A DPD 60 borrower who has lost their job requires a different conversation than a DPD 60 borrower who has the money but is avoiding the institution. Treating them the same — applying the same escalating urgency regardless of circumstance — is both commercially ineffective and, in the case of genuinely vulnerable borrowers, ethically and legally problematic. The Mid Bucket Agent AI detects hardship before it is declared and routes vulnerable borrowers to the appropriate support pathway before the collections process causes further harm.
Why hardship detection matters in mid-bucket collections
The DPD 31–90 population is not homogeneous. It contains deliberate non-payers who have the means and are testing the institution's follow-through. It contains disorganised payers who simply need more structure and urgency to prioritise the payment. And it contains genuinely distressed borrowers — people who have experienced job loss, medical emergencies, family bereavement, business failure, or domestic disruption that has materially changed their ability to pay.
For deliberate non-payers and disorganised payers, escalating collections pressure works. For genuinely distressed borrowers, the same escalating pressure produces the opposite of the desired result: it adds stress to a situation that is already overwhelmed, damages the relationship beyond repair, generates FPC complaints, and ultimately produces worse recovery outcomes than a restructure offer made at the right moment.
The challenge is identification. Genuine hardship is not always declared voluntarily — sometimes because borrowers are embarrassed, sometimes because they do not know that alternatives to payment exist, and sometimes because they have disengaged from the institution entirely. The Mid Bucket Agent AI detects hardship signals in the conversation and in the data before the borrower declares it.
The 8 hardship signals — conversational and data-based
Borrower states a hardship reason directly
The most reliable signal — and the most commonly missed in automated collections — is when the borrower explicitly names a hardship reason: job loss, medical emergency, business failure, bereavement, or divorce. The Mid Bucket AI's speech processing detects these statements and flags them immediately, pausing the standard collections script and switching to the hardship response pathway.
→ Trigger: "lost my job", "hospital", "business closed", "husband passed", "divorce" — immediate flagUPI inflows or bank credit have dropped sharply vs onboarding baseline
For borrowers who consented to Account Aggregator data at onboarding, the Mid Bucket AI checks whether the current bank statement inflow pattern has materially diverged from the onboarding baseline. A borrower whose monthly inflows have dropped from ₹85,000 to ₹12,000 in the period coinciding with their first missed payment is exhibiting a genuine income disruption signal — not a behavioural avoidance signal.
→ Trigger: Bank inflows <40% of 6-month onboarding average for 2+ consecutive monthsBureau shows simultaneous delinquency across multiple lenders
A borrower who has gone delinquent simultaneously across three or more lenders is not selectively avoiding this institution — they are in systemic financial distress. The bureau report shows this pattern: multiple accounts with similar DPD ranges, all deteriorating at the same time. This systemic delinquency pattern is a strong hardship indicator — the borrower is not managing multiple institutions strategically, they are unable to service any of their obligations.
→ Trigger: 3+ accounts at other institutions with DPD 30+ in the same calendar windowEPFO contribution history shows employment disruption
For salaried borrowers, EPFO contribution data provides a verifiable employment signal. A borrower whose employer has stopped making EPFO contributions in the months coinciding with their first missed payment has likely experienced a job change, layoff, or employer insolvency. The EPFO signal does not confirm job loss — the borrower may have switched to a non-EPFO employer — but combined with other signals, it is a significant hardship indicator.
→ Trigger: EPFO contributions ceased or changed employer in prior 2 monthsSpeech pattern analysis detects distress, crying, or emotional fragility
The Mid Bucket AI's voice analysis detects acoustic markers of emotional distress — vocal tremor, irregular breathing pattern, crying, extended pauses, or flat affect — that indicate the borrower is in a psychologically vulnerable state during the call. This signal alone does not confirm financial hardship, but it triggers an immediate shift to a more supportive conversational mode and flags the account for human agent follow-up regardless of what the borrower says about their financial situation.
→ Trigger: Acoustic distress markers detected → immediate tone shift + human flagBorrower engages, makes commitments, breaks them — repeatedly and with explanation
A deliberate non-payer eventually stops engaging. A borrower who keeps engaging, makes promises, breaks them, and explains why — "the money came in but I had a medical expense," "my daughter's school fees were due" — is exhibiting the chaotic financial management of genuine hardship, not the strategic silence of deliberate avoidance. Multiple broken PTPs with consistent engagement and explanation is a moderate hardship indicator that warrants a different conversation.
→ Trigger: 3+ broken PTPs with engagement + explanation in each instanceInsurance claim or large medical transaction in account statement
Where Account Aggregator data is available, a large medical payment or an insurance claim lodging in the period coinciding with first delinquency is a contextual hardship signal. A borrower who paid ₹1.4 lakh to a hospital in the month their EMI was first missed is not avoiding payment — they are managing a medical crisis with limited resources. This signal is contextual rather than definitive, but it significantly raises the probability weight of genuine hardship.
→ Trigger: Medical transaction >25% of monthly income in delinquency onset periodSelf-employed borrower has stopped or delayed GST filing
For self-employed borrowers, GST filing regularity is a real-time business health signal. A borrower who has filed GST consistently for three years and has now missed or delayed two consecutive filings in the period coinciding with their delinquency is likely experiencing a business disruption — not making a strategic choice to avoid loan repayment. GST filing lapse combined with income signal collapse produces a high-confidence hardship classification.
→ Trigger: 2+ consecutive GST filings missed or significantly late in delinquency onset periodThe hardship score: how signals combine into a classification
The hardship case that is not detected costs more than the one that is
A genuinely vulnerable borrower who is subjected to standard mid-bucket escalation — increasing urgency, consequence disclosure, formal recovery timeline — without any hardship intervention does not pay faster. They become more distressed, less communicative, and more likely to seek informal financial sources to make partial payments, accelerating a debt spiral. The eventual recovery cost is higher. The relationship is destroyed. And the institution may have a regulatory exposure if the borrower is subsequently classified as a vulnerable customer who was not appropriately supported. The 14.8% of mid-bucket accounts that the Mid Bucket Agent AI classifies as hardship cases are not being given leniency — they are being given the intervention that produces the highest probability of eventual recovery and the lowest probability of regulatory and reputational harm.