents focus on complex, late-bucket and legal-stage accounts that need negotiation and judgment.
Conclusion
A modern NBFC collections stack is less a single tool than a disciplined funnel: segment first, automate the early buckets with compliant voice AI, and reserve human agents for the hard accounts — all inside RBI's conduct rules. Read the public data, and the pattern is clear: the lenders winning on recovery are the ones treating compliance and automation as the same design problem.
Ready to build a compliant, AI-driven collections stack? Talk to the YuVerse team.
References
- Reserve Bank of India — Recovery Agents notification (RBI/2022-23/108, 12 August 2022) — https://www.rbi.org.in/scripts/NotificationUser.aspx?Id=12378&Mode=0
- Reserve Bank of India — Guidelines on Recovery Agents — https://www.rbi.org.in/commonman/english/scripts/Notification.aspx?Id=787
- Reserve Bank of India — Fair Practices Code — https://www.rbi.org.in/Commonman/English/Scripts/Notification.aspx?Id=1572
- Business Standard — RBI directs recovery agents: no calls before 8am or after 7pm — https://www.business-standard.com/article/finance/rbi-directs-loan-recovery-agents-not-to-intimidate-borrowers-no-calling-before-8am-after-7pm-122081201144_1.html
- Business Standard — NBFCs may incur additional cost to train recovery agents under RBI norms — https://www.business-standard.com/industry/banking/nbfcs-additional-cost-train-outsourced-recovery-agents-rbi-norms-126021500293_1.html
