The agent is triggered by a lead event — web form submission, missed call, or inbound inquiry — and places an outbound call within 4 minutes of lead creation. It confirms identity via PAN and date of birth, then initiates a soft-pull bureau check in the background while conducting the income and obligation collection conversation.

During the call the agent collects: employment type (salaried, SEP, SENP), employer or business name, net monthly income, existing EMI obligations, desired loan amount and tenure, and purpose. It cross-references these against pre-approved offer data from the CBS and the bureau response to compute a real-time eligibility assessment, including FOIR on reducing-balance EMI.

If the applicant qualifies, the agent delivers the KFS — all 7 mandatory disclosures per RBI Digital Lending Guidelines September 2022 — and schedules the V-CIP session before the call ends. For borderline applicants it explores co-applicant options, loan amount reduction, or tenure extension before reaching a decline.

Production benchmarks across 60+ lending partners: 68-76% lead-to-qualified-applicant conversion vs 42-51% for manual pre-qualification teams. Average handling time: 7.2 minutes per qualified applicant vs 18-24 minutes for a manual credit officer call.

Each eligibility dimension maps to a specific data source and threshold. Credit score is retrieved via soft pull from the bureau selected by the lender — CIBIL is default; Experian, Equifax, or CRIF High Mark as fallback or for NH/NTC profiles. FOIR is computed as (sum of all monthly obligations including proposed EMI) divided by (net monthly income), with the proposed EMI calculated using the reducing-balance method on the requested loan amount and tenure.

Income thresholds vary by lender type: PSU banks typically require Rs 15,000-25,000 NMI; private sector banks require Rs 20,000-35,000; NBFCs and fintechs extend to Rs 12,000-15,000 for digital-first products. The agent applies the specific lender's policy table, not generic industry averages.

For salaried applicants the agent captures employer category: Category A (listed, government, MNC) commands the lowest rate and highest loan amount; Category B (medium enterprises) mid-tier; Category C (small or unverified) highest rate and lowest limit. For self-employed applicants, business vintage, GST registration, and ITR filing history are collected verbally and later verified against Account Aggregator data.

Age eligibility is checked against the declared date of birth confirmed at call start. The agent screens out-of-age-band applicants before initiating the bureau inquiry — avoiding a wasted hard pull fee and informing the applicant of the age restriction politely.

The agent is connected to the lender's existing bureau API credentials — Kallix holds no independent bureau membership. This protects the lender's data governance and allows the lender to maintain their existing volume-based bureau pricing agreements.

Consent collection follows a structured sequence. The agent states: the purpose of the inquiry (personal loan pre-qualification), the bureau to be accessed, that this is a soft inquiry with no score impact, and the data retention period. The applicant provides verbal consent, captured in a timestamped recording compliant with DPDP Act 2023 and CIC Act Section 20.

For NH (No History) or NTC (New to Credit) applicants — estimated at 180-200 million adults in India — the agent switches to an alternate eligibility track: bank statement analysis via Account Aggregator, or bureau-light products offered by NBFC partners. Rather than declining NTC applicants, the agent presents the AA-based path and schedules a callback for document-based underwriting.

Bureau response parsing includes: current score, number of active tradelines, DPD (Days Past Due) history for the last 24 months, hard inquiry count in the last 90 days (loan stacking signal), and any written-off or settled accounts. The agent generates a real-time qualification decision in under 12 seconds of bureau API response.

FOIR calculation requires two inputs: verified monthly income and total fixed obligations. Income is declared verbally and later verified via Account Aggregator bank statement analysis or document upload. The proposed EMI is calculated on the fly: EMI = [P x r x (1+r)^n] / [(1+r)^n - 1], where P = loan amount, r = monthly interest rate, n = tenure in months.

Obligations data comes from two sources: the bureau report (active tradeline EMIs) and the applicant's verbal declaration. The agent cross-checks declared obligations against bureau-reported EMIs and flags significant discrepancies for underwriter review rather than declining automatically.

For borderline applicants the agent applies a decision tree. First: can the loan amount be reduced to bring FOIR to 50%? The agent proposes the revised amount and asks for acceptance. Second: can tenure be extended to 60 months to reduce the EMI? The agent re-calculates and proposes. Third: is a co-applicant (spouse, parent, earning sibling) available to increase combined NMI? If yes, the agent collects co-applicant details and re-runs FOIR on combined income.

