How AI-Powered Collections Improve Loan Recovery in Vietnam
AI-powered collections improve loan recovery in Vietnam by placing consistent, empathetic reminder calls in Vietnamese at scale, lifting contact and promise-to-pay rates in early buckets, and logging every interaction. They let lenders recover more while respecting State Bank of Vietnam (SBV) consumer-finance debt-collection conduct rules and personal-data law.
Recovery pressure is real in Vietnam's consumer-finance sector. Three companies — FE Credit, HD Saison, and Home Credit — dominate roughly 80% of consumer lending (Vietnam News), and finance-company asset quality has been under strain, with sector non-performing ratios reported in double digits during the post-2022 stress. At the same time, SBV has tightened conduct around how overdue debt may be collected. AI-powered collections — automated voice agents that call, remind, negotiate simple arrangements, and escalate — help lenders recover more without breaching those rules.
Why Are Collection Costs So High for Vietnamese Lenders?
Traditional collections is labour-intensive and uneven. Field and tele-calling teams are expensive, agent scripts drift, and contact rates in early Days Past Due (DPD) buckets are low because dialling is manual and mistimed.
Three cost drivers stand out:
Manual dialling wastes capacity. Agents spend much of the day on unanswered calls and wrong numbers rather than live conversations.
Inconsistency creates compliance risk. Human agents under target pressure may stray from approved language — precisely the behaviour SBV conduct rules restrict.
Early buckets are under-served. Most recoverable value sits in the 1–30 and 31–60 DPD buckets, but teams often concentrate on hard, late accounts, missing the cheap wins.
What Do SBV Consumer-Finance Debt-Collection Rules Require?
Vietnam regulates how finance companies may pursue overdue consumer loans. Circular 43/2016/TT-NHNN, as amended by Circular 18/2019/TT-NHNN, governs consumer lending by finance companies and directs how debt reminders must be conducted — including limits on reminder frequency and permitted hours, and prohibitions on threatening, coercive, or reputation-damaging measures and on pressuring people who are not the borrower (Circular 43/2016/TT-NHNN, full text). More recently, the industry has moved toward a standardised code of conduct for debt collection to reinforce fair, professional practice (Vietnam.vn, 2025). This is an explainer, not legal advice.
The practical implications for any collections channel — human or AI:
Requirement area | What it means in practice |
|---|---|
Contact frequency & timing | Reminders limited to reasonable frequency within permitted daytime hours |
Conduct | No threats, coercion, insults, or reputation-damaging tactics |
Third parties | Do not pressure relatives, colleagues, or non-borrowers to force repayment |
Records | Keep clear records of collection contact and content |
Data protection | Process borrower data lawfully under Decree 13/2023/ND-CP and the 2025 Personal Data Protection Law |
AI collections have a structural advantage here: because every call follows an approved script within configured time windows and is fully logged, compliance is enforced by design rather than left to individual agent discipline.
How Does AI Improve Loan Recovery?
Contact rates rise. Automated agents dial thousands of accounts in parallel at the best-answer times, so more calls connect and more reach a live conversation.
Early buckets get covered. Cheap, high-frequency reminders blanket 1–30 and 31–60 DPD accounts where recovery is easiest, before balances harden. This mirrors the collections scenarios where voice AI outperforms manual teams.
Conversations stay compliant and consistent. Every call uses approved, non-coercive Vietnamese language, keeping tone empathetic — "Xin chào anh, đây là nhắc nhở về khoản trả góp sắp đến hạn của anh" ("Hello, this is a reminder about your upcoming instalment") — the same fair-practice compliance discipline regulators expect.
Promises-to-pay are captured and followed up. The agent records commitments in ₫ and schedules automated follow-ups, feeding a data-driven collections playbook rather than ad-hoc chasing.
Humans focus where they matter. Distressed, disputed, or complex accounts route to trained agents with full context, following the same outbound-collections model proven at scale.
How AI Helps
YuVoice runs early- and mid-bucket collections as automated Vietnamese voice conversations — reminding, confirming amounts in đồng, capturing promises-to-pay, and escalating hardship cases to humans with full context. Every call stays inside configured time windows, uses approved non-coercive language, and is transcribed for audit against SBV conduct rules and Vietnam's data-protection law. Handling over 25 million calls a month across financial-services deployments, it gives lenders elastic capacity at Tết and month-end peaks without adding agents — improving contact and recovery rates while keeping collections consistent, humane, and fully logged.
FAQ
Is it legal to use AI voice agents for debt collection in Vietnam? There is no ban on automated reminders. AI agents must operate within the conduct rules that apply to all collections under Circular 43/2016/TT-NHNN (as amended) — permitted hours and frequency, no coercion, and lawful data handling. This is an explainer, not legal advice.
Which DPD buckets suit AI collections best? Early buckets — 1–30 and 31–60 Days Past Due — where borrowers are most responsive to a polite reminder and recovery is cheapest. Late-stage, disputed, or hardship accounts should route to human specialists.
Can AI handle borrowers who say they cannot pay? It can acknowledge hardship, offer configured options such as a revised date, and escalate to a human for anything requiring judgment or a settlement decision — never pressuring or threatening.
How does AI stay compliant on contact timing and frequency? Calling windows and per-account frequency caps are configured centrally, so the system cannot dial outside permitted hours or over the limit — compliance is enforced by design.
Does AI collections work for finance companies as well as banks? Yes. Consumer-finance companies, which hold much of Vietnam's unsecured lending, use the same early-bucket reminder and promise-to-pay flows, tuned to their products and conduct obligations.
What proof of each interaction is kept? Every call is transcribed and time-stamped, giving a complete, auditable record of what was said — supporting both SBV oversight and internal quality assurance.
Conclusion
AI-powered collections let Vietnamese lenders recover more from the buckets that matter most, while making compliance the default rather than a hope. By scaling empathetic, script-consistent Vietnamese reminders inside SBV's conduct rules and logging everything, banks and finance companies cut cost-to-collect and protect their reputation at the same time.
Recover more, compliantly, at scale. Talk to the YuVerse team to see AI-powered collections built for Vietnam.
References
- Vietnam News — Three companies dominate consumer lending market in VN — https://vietnamnews.vn/economy/898552/three-companies-dominate-consumer-lending-market-in-vn.html
- Circular 43/2016/TT-NHNN prescribing consumer lending by finance companies (full text) — https://vanbanphapluat.co/circular-43-2016-tt-nhnn-prescribing-consumer-lending-by-finance-companies
- Vietnam.vn — Code of Conduct in Debt Collection: a guideline for banks to follow — https://www.vietnam.vn/en/bo-quy-tac-ung-xu-trong-thu-hoi-no-kim-chi-nam-de-ngan-hang-soi-chieu
- State Bank of Vietnam — https://www.sbv.gov.vn
