DPD 1–15
Pre-delinquency nudge
Friendly reminder + pay-now via UPI deep-link. 90% resolvable end-to-end.
AI-owned
Virtual Recovery Agent
An AI collection agent that respects the RBI Fair Practice Code.
Dials the delinquent portfolio, holds compliant conversations in Hindi + regional languages, captures promise-to-pay, and hands off to your human collectors for the hard cases. 4× cheaper per meaningful conversation than a human tele-caller. Two weeks to first live batch.
Bucket strategy
Automate the routine buckets. Free your senior collectors.
The economic argument is DPD 1–30, which is 70–80% of the call volume on any collections floor. Automating even 60% of those buckets redeploys your senior collectors to the 60+ buckets where their judgment moves the recovery-rupee number.
DPD 15–30
Bucket X
Still high volume. AI handles ~70%; escalates the 30% that need judgment.
AI-owned
DPD 30–60
Bucket 1 · triage
AI qualifies (willing/unable/disputed), books a call-back with a human agent.
Hybrid
DPD 60–90
Bucket 2
Human leads. AI handles PTP follow-ups after the human contact.
Hybrid
DPD 90+
Settlement · legal
Human-only. Senior collectors + legal team. AI stays out entirely.
Human-only
Compliance
Built to survive the audit.
RBI FPC, TRAI TCCCPR 2018, and DPDP Act 2023 are non-negotiable for any BFSI collections process — in-house or outsourced, human or automated. The virtual recovery agent bakes them in by default; nothing is opt-in.
RBI FPC time windows
No calls before 8am or after 7pm. Enforced at the dialer, not the script — never a compliance workaround.
Automated-system disclosure
Every call opens with clear identification as an automated collections system. Compliance-safe by design.
DLT template registration
Every scripted line the AI may say is DLT-registered under TRAI TCCCPR 2018 before it goes live.
Immediate human handoff
"Speak to a person" triggers a warm transfer with real-time context. No compliance risk, no rework.
5–7 year recording retention
RBI-mandated evidentiary retention. Retrievable per-loan for audits + regulator queries.
DPDP data locality
All recordings + transcripts + CRM data live in India (Mumbai / Chennai / Hyderabad). Right-to-erasure workflow built in.
Portfolio math
What ~4× cheaper actually means at portfolio scale.
A single collections floor of 40 tele-callers handles roughly 8,000 – 12,000 meaningful conversations a month at bucket 1–30. Here is what a virtual recovery agent does to that unit economic.
100K loans · 15% delinquency
₹8 – 12 lakh / month
~15,000 meaningful conversations shifted from human to AI. Redeploys 5–8 senior collectors to bucket 30+.
500K loans · 12% delinquency
₹40 – 60 lakh / month
~60,000 meaningful conversations. Frees a 25-collector floor for bucket 30+ and legal escalations.
Most common tier
1M+ loans · 10% delinquency
₹80 lakh – 1.2 cr / month
Full recovery-agent stack — bucket 1–15 fully autonomous, bucket 15–30 triage-and-hand-off, bucket 30+ human-owned.
Numbers from live NBFC + PSU-bank + private-bank deployments in India. See the deployment walkthrough for the two-week rollout plan.
FAQ
Answers to what BFSI collections heads actually ask.
What is a virtual recovery agent?
A virtual recovery agent is an AI voice agent purpose-built for the collections funnel. It dials debtors, holds a compliant multi-turn conversation about the outstanding, captures payment commitments, and hands off to a human collector when the case requires judgment. Different from an IVR reminder — it's a real conversation, not a menu tree.
Is it RBI Fair Practice Code (FPC) compliant?
Yes. Baked in: no calls before 8am or after 7pm, clear identification as an automated system at call start, no threatening or harassing language, RBI-mandated 5–7 year recording retention, and immediate warm handoff to a human on caller request. Every scripted line is DLT-registered under TRAI TCCCPR 2018.
Which collections buckets does it work best on?
Pre-delinquency and buckets DPD 1–15 and 15–30 are the sweet spot — high volume, relatively straightforward negotiations, and 70–90% resolvable end-to-end. Bucket 30–60 as a triage layer (qualify then hand off). Bucket 60+ belongs to human senior collectors and legal teams.
What is the cost saving vs a human tele-caller?
~4× cheaper per meaningful conversation. Human tele-caller: ₹18–32 per successful right-party-contact all-in. Virtual recovery agent: ₹3.50–7 per successful RPC. On a portfolio of 100,000 loans with 15% delinquency, monthly savings run into ₹8–12 lakh. On 500,000 loans, ₹40–60 lakh/month.
How does it handle Hindi and regional languages?
Production-ready in Hindi, English, Tamil, Telugu, Marathi, Bengali, and Kannada. Handles Hinglish and code-mixing natively — debtors switching between English and Hindi mid-sentence stay on the same conversation flow.
What happens when the debtor asks to speak to a human?
Immediate warm transfer to a human collector with a real-time context summary — account number, loan details, what the AI already discussed, sentiment, PTP if captured, and the reason for handoff. The human agent picks up with full context; the debtor never has to repeat the story.
How long does deployment take?
Two weeks from PO to first live production batch. Week 1: DLT registration + script design + compliance sign-off. Week 2: sandbox pilot on 1,000 accounts, then staged ramp to 50,000 accounts/day. Existing dialer stack + CRM stay in place.
Where is my data stored?
India-region only — Mumbai, Chennai, or Hyderabad DC. Full DPDP Act 2023 compliance: right-to-erasure workflow, consent artefact retention per the sanction letter, and audit logs for every regulator query.
Try it free
Make your reps faster.
Spin up a free workspace, import a list, and place 50 calls today. No credit card, no sales call required.