If FOIR remains above policy cap after all three remediation paths, the agent provides a structured explanation: current FOIR percentage, the cap, the shortfall amount, and a specific action the applicant can take — for example, 'Close one existing EMI of Rs X and reapply, or add a co-applicant earning Rs Y or more per month.'

Salaried income verification has a two-stage flow. During the pre-qualification call, the agent captures employer name for category classification (Category A: listed/govt/MNC; Category B: medium enterprise; Category C: small/unverified), net monthly in-hand salary, and whether salary is credited to a bank account. If the applicant consents to AA data sharing, the agent verifies the 6-month salary credit pattern in real time, reducing reliance on manual document upload.

For self-employed professionals (SEP: doctors, CAs, architects, consultants) and self-employed non-professionals (SENP: traders, manufacturers, retailers), income assessment is more nuanced. The agent collects: business vintage (years of operation), GST registration and monthly turnover, whether ITR has been filed for the last 2 years, and whether a CA-certified P&L is available. SENP income is assessed at 70-80% of net profit after lender-specific haircuts.

If the applicant cannot provide ITR (new GST registrant, cash-to-formal transition), the agent explores bank statement-based lending: AA consent for 12 months of bank statements, with the lender's credit model assessing average monthly balance, inflow regularity, and bounce rate as income proxies. This track is important for the estimated 60-70 million SENP borrowers in India who are underserved by traditional income verification.

The agent never fabricates or estimates income on behalf of the applicant. If the applicant cannot provide a credible income figure, the call is flagged for manual credit officer callback rather than proceeding with an incorrect FOIR calculation.

Pre-approved offers are generated by the lender's credit engine based on existing customer data: transaction history, repayment track record, tenure as a customer, and periodic bureau refreshes. Kallix receives the offer parameters (amount, rate, tenure) via CBS API before initiating the call.

The pre-approved call follows a condensed flow: confirm identity (PAN + date of birth), present the specific offer ('You are pre-approved for Rs 5 lakh at 11.5% for 48 months — an EMI of approximately Rs 13,000 per month'), confirm the applicant's current NMI and obligations to ensure no material change since the offer was generated, collect consent to proceed, and schedule V-CIP.

KFS delivery is mandatory even for pre-approved offers — the agent delivers all 7 disclosures before obtaining consent. This is non-negotiable under RBI Digital Lending Guidelines September 2022 regardless of whether the loan is pre-approved or freshly originated.

If the applicant requests a higher amount than the pre-approved limit, the agent explains the limit basis without disclosing internal credit scores per RBI credit information guidelines, offers the pre-approved amount as a starting point, and flags the top-up request for underwriter review post-disbursement. Decline reasons (high interest rate, no current need, competitor offer) are captured as structured CRM disposition data for re-engagement triggers.

RBI Digital Lending Guidelines September 2022 require KFS delivery to the borrower before loan execution for all digital loans. The 7 mandatory elements are non-negotiable — omitting or obscuring any element creates direct regulatory liability for the Regulated Entity (RE: bank or NBFC).

Kallix's agent handles KFS delivery in a structured verbal sequence. The agent reads each element clearly, pauses for the applicant to confirm understanding, and asks clarifying questions on the two most commonly misunderstood items — APR and prepayment terms. If the applicant asks 'What is the difference between flat rate and APR?', the agent explains: 'APR is the annual percentage rate calculated on the reducing balance — the actual interest cost to you. The EMI quoted here is Rs X per month, meaning your total interest payout over 48 months is Rs Y.'

The written KFS PDF is generated from the lender's LOS template, populated with the applicant's specific offer parameters, and dispatched via email and WhatsApp within 2 minutes of verbal delivery. The agent confirms delivery by asking the applicant to check their phone before proceeding to consent.

A 3-day cooling-off period applies to all digital loans under RBI Digital Lending Guidelines 2022: the applicant may cancel without penalty within 3 working days of disbursal. The agent explicitly mentions this right during KFS delivery, which reduces post-disbursement disputes and RBI Ombudsman complaints.

Low credit score or thin-file applicants represent 20-35% of inbound personal loan inquiries depending on channel mix. A flat decline on the first call permanently loses the customer and damages brand perception — particularly for NBFCs and new-age lenders targeting the underbanked segment.

