The Reserve Bank of India's Fair Practice Code for NBFCs governs how lenders can contact borrowers for collections — and violations carry serious consequences, from regulatory censure to license risk. As more NBFCs adopt voice AI for scale, understanding exactly what the RBI permits (and prohibits) is essential.
This guide covers every compliance dimension your collections AI must handle automatically.
1. Permitted Call Hours
RBI guidelines restrict collections calls to 8:00 AM to 7:00 PM local time, seven days a week. No calls — automated or human — can be placed outside this window.
For AI systems, this means the calling engine must enforce time-zone-aware scheduling for every account. A borrower in Kerala (IST) and one in Meghalaya (also IST, but different dawn/dusk) must both receive calls within the permitted window.
Agni enforcement: Every outbound campaign has call-window enforcement baked in at the infrastructure level. A campaign cannot be scheduled outside permitted hours, regardless of how the API call is made.
2. Maximum Contact Frequency
While the RBI does not specify a single national limit in all circulars, market practice (upheld in Banking Ombudsman rulings) restricts collections contact to a maximum of 2–3 contacts per day per account, across all channels combined.
Your AI system must have cross-channel contact counting — so a WhatsApp message + a voice call counts as 2 contacts toward the daily limit, not 1.
3. Consent and Disclosure at Call Initiation
Every collections call must open with a clear disclosure: who is calling, on behalf of whom, and for what purpose. Under the DPDP Act 2023, this disclosure also constitutes the consent capture — and must be recorded.
The disclosure must be in the borrower's preferred language (or the language they respond in). A borrower who responds in Marathi must receive disclosures in Marathi, not in Hindi or English.
4. Two-Year Recording Retention
The RBI requires all loan-related calls to be recorded and retained for a minimum of two years. These recordings must be:
- Tamper-proof (hash-verified)
- Accessible for audit within 48 hours of a regulator request
- Stored on India-based infrastructure (DPDP Act requirement)
- Segregated by account with clear metadata (timestamp, agent ID, outcome)
5. Harassment Prohibition
The RBI explicitly prohibits recovery agents from using intimidating, abusive, or threatening language. For voice AI, this means the AI model must be tested against the full range of emotional responses — frustration, distress, anger — and must not escalate tone in response to borrower hostility.
Agni's emotion-aware model is specifically calibrated for Indian collections contexts: firm but empathetic, persistent but never threatening. Every response is within the Fair Practice Code's permitted tone parameters.
6. Do Not Disturb Registry
The TRAI's NDNC (National Do Not Call) registry must be scrubbed before any outbound campaign. For financial institution communications, DLT-registered sender IDs are required for SMS components of multi-channel campaigns.
"We went from three RBI audit findings per year to zero in the 12 months since deploying Agni. Every compliance requirement is automated — we don't depend on individual agent judgment anymore." — Head of Compliance, NBFC (Mumbai)
7. What AI Compliance Looks Like in Practice
A fully compliant AI collections call should:
- Only dial between 8AM–7PM in the borrower's local time zone
- Open with a language-matched disclosure (DPDP consent capture)
- Identify itself as an automated system (AI, not human impersonation)
- Not contact the same account more than 2× in 24 hours
- Record the full call and store it in India, tamper-proof, for 2 years
- Immediately cease contact if the borrower invokes their right to withdraw consent
All six of these are handled automatically by Agni's infrastructure. No agent judgment is required — and no agent can override them.
The Compliance Dividend
NBFCs that switch from human-only collections to Agni-assisted collections consistently report the same unexpected benefit: their compliance posture improves, not despite the AI, but because of it. Humans are variable. AI is consistent. And in RBI compliance, consistency is everything.
