EMI reminder calls with AI: a compliant playbook for NBFCs

Updated 2026-09-19 · 5 min read · By the Convo360 team

Why NBFCs are automating EMI reminders

NBFCs running consumer or vehicle loan books across tier-2 and tier-3 India often have a calling team too small for the volume of due-date reminders each month. Manual calling also means inconsistent tone: some agents are too soft, others come across as aggressive, both of which cause complaints.

AI voice calls can run the same, compliant script consistently across a whole due-date cohort, at whatever volume the day needs.

Inconsistency is the real risk being solved here, not just headcount. A script that's approved by compliance and legal, once, and then run identically on every call removes the variation between individual agents having a bad day, improvising a line that sounds like a threat, or forgetting to mention the payment link. That consistency matters as much for audit purposes as it does for the borrower's experience.

Pre-due vs post-due: two different tones

A pre-due call, 3-5 days before the EMI date, is a courtesy reminder: confirm the amount, confirm the date, offer the payment link. Nothing about it should sound like collections. A sample opening line: "This is a reminder call regarding your loan EMI of ₹8,200 due on the 5th. Would you like me to send the payment link to this WhatsApp number?"

A post-due call, after a missed date, stays strictly factual: state the amount overdue, ask if there's a reason for the delay, and offer to connect to an officer if the customer wants to discuss restructuring or hardship. It must never threaten, raise its voice, or imply legal action the NBFC hasn't actually initiated. A safe opening: "Your EMI of ₹8,200 due on the 5th has not been received yet. Is there a reason for the delay, or would you like the payment link resent?"

A table helps keep the two scripts distinct when briefing compliance:

Calling hours, DND and consent

All calls, AI or human, must stay within RBI/TRAI-compliant calling hours and skip numbers registered on DND for promotional-type contact where applicable. Since this is a servicing communication tied to an existing loan, most NBFCs treat it as compliant transactional contact, but you should confirm this against your own compliance policy before switching on AI calling.

The AI never asks for card numbers, OTPs or passwords on the call, only confirms amount and date, and offers the payment link.

Consent should also cover the channel itself. Most loan agreements already include a clause allowing servicing calls and messages, but it's worth confirming this explicitly covers AI-placed calls specifically, not just human agent calls, before rolling this out across your full portfolio.

Building it on Convo360

1. Connect your telephony provider, Exotel, Plivo, Twilio or your own SIP trunk, under Voice settings.
2. Create two voice flows: "EMI pre-due" and "EMI post-due", each with the AI agent's script and tone set separately.
3. Use a DTMF menu voice block so the customer can press 1 to hear the payment link sent via WhatsApp, or press 2 to speak to an officer, which uses the transfer block.
4. Trigger calls from your loan management system via the API trigger, passing due date, amount and customer number.
5. Log every call outcome to your CRM automatically.

See voice docs and compliance docs.

Mistakes that create complaints

The most common compliance mistake isn't a rogue line in the script, it's calling frequency. Placing a pre-due and a post-due call on the same customer within a short window, or calling the same overdue account daily, reads as harassment even when each individual call is polite. Build a minimum gap between calls to the same number directly into the flow, not just as a policy on paper.

A second mistake is skipping the officer-callback path when a customer sounds confused rather than clearly disputing the amount: genuine confusion (a joint loan, a recent part-payment) deserves the same human handoff as an outright dispute.

Handling call outcomes

Not every call reaches the borrower directly. A spouse or family member may answer, the number may be switched off, or the call may go to voicemail. The flow should have a distinct branch for each: a family member gets a brief, discreet message asking the borrower to call back rather than disclosing loan details to a third party, while an unreachable number gets logged for a retry the next day rather than a repeated same-day attempt.

Log every outcome, reached, not reached, disputed, transferred to officer, against the account, so your collections dashboard reflects actual contact attempts rather than just calls placed.

What it costs and what to review

AI voice calling needs the Pro plan (₹2,999/month, ex-GST) or above, plus a connected telephony account. Convo360 adds no markup on call minutes. You pay Exotel, Plivo or Twilio directly at their rate, and your AI provider (OpenAI, Claude or Gemini) at cost for the model used to run the conversation.

Review call transcripts and outcomes monthly against three things: how many calls ended in a human transfer (and why), how many payment links were actually opened after the call, and whether any transcript contains wording your compliance team wouldn't sign off on. Since every AI Reply is generated from your script and knowledge base, a wording problem is fixed once, centrally, rather than retrained into dozens of individual agents.

Example conversation

Frequently asked questions

Related reading

Available across India — see WhatsApp & AI calling automation by state and city.