For scores in the 600-649 range the agent checks the reason for the low score from the bureau report: a single missed payment on one tradeline (potentially explainable); a settled or written-off account (more serious); a pattern of DPD across multiple tradelines (highest risk); or high utilisation on revolving credit (often reversible). For single-event and high-utilisation cases the agent can still qualify the applicant at a higher rate or lower amount under risk-based pricing.

For NH (No History) or NTC (New to Credit) applicants the agent pivots to the AA track: requests consent to access 12 months of bank statements via Account Aggregator, explaining that the lender can assess income regularity, balance maintenance, and bill payment history as alternative credit signals. This track is particularly relevant for young professionals aged 22-28, first-time borrowers, and the estimated 180-200 million credit-invisible adults in India.

The co-applicant path is the most effective conversion lever: adding a spouse or parent with a 750+ score can move a combined application from decline to prime eligibility in most product policies. The agent collects the co-applicant's name, relationship, income, and PAN, then re-runs the eligibility check on the combined profile.

Loan stacking — simultaneously applying to multiple lenders to extract maximum credit before any single lender's FOIR calculation catches the cumulative obligation — is a significant fraud and credit risk in India's personal lending market. Bureau data shows hard inquiry spikes: a borrower with 5+ inquiries in 30 days across different lenders is an established high-risk signal.

The agent applies a tiered response. For 1-2 inquiries in 90 days: normal competitive shopping behaviour, no action. For 3-5 inquiries: direct clarifying question, response captured as structured note. For 6+: application flagged for mandatory underwriter review before advancing to document collection.

Beyond inquiry count, the agent looks for a secondary signal: declared obligations vs bureau-reported active tradeline EMIs. A Rs 15,000/month declared obligation vs Rs 28,000/month bureau-reported suggests active suppression of obligations — a fraud indicator captured for the underwriter.

Indirect dialogue is used to surface stacking intent: 'Have you received preliminary offers from other lenders in the last few weeks?' and 'What is your timeline for accessing the funds?' Urgency signals combined with high inquiry counts are weighted as a compound risk indicator.

Kallix's fraud intelligence model, trained on 80,000+ annotated loan applications, identifies 78-85% of loan stacking cases during the pre-qualification call, reducing downstream fraud losses by an estimated 2.1x compared to inquiry-count-only screening.

Co-applicant addition is the highest-conversion remediation lever in borderline cases. Lender policy typically allows a co-applicant who is a first-degree family member (spouse, parent, adult child, sibling) and an earning member. Combined income is used for FOIR calculation; both bureau reports are pulled via soft inquiry.

The agent initiates the co-applicant conversation with a clear rationale: 'Based on your current income and obligations, adding a co-applicant would strengthen your application significantly. Does your spouse or a family member earn a regular income and would be willing to be co-applicant?' If the applicant agrees, the agent collects name, relationship, net monthly income, existing EMIs, and PAN number for bureau consent.

A second soft pull on the co-applicant PAN is initiated after verbal consent. The agent informs the primary applicant: 'I will be sending an SMS to your co-applicant's registered mobile for their consent to the credit inquiry.' Verbal family consent in the same call is accepted by most lenders for pre-qualification, with formal co-applicant consent collected during document upload.

Guarantor arrangements are less common in personal lending and are not handled within the automated pre-qualification flow. If a guarantor is required by underwriting, the agent flags the application for manual credit officer handling and informs the applicant of the next steps.

The Account Aggregator ecosystem, established under RBI's 2016 NBFC-AA Master Direction and expanded to banks as Financial Information Providers in 2021, allows lenders to access an applicant's financial data with explicit, consent-based, digitally-signed consent artifacts. Kallix's agent integrates with the AA ecosystem through the lender's FIU (Financial Information User) license and API credentials.

The consent flow: the agent explains that the applicant can share bank statements digitally in 2 minutes via the AA app, without uploading physical PDFs. 'You'll receive an SMS from your bank's AA application — typically Finvu or OneMoney — asking you to approve sharing of your last 6 months of bank statements with [lender name]. This is secure, doesn't share your passwords, and you can revoke consent at any time.'

Once the applicant consents via the AA app, the bank (as FIP) transmits encrypted financial data to the lender's FIU within 60-120 seconds. The agent waits on the call and confirms: 'Your bank statements have been shared. Your salary credits of Rs X over the last 6 months have been verified.'

