Discover how an SMA and NPA Classification AI Agent runs daily SMA tagging, NPA classification and provisioning under RBI norms, handles borrower-level contagion and upgrades, and prevents misstatements.
An SMA and NPA Classification AI Agent is an AI system that runs daily Special Mention Account tagging, NPA classification and provisioning under RBI's asset classification norms. It handles borrower-level contagion, upgrades and exceptions, and prevents the audit and regulatory misstatements that follow from classification errors.
Asset classification decides how much a lender must provide, what it reports to RBI and the credit bureaus and what its balance sheet shows. Small errors scale quickly across large books.
RBI's November 2021 clarification tightened the rules. Lenders must classify accounts as part of the day-end process, apply classification borrower-wise and upgrade NPAs only after all arrears are cleared.
The agent is a classification engine and control layer. It computes overdue status for every facility daily, applies the rules and produces classification, provisions and evidence.
Once an account is an NPA, it moves through substandard, doubtful and loss categories based on how long it has been an NPA and the value of security. Each category carries its own provisioning rate.
Chief financial officers, chief risk officers, credit monitoring teams, collections heads, regulatory reporting teams and auditors.
Classification errors lead to misstated profits, wrong regulatory reports and supervisory action.
RBI requires the classification date to match the actual date of slippage. For example, a loan due on March 31 that is not paid by day-end becomes SMA-0 on that date, SMA-1 on April 30, SMA-2 on May 30 and NPA on June 29 if dues remain unpaid.
If one facility slips, all of the borrower's facilities with the lender must follow. Linking facilities across systems is hard without a unified borrower view.
Partial payments do not allow upgradation. Systems that upgrade on reduced overdues create misstatements.
RBI expects classification to run through automated systems without manual intervention. Our guide on system-based asset classification explains this expectation.
Loans covered by fintech default loss guarantees must still be classified on their own performance. The Default Loss Guarantee Monitoring AI Agent tracks cover separately.
The agent calculates overdue days, applies rules, links borrowers, calculates provisions and handles exceptions.
Every day, the agent receives facility balances, due dates, receipts, limits, drawing power and collateral values from core systems.
It calculates days past due for each facility using the lender's approved method. Our guide on DPD calculation for NBFCs explains common pitfalls.
Each facility receives its SMA category for the day. Cash credit and overdraft accounts follow out-of-order rules based on limits and drawing power.
Facilities are linked by customer identifiers such as PAN and customer ID. If any facility is an NPA, the agent classifies all linked facilities as NPA.
For NPA accounts, the agent checks whether all arrears of interest and principal have been cleared across facilities before upgrading.
The agent tracks the NPA date and moves accounts into substandard, doubtful and loss categories over time.
It applies provisioning rates by category and security, using realisable values, and shows movement from the previous period. Interest on NPA accounts is reversed as required, as described in our guide on interest reversal on NPA accounts.
Technical overdues, data errors and disputed items go to authorised officers with evidence. Overrides need reasons and approvals.
The agent gives lenders confidence that their reported asset quality is right.
The agent works alongside core systems rather than replacing them.
Daily balances, dues and receipts flow in. Classification results flow back for use in servicing.
Security values support provisioning for secured portions.
Provision entries and interest reversals are posted with full reconciliation.
Classification feeds expected loss modelling, supported by the Expected Credit Loss Intelligence AI Agent.
SMA lists drive collection strategies such as bucket-wise collection strategy and field allocation through the Field Collections Route Optimization AI Agent.
Lenders measure results in accuracy, audit outcomes and early action.
The agent supports daily operations and periodic reporting.
Classification as part of day-end processing.
Classifying all facilities of a borrower consistently.
Confirming full clearance of arrears before upgrade.
Calculating provisions and movement reports for each period.
Producing classification evidence and reconciliations for statutory and concurrent auditors. Our guide on statutory audit preparation for NBFCs covers the wider audit process.
Accurate classification data for decisions such as an NPA portfolio sale to an ARC.
Flagging accounts moving quickly from SMA-0 to SMA-1 and SMA-2.
Accurate classification supports better decisions across the lender.
Daily SMA data shows where effort will prevent slippage.
SMA trends by product and channel inform policy changes.
Reliable provision forecasts support capital decisions.
Clear NPA ageing helps decide between recovery, restructuring and sale.
Boards see accurate asset quality trends.
Classification depends on data quality and policy clarity.
