Finance ministry reviews banks' AI tools for loan recovery: what it means for borrowers

The Finance Ministry has started consulting banks on how they are using artificial intelligence in loan collection, and is taking stock of how far this adoption has gone, according to reporting by The Economic Times. This is described as an information-gathering exercise, not a finished rule, so nothing changes today in how your bank or NBFC contacts you about a missed EMI.

For a borrower with an ongoing personal loan, home loan, gold loan or instant loan, the practical point is this: whether a human recovery agent or an AI-driven system initiates contact, the existing RBI conduct rules for loan recovery still apply. The ministry's exercise is about understanding current practice and thinking ahead — it is not, on the reporting available, an announcement of new restrictions or a new AI-recovery scheme.

That said, the direction of travel is worth tracking. Lenders have used AI and machine learning for credit scoring, fraud detection and early-warning signals on stressed accounts for several years. Collections — the calls, messages and reminders that follow a missed payment — is one of the more sensitive areas where that technology could expand next, because it touches borrowers directly and repeatedly.

Key takeaways

• The Finance Ministry is gathering inputs from banks on their AI adoption in loan collection; this is reported as an assessment, not a new policy.
• No new recovery rule has taken effect as a result of this exercise — current RBI norms on recovery agent conduct continue to govern how banks and NBFCs can contact defaulting borrowers.
• AI in collections typically means automated reminder calls, chatbots, risk-scoring of which accounts to prioritise, and predictive models flagging early delinquency — not necessarily a change in the legal recourse available to lenders.
• Unsecured retail credit — personal loans, credit cards and instant/app-based loans — is the segment most likely to see AI-driven collections first, since these portfolios are large, high-volume and already digitised end to end.
• Secured loans such as home loans and gold loans follow separate, collateral-backed recovery routes and are less likely to be the immediate focus of this review.
• Borrowers should treat this as a good moment to know their existing rights under the RBI's Fair Practices Code, regardless of whether recovery communication becomes more automated.

Why the government is looking at AI in loan recovery now

India's retail credit book has grown quickly over the past several years, and a growing share of that growth has come from unsecured, digitally originated loans — the kind that instant loan apps and buy-now-pay-later products popularised. Where loan books are large and originated digitally, collections has naturally started following the same path: banks and NBFCs already use algorithms to decide who gets a loan and at what rate, so it is a short step to also using algorithms to decide who gets called first when a payment is missed, and what tone or channel that outreach should take.

The Finance Ministry, as the owner of public sector banks and the ministry that sets the broader policy tone for the financial sector, has an interest in knowing how widespread this practice already is before any guardrails are considered. Gathering inputs from banks first — rather than legislating without data — is consistent with how most fintech-adjacent policy in India has developed over the last decade, from digital lending guidelines to the co-lending framework.

How loan recovery is regulated today

Regardless of whether a human being or a piece of software initiates contact, recovery activity in India already sits inside a well-established rulebook. Banks are expected to follow a Fair Practices Code that governs recovery conduct, and the RBI's Master Directions lay out expectations for outsourced recovery agents, including:

1. Contact only during reasonable hours, generally between 8 a.m. and 7 p.m.
2. No use of intimidation, threats or persistent harassment to recover dues.
3. A single, identifiable point of contact for a borrower's account wherever agents are used.
4. Proper authorisation and identification for anyone contacting a borrower on the lender's behalf.
5. A functioning grievance redress mechanism if a borrower feels the process was mishandled.

These rules were written with human recovery agents in mind, but their substance — reasonable hours, no harassment, proper authorisation, a grievance channel — does not stop applying just because a chatbot or an automated dialer is doing the contacting instead of a person. If AI-driven collections expand, the open policy question is less about whether these principles apply and more about how they get enforced against automated systems that can call or message at a scale no human team could match.

What AI could actually change in the collections process

It helps to separate what AI realistically does in collections from what people sometimes assume it does. AI does not create new legal powers to recover a loan — a lender still needs the same notices, the same escalation to written-off or NPA status, and, for secured loans, the same SARFAESI or civil recovery route it always did. What AI changes is speed, scale and targeting:

Overdue stage	RBI classification	Typical contact intensity today
1–30 days overdue	SMA-0	Automated reminders (SMS/app), light-touch
31–60 days overdue	SMA-1	Reminder calls, in-app nudges
61–90 days overdue	SMA-2	Recovery agent calls begin, escalation letters
90+ days overdue	NPA (Non-Performing Asset)	Formal recovery process, possible legal notices

AI mainly operates in the earlier stages of this table — deciding which SMA-0 or SMA-1 accounts are more likely to self-cure with a gentle nudge versus which need a phone call sooner, and automating the reminder itself. Where it could meaningfully change the borrower experience is in the sheer volume and immediacy of contact: a model that flags risk daily, rather than a human team reviewing accounts weekly, can generate reminders far faster after a due date is missed.

A worked example: how the timeline plays out

Consider a borrower with a ₹6 lakh personal loan and an EMI of roughly ₹13,000 a month, who misses a payment because of a temporary cash flow gap.