AA data quality varies by FIP: PSU banks (SBI, Bank of Baroda, PNB) typically have complete 12-month transaction data; some regional cooperative banks and newer small finance banks may have partial data. Where AA data is incomplete, the agent requests targeted document upload only for the missing period. For lenders not yet integrated with AA, the agent routes to NSDL DigiLocker salary slip download, net banking statement export, or portal upload in priority order.

V-CIP (Video Customer Identification Process) is the RBI-approved remote KYC method introduced via circular DBOD.AML.BC.No.81/14.01.001/2019-20 (January 2020). It replaces in-branch KYC for digital loans and requires: a live video call with the customer, capture of Aadhaar/PAN document in the customer's hand, a live photograph, and an audio-video recording retained for minimum 6 months.

The AI agent's role in V-CIP is scheduling coordination only. The agent presents available slots (typically 9 AM to 7 PM across 7 days a week for most lenders' V-CIP teams), confirms the applicant's preferred time, and registers the appointment in the lender's scheduling system. The calendar invite includes the V-CIP meeting link, a document checklist (original PAN, original Aadhaar, latest bank statement if not AA-verified), device requirements (smartphone with camera, stable internet), and instructions to be in a well-lit location.

V-CIP no-show is a significant drop-off point in the personal loan funnel. Kallix sends 3 reminders: 24 hours before (SMS + email), 2 hours before (WhatsApp + call), and 15 minutes before (WhatsApp). This reminder cadence reduces V-CIP no-show rates from 38-44% to 18-24%, increasing funnel conversion at this stage by 20-26 percentage points.

If an applicant misses their V-CIP slot, the agent calls within 30 minutes to reschedule. Applications that miss two consecutive slots are flagged for manual credit officer review before a third slot is offered, as this pattern correlates with elevated fraud risk.

Interest rate communication is one of the most compliance-sensitive and conversion-sensitive steps in the pre-qualification call. Misrepresented or unclear rates are the leading cause of post-disbursement disputes, RBI Ombudsman complaints, and regulator scrutiny of digital lenders.

Kallix's agent is configured with the lender's risk-based pricing model: prime applicants (CIBIL 750+, FOIR below 40%, Category A employer) receive the lowest rate band; near-prime applicants receive the higher band. The agent communicates the specific rate for this applicant's profile — not a generic range — and explains what factors drove the rate: 'Based on your credit profile, you qualify for our 12.5% per annum rate. If your CIBIL score improves above 750, the rate could reduce to 11.5%.'

Processing fee is disclosed as both percentage and absolute rupee amount with GST: 'Processing fee: 1.5% of Rs 3 lakh = Rs 4,500 plus 18% GST = Rs 5,310.' This prevents the common applicant confusion about the all-in loan cost. Prepayment charges are disclosed by category: many banks offer zero prepayment after 12-24 months, but charge 2-4% on outstanding principal during the lock-in period.

Total cost of credit (TCC) — including processing fee, GST, and total interest — is disclosed as a lump sum before consent, giving the applicant a single number to compare across lenders. This is part of KFS delivery and non-negotiable per RBI Digital Lending Guidelines 2022.

Rejection handling is a compliance obligation, not just a customer experience consideration. RBI's Fair Practices Code (Master Circular for NBFCs) requires: all loan applications disposed of within a reasonable period, applicants notified of rejection with the reason, and rejection communications in writing.

Kallix's agent provides a tiered rejection response based on the primary decline factor. For credit score decline: 'Your CIBIL score is currently below our minimum threshold. I'll send you a report with steps to improve your score — reducing credit card utilisation is typically the fastest improvement. You can reapply in 90 days.' For FOIR decline: 'Your EMI obligations relative to income exceed our cap. Closing one existing EMI would bring you within range, or adding a co-applicant earning Rs X per month would qualify the joint application today.'

The agent never discloses the exact cut-off score or FOIR threshold — this is proprietary lender policy — but provides enough information for meaningful corrective action. Under-informing leads to repeat applications that continue to fail; over-disclosing creates model gaming.

For applicants on the negative list (willful defaulters, fraud list, CIBIL CAIS settled/written-off), the agent follows a sensitive protocol: it does not explain the specific list on the call, states the application cannot be processed at this time, and provides the RBI Ombudsman number and SACHET portal reference for applicants who believe the listing is erroneous.

All rejection dispositions are written to the LOS and CRM with structured reason codes, enabling analytics on decline patterns by channel, product, and segment.