Weak customer linking breaks contagion rules. Deduplication must be strong.
Late posting of receipts can create false overdues. Cut-off rules must be clear.
Agricultural loans, restructured loans and revolving facilities follow special rules that need careful setup.
Every manual change must be justified, approved and logged.
RBI's move towards expected credit loss provisioning will change methods. The agent's rules must be updated as new norms take effect.
Classification will become more automated and more forward-looking.
Provisioning will move from incurred loss towards expected loss, linking classification with forward-looking models.
Classification may update continuously as payments arrive.
Supervisors increasingly analyse granular data, making accurate classification more visible.
SMA data will combine with cash-flow and behavioural signals for earlier action.
A careful parallel run builds trust in the agent's results.
Compare the agent's classification with the current system for every facility.
Group differences by cause, such as date logic, borrower linking or product rules.
Correct customer linkages, due dates and limits found during the run.
Document how each special case is handled, with compliance and audit agreement.
Move to the agent once differences are explained and approved.
Several loan types follow rules that differ from standard term loans. The agent keeps a separate rule set for each.
These become out of order when the outstanding balance stays above the sanctioned limit or drawing power, or when credits do not cover interest. The agent tracks drawing power from stock statements and flags stale statements.
Short-duration crop loans become NPA when dues remain unpaid for two crop seasons, and long-duration crop loans after one crop season. The agent uses the crop season calendar for each region.
Restructured loans follow separate classification and upgradation conditions under RBI's resolution framework. The agent tracks the specified period and performance conditions.
Where a moratorium or capitalisation is permitted, overdue calculation must follow the approved terms. The agent applies these terms only when they are documented.
Each partner must classify the borrower consistently. The agent shares classification status with partners to avoid mismatches.
Special Mention Account categories flag early stress before an account becomes NPA. For term loans, SMA-0 covers principal or interest overdue up to 30 days, SMA-1 covers 31 to 60 days and SMA-2 covers 61 to 90 days. For cash credit and overdraft accounts, SMA-1 and SMA-2 are based on the outstanding balance remaining above the limit or drawing power for 31 to 60 and 61 to 90 days.
A term loan becomes a non-performing asset when interest or principal remains overdue for more than 90 days. Other triggers apply to cash credit and overdraft accounts that stay out of order, and to agricultural loans based on crop seasons. RBI requires lenders to apply these rules on a borrower-wise basis.
RBI's November 2021 clarification requires lenders to flag borrower accounts as SMA or NPA as part of the day-end process for the relevant date, so that the classification date matches the actual running date. Running classification only at month-end can misstate dates, provisions and reporting.
Asset classification is borrower-wise, not facility-wise. If one loan of a borrower becomes an NPA, all other facilities of the same borrower with the same lender are also classified as NPA, even if they are being paid on time. The agent links all facilities of each borrower to apply this rule.
Under RBI's clarification, an NPA account can be upgraded to standard only when the entire arrears of interest and principal are paid by the borrower across all facilities. Partial payments that reduce overdues do not allow upgradation. Restructured accounts follow separate upgradation conditions.
After classification, the agent applies the lender's provisioning rates for each asset category, ageing bucket and secured or unsecured portion, using realisable security values where relevant. It shows the movement of provisions from one period to the next so that finance and audit can verify the numbers.
It identifies exceptions such as technical overdues from delayed credit, disputed charges, moratoriums, restructured accounts and system data errors. Each exception goes to an authorised officer with evidence, and any override is recorded with a reason and approval so that the audit trail remains complete.
A first deployment that connects the core banking or loan management system, collateral data and the general ledger typically takes 8 to 12 weeks, including parallel runs against existing classification to reconcile differences before go-live.
Run daily SMA tagging, borrower-level NPA classification and provisioning with full audit trails. Talk to our AI experts about an SMA and NPA Classification AI Agent.
Ahmedabad
B-714, K P Epitome, near Dav International School, Makarba, Ahmedabad, Gujarat 380051
+91 99747 29554
Mumbai
C-20, G Block, WeWork, Enam Sambhav, Bandra-Kurla Complex, Mumbai, Maharashtra 400051
+91 99747 29554
Stockholm
Bäverbäcksgränd 10 12462 Bandhagen, Stockholm, Sweden.
+46 72789 9039

Lewes
16192 Coastal Highway, Lewes, Delaware 19958, USA