• Day 1–5 after due date: Under a traditional process, a missed EMI might trigger an SMS and perhaps an app notification within a day or two. Under an AI-assisted process, the same reminder could go out within hours of the auto-debit failing, and a second, more specific nudge (e.g., referencing the exact overdue amount) could follow within the week.
• Day 30 (SMA-0 to SMA-1): A human-led process typically escalates to a phone call around this point. An AI-prioritised system might have already flagged this borrower as lower-risk (say, a first missed EMI in a two-year loan history) and routed them to a soft-touch automated call rather than a live agent — or flagged them as higher-risk based on other signals and escalated faster.
• Day 90 (NPA): The account becomes a non-performing asset regardless of whether AI was used anywhere in the process. This threshold, and the accounting and reporting consequences that follow it, are set by RBI prudential norms and are not something AI adoption changes.

The arithmetic of the loan itself — principal, rate, tenure, and how a missed EMI affects amortisation — is unaffected by whether AI is involved in collections. Borrowers who want to see exactly how a missed or delayed EMI reshapes their repayment schedule can work through the numbers on an EMI calculator rather than relying on what a collections call implies.

Who this is likely to affect, and who it isn't

More likely to be affected first:

• Borrowers with unsecured personal loans, credit cards, and instant loans from digitally native lenders and NBFCs, since these books are large, high-turnover and already run on algorithmic underwriting.
• Borrowers who interact with their lender mostly through an app, since AI-driven nudges are easiest to deploy through the same channel a loan was originated on.

Less likely to see near-term change:

• Borrowers with home loans or gold loans, where recovery for genuine default follows collateral-based routes (foreclosure, gold auction) that are procedurally distinct and slower-moving than unsecured-loan collections calls.
• Borrowers who are current on their EMIs — this entire discussion is about how overdue accounts get contacted, not about performing loans.

What borrowers should do now

Since no rule has actually changed, the sensible response is preparation rather than alarm:

1. Know your lender's grievance channel before you need it — every regulated lender is required to have one, and it is the right first stop if you feel contact has crossed into harassment, whether from a person or an automated system.
2. Keep your contact details and loan app updated, since most reminder systems — AI-driven or not — rely on accurate phone numbers and app notifications to reach you early, when a soft nudge is all that's needed.
3. If you know an EMI is at risk, contact your lender proactively rather than waiting for the reminder cycle to catch up; most lenders would rather restructure a payment than escalate a file.
4. Check your eligibility for a top-up, balance transfer or restructuring option if a temporary shortfall looks like it could recur, instead of letting several EMIs slip in a row.
5. If you ever receive contact that you suspect is not from your actual lender or an authorised agent, verify independently and, if it looks like an unauthorised or fraudulent recovery attempt, it can be reported through RBI's Sachet portal.

Common mistakes and the likely outlook

The most common mistake here is over-reading a consultation as a completed policy. Government ministries routinely gather inputs from regulated entities before deciding whether any rule is even needed; plenty of these exercises conclude with guidance, some with nothing at all, and a smaller number with binding rules — usually after further consultation, not immediately. Assuming your bank's collections process has already changed because of this news is premature.

The second mistake is assuming AI in collections automatically means a harsher experience. In practice, better risk-scoring can just as easily mean fewer unnecessary calls to borrowers who are likely to self-cure a missed payment, and more consistent, compliant scripting than a poorly trained human agent might use. The outcome depends entirely on how it's implemented and supervised — which is presumably a large part of why the ministry wants a clearer picture of current practice before deciding whether to act.

Looking ahead, the realistic path is further consultation, possibly followed by guidance from the RBI on how existing Fair Practices Code obligations extend to automated and AI-assisted collections channels, rather than a wholly new regulatory regime. Borrowers should watch for RBI circulars or notifications on this topic over the coming months rather than expecting immediate change. You can track how this and other lending-policy stories develop on our news section.

Frequently asked questions

Has any new rule on AI-based loan recovery been announced?

No. Based on current reporting, the Finance Ministry is gathering inputs and assessing how banks already use AI in collections. No new recovery rule has been announced as a result of this specific exercise.

Do existing RBI recovery rules apply if a bank uses an AI chatbot to contact me?

Yes. The Fair Practices Code and recovery-agent conduct norms are about how a borrower is treated — reasonable hours, no harassment, proper identification, a grievance channel — and these principles do not depend on whether the contact came from a person or an automated system.

Which types of loans are most likely to see AI-driven collections first?

Unsecured, digitally originated credit such as personal loans, credit cards and instant/app-based loans is the most likely early focus, since these portfolios are already run end-to-end on digital and algorithmic infrastructure. Secured loans like home loans and gold loans follow separate, collateral-based recovery processes.

What should I do if I think a recovery call or message is harassing or unauthorised?

First raise it with your lender's grievance redress channel, which every RBI-regulated bank and NBFC is required to maintain. If you suspect the contact isn't from your actual lender or an authorised agent at all, it can be reported through RBI's Sachet portal for unauthorised entities.

Will this make it harder to get an EMI due date extended or a loan restructured?

There's nothing in this reporting to suggest that. If anything, better risk-scoring is more likely to help lenders identify genuinely stressed borrowers earlier and offer restructuring or revised schedules sooner, rather than making that harder to access.