Top-up loans are the highest-conversion product conversation in retail lending: the borrower is a known customer with a verified repayment track record, credit risk is quantified, and marginal acquisition cost is near zero. Kallix's agent identifies top-up eligible borrowers from CBS data before calling and leads with the offer: 'You've completed 18 of your 48 EMIs without delays. You're eligible for a top-up of up to Rs 2 lakh at your existing rate.'

Top-up FOIR calculation adds only the incremental EMI — not the original loan balance — to existing obligations. This is the correct method per most lender policies: the existing EMI is already factored into the current FOIR; the top-up assessment tests whether the incremental amount is affordable.

Balance transfer is more complex. The agent must: verify the outstanding principal (via AA or verbal declaration), confirm remaining tenure, compute the new EMI at the target lender's rate, verify that the borrower saves money net of BT processing fee, and confirm FOIR on the new EMI. If the BT saves less than 1.5-2% on effective rate after fees, the agent discloses the net savings clearly and lets the applicant decide.

Debt consolidation — combining multiple personal loans, credit card outstanding, and consumer EMIs into a single personal loan — follows BT methodology but with multiple source obligations. The agent walks through each obligation, confirms the outstanding balance, and calculates whether consolidation reduces total monthly outflow and total interest payout. In production, top-up calls convert at 45-58% — the highest of any pre-qualification call type.

SENP income assessment is fundamentally different from salaried income. Declared monthly income cannot be taken at face value — business income is gross revenue, not net profit, and many SENP applicants understate income on ITR for tax optimisation. The agent resolves this through a structured verbal income audit.

The agent asks: (1) What is your line of business and GST-registered turnover for the last financial year? (2) What is your net profit after business expenses as reported in your last ITR? (3) Do you have a CA-certified P&L for the last 2 years? (4) Does your business have seasonal peaks that affect monthly cash flow? This structured collection gives the underwriter a consistent, comparable data set.

For seasonal businesses (retail during Diwali, agriculture-linked, construction), the agent applies a 12-month bank statement analysis via AA: the average of the 6 highest income months, discounted by 60% of the seasonal peak, provides a conservative income base. This method is used by progressive NBFCs and is increasingly accepted by private sector banks for bureau-verified SENP applicants.

GST return data (GSTR-1 and GSTR-3B) is an emerging income verification source: monthly GST filing shows revenue regularity and business scale. Kallix's agent seeks consent to access GST data via the government's sandbox API where lenders use GST-based underwriting, which can replace CA P&L as the primary income verification document.

SENP applicants without ITR filing are guided to the AA bank-statement track and advised to begin ITR filing to improve future eligibility — this advisory is captured in the CRM for follow-up in 3-6 months.

Personal loans in India are legally general-purpose unsecured credit — the borrower is not required to prove expenditure post-disbursement. However, purpose affects: documentation required (home renovation invoices, medical bills, wedding quotes for higher loan amounts), eligibility for specific lender products (wedding loans with florist tie-ups, travel loans with airline integrations), and risk model calibration (some lenders apply higher risk weights to business working capital vs medical).

Kallix's agent collects loan purpose early in the call, as it affects the script path. Medical emergency calls follow the fastest flow: after basic eligibility confirmation, the agent offers same-day V-CIP and initiates expedited disbursement where lender infrastructure supports it. The agent confirms: 'For medical purposes, [lender name] offers expedited processing — subject to eligibility, funds can be disbursed within 4-6 hours of V-CIP completion.'

For business working capital, the agent flags a product routing issue: if declared purpose is business working capital or expansion, RBI Digital Lending Guidelines and several lender policies require a separate MSME loan assessment rather than a personal loan. The agent redirects: 'For business funding, [lender name] has an MSME loan with a higher eligibility amount and lower rate for GST-registered businesses. Would you like me to check your eligibility on that track instead?'

Debt consolidation is handled with a specific calculation: the agent adds up all declared EMIs, computes the consolidated personal loan EMI at the proposed amount and rate, and presents the net monthly saving and total interest saving. This is a strong conversion tool for high-FOIR applicants near the policy cap.

NACH (National Automated Clearing House) mandate setup is a critical step in the personal loan disbursement funnel. Without an active NACH mandate, the lender cannot auto-debit EMIs, which delays disbursement and increases post-disbursement operational overhead. Initiating the NACH conversation during pre-qualification — rather than as a separate post-V-CIP step — significantly improves same-session completion rates.

The agent's e-NACH flow works as follows: after V-CIP scheduling is confirmed, the agent transitions to NACH setup: 'To complete your application, we need to set up the auto-debit mandate for your EMI payments. This takes about 2-3 minutes and can be completed via your net banking or UPI. Would you like to set this up now or receive the link later?' For applicants who agree, the agent collects the repayment account details verbally and dispatches the e-NACH link via SMS.

E-NACH through NPCI's mandate registry has two authentication paths: net banking (bank portal login, mandate approval) and Aadhaar e-sign (Aadhaar OTP, biometric signature). For applicants without net banking access, the Aadhaar e-sign path is the recommended alternative.

If the applicant declines e-NACH setup during the call, the agent schedules a follow-up: 'I'll send you the NACH setup link along with your V-CIP reminder. Please complete it before your video KYC appointment — it takes under 3 minutes.' The LOS flags applications with incomplete NACH for credit officer follow-up before disbursement approval.

In production, same-session e-NACH initiation (during or immediately after the pre-qualification call) produces a 62-72% completion rate within 24 hours, vs 38-48% when e-NACH is first introduced post-V-CIP.

The integration architecture involves four system layers: (1) telephony (Kallix's managed Twilio or Exotel infrastructure), (2) AI reasoning and conversation layer (Kallix platform), (3) data APIs (bureau, CBS, LOS, Account Aggregator), and (4) CRM and communication (Salesforce, LeadSquared, Zoho, or lender-proprietary).

Bureau API integration uses the lender's existing CIBIL/Experian/Equifax/CRIF membership and API keys — Kallix holds no independent bureau access. This protects the lender's data governance and preserves their existing volume-based bureau pricing.

LOS integration is the critical path for eliminating manual re-entry. After a successful pre-qualification call, Kallix writes a structured application record to the LOS: applicant PAN, Aadhaar (masked), name, mobile, email, employment type, employer name, NMI, existing obligations, declared purpose, requested amount and tenure, FOIR calculated, credit score band (not raw score), AA consent artifact if obtained, and V-CIP appointment details. The credit officer's queue shows a pre-populated application rather than a blank form.

CRM integration handles lead lifecycle: on qualification, lead status advances from 'Pre-Qualification Pending' to 'Qualified — Awaiting V-CIP'; on V-CIP completion to 'Underwriting'; on rejection to 'Declined' with structured reason code and re-engagement trigger (90 days for credit score decline, 30 days for FOIR decline post-EMI closure). This eliminates the manual CRM update burden that typically consumes 15-20% of a credit officer's time.

For Salesforce Financial Services Cloud, Kallix provides a native integration package with pre-built field mappings for personal loan origination objects. For LeadSquared (common among NBFC and FinTech lenders), the integration uses the Lead API to write structured pre-qualification outcomes as lead activity records.

Compliance architecture for an AI voice agent in personal lending is more complex than most financial verticals because a single pre-qualification call touches six separate regulatory frameworks, each with its own consent and disclosure requirements.

RBI Digital Lending Guidelines September 2022 apply to all digital loans by Regulated Entities. Key obligations: KFS delivery before execution, loan disbursement only to the borrower's bank account, a 3-day cooling-off period, mandatory LSP disclosure ('This call is conducted by Kallix on behalf of [lender name], a Reserve Bank of India-regulated entity'), and FLDG cap (5% of digital loan portfolio for credit enhancement arrangements).

DPDP Act 2023 governs the processing of personal data including financial information, classified as sensitive personal data. The agent must: state the purpose of data collection at the start of the call, obtain explicit consent for bureau inquiry, limit collection to what is necessary for pre-qualification, and store call recordings for the minimum required period (typically 6 months for credit calls, longer if a dispute is raised).

TRAI TCCCPR 2018 classifies personal loan pre-qualification outbound calls as 'transactional' for existing customers with pre-approved offers (DND-exempt), or 'commercial' for new leads from marketing lists (NDNC check required before dialling).

RBI Account Aggregator consent artifact requirements are specific: consent must be purpose-specific, time-limited, and revocable. Kallix does not store bureau data or AA-retrieved financial data — both are processed in real time and written to the lender's LOS under the lender's data governance policy.

Personal loan pre-qualification is a volume-driven funnel: lenders with 500-5,000 leads per day cannot afford the capacity or consistency constraints of manual pre-qualification teams. Kallix's metrics framework measures AI agent impact at every funnel stage.

Lead-to-qualified rate measures the percentage of leads that exit the pre-qualification call as qualified applicants with V-CIP scheduled. The 68-76% production benchmark includes co-applicant addition, FOIR remediation, and AA-based income verification — these three interventions collectively rescue 25-30% of initially borderline leads. Without them, the conversion rate would be 45-55%.

Average handling time (AHT) of 7.2 minutes includes identity verification, bureau check wait time, FOIR calculation, KFS delivery, and V-CIP scheduling. The manual credit officer equivalent (18-24 minutes) includes time to log into CBS, run the bureau inquiry, input data to LOS, and compose the KFS email. One Kallix AI agent instance handles the equivalent throughput of 6-8 manual credit officers at peak.

V-CIP show rate is the most operationally sensitive KPI. A 20-percentage-point improvement (from 58% to 78%) translates directly to 20% more disbursals from the same qualified applicant pool. Kallix's reminder cadence is the primary driver, and lenders with limited SMS/WhatsApp communication infrastructure see the largest gains.

Time-to-disbursal for prime applicants with AA-verified income: T+3 to T+6 days via AI-assisted funnel vs T+7 to T+12 days via manual pre-qual + manual LOS entry.

ROI for personal loan pre-qualification automation has three components: direct cost saving, revenue uplift from higher conversion, and risk reduction from better fraud detection.

Direct cost saving: at 1,000 leads per day, a manual pre-qualification team requires 60-80 credit officers at Rs 30,000-50,000 per month salary plus benefits, training, and attrition overhead. Total annual cost: Rs 3.5-6.0 crore. Kallix's per-qualified-applicant model for the same volume: Rs 1.8-3.2 crore annually. Direct saving: Rs 1.3-2.8 crore.

Revenue uplift: the 25-30 percentage point improvement in lead-to-qualified-applicant conversion generates 27% more qualified applicants from the same lead pool. For a lender disbursing an average Rs 3.5 lakh personal loan with a 2.5% net interest margin over 36 months, each additional qualified applicant who disburses is worth approximately Rs 8,750 in annual net interest revenue. At 1,000 leads per day with a 12% disbursal rate, each percentage point improvement in qualification rate adds approximately Rs 32 lakh annually in net interest revenue.

Fraud reduction: loan stacking detection (78-85% catch rate) and income mismatch flagging reduce post-disbursal NPAs from fraudulent applications. For a lender with a 3.2% personal loan NPA rate, a 12-18% reduction in fraud-driven NPAs is worth Rs 8-15 lakh annually per Rs 10 crore of personal loan portfolio disbursed through the AI-assisted funnel.

Break-even is typically at 8-12 weeks post-deployment, with full production metrics stabilising by week 10-14.

Personal loan pre-qualification deployment is more involved than simpler voice agent use cases because it requires ingesting lender-specific policy tables, integrating with live bureau APIs, and passing compliance review for KFS delivery and consent flows.

Week 1-2 (Policy Ingestion and API Integration): Kallix ingests the lender's underwriting policy as a structured rule set: minimum credit score by segment, FOIR caps by employment type, minimum NMI by product, employer category matrix, maximum loan amount by income band, and risk-based pricing table. Bureau API integration is tested with sandbox credentials before going live. LOS and CRM API connections are established and data flow validated end-to-end.

Week 3-4 (Script Tuning and Compliance Review): The call script is reviewed by the lender's compliance team for KFS accuracy (every number must match the LOS template), consent language per DPDP and CIC Act requirements, rejection language per Fair Practices Code, and TRAI classification of call types. The 3-day cooling-off disclosure is reviewed by legal. Kallix provides a compliance checklist with regulatory citations for each disclosure element.

Week 5-6 (Pilot and Tuning): 500 live pre-qualification calls with full monitoring. Common tuning requirements: income range questions for self-employed applicants, FOIR remediation dialogue for borderline cases, V-CIP scheduling flow wording. Lender credit officer reviews 50-100 calls and provides structured feedback. Sign-off by compliance and credit head before full production launch.

Deployment requires no changes to the lender's existing infrastructure. Ongoing support includes weekly performance reports, monthly model tuning, and quarterly compliance review.

Web form-based pre-qualification has a fundamental engagement problem: the form has no agency to respond to confusion, push back on incorrect inputs, or explain why a field matters. When an applicant sees a 20-field form asking for employer category, FOIR, gross vs net income, and PAN, they either guess (producing low-quality data) or abandon (zero conversion).

A McKinsey analysis of retail lending funnels found that 62-68% of personal loan applicants who start an online eligibility form do not complete it. Of those who complete it, 30-40% enter incorrect or estimated data (particularly self-employed income), creating inaccurate FOIR calculations that either over-approve risk or under-approve eligible applicants.

AI voice agents solve both problems simultaneously. The agent collects the same data points as the web form through conversation: the applicant answers a question rather than filling a field. When the applicant doesn't know their FOIR, the agent explains and walks through the calculation. When they're unsure about employer category, the agent asks for the company name and classifies it. When the bureau score is borderline, the agent doesn't decline — it explores alternatives.

For mobile-first leads (70-80% of India's personal loan applicants access via mobile), voice is a significantly lower-friction channel than a 20-field mobile form. The agent-to-applicant session lasts 7 minutes on average — equivalent to a natural phone conversation, not a form-filling exercise.

Data quality advantage: income figures collected verbally with clarifying questions (gross vs net, regular vs variable) produce 35-40% more accurate FOIR calculations than web form entries, reducing the approval-to-disbursal drop-off when underwriting re-checks the form data.

Language handling in personal loan pre-qualification is more complex than in most financial voice agent use cases because the vocabulary is technical (FOIR, APR, reducing balance, co-applicant) and applicants have widely varying financial literacy. A single-language agent that speaks only formal Hindi or formal English will produce high drop-off rates among applicants who naturally code-switch.

Kallix's multilingual model is trained on 200,000+ annotated BFSI call transcripts in Hindi, English, and Hinglish. The agent detects the applicant's language preference within the first 2 exchanges and matches it for the rest of the call. Code-switching within a sentence is handled naturally — the agent does not ask 'Which language do you prefer?' but simply adapts.

Technical terms are handled differently in Hindi vs English. FOIR has no standard Hindi equivalent — the agent explains it in plain language: 'Aapki maasik income mein se kitna hissa existing EMIs pe ja raha hai, aur naya loan lene ke baad woh ratio policy ki limit se kam hona chahiye.' APR is explained as the effective interest rate calculated on the reducing balance — the actual cost of the loan in rupees per year.

For lenders serving specific geographies — a Tamil Nadu-focused NBFC, a Gujarat cooperative bank — Kallix deploys the regional language variant with lender-specific terminology calibration during Week 3-4. Regional language models are validated with native-speaking credit officers from the lender's team before production launch.

The AI vs manual comparison should be framed as a specialisation decision, not a replacement decision. AI handles the high-volume, protocol-driven portion of the pre-qualification funnel; manual officers handle complex, judgment-intensive edge cases requiring senior credit expertise.

Throughput is the most dramatic advantage: a single Kallix AI instance can run 100 simultaneous pre-qualification calls, 24x7, with no degradation in call quality. A 100-officer manual team can run 100 simultaneous calls during business hours, but with significant quality variance: the top 20% of officers convert 55-65% of leads; the bottom 20% convert 30-38%. The AI agent delivers consistent 68-76% conversion across all call slots, including early morning, late evening, and weekend calls when manual team performance typically drops 15-20%.

Policy consistency is the second major advantage. Kallix's production audits across 60+ lending partners show manual credit officers deviate from stated FOIR policy in 15-25% of calls — approving borderline cases based on rapport or declining based on perceived socioeconomic signals, or failing to deliver the full KFS. The AI agent is 100% policy-compliant on every call with a complete audit trail.

Manual teams retain advantage in three categories: (1) complex SENP cases where income verification requires significant professional judgement; (2) estate or inheritance-backed applications where credit history doesn't apply; (3) HNI relationship management where the personal banker relationship is a core retention lever. These cases represent 8-15% of pre-qualification call volume for most lenders.

The optimal model is hybrid: AI handles 85-92% of calls with automatic escalation to senior credit officers for flagged complex cases. This allows a 70-80% reduction in pre-qualification team size while improving both throughput and consistency.

Personal loan foreclosure generates high inbound call volume — borrowers want to know the exact payoff amount, pay it, and receive their NOC without branch visits. The process involves three steps that Kallix automates end-to-end.

Step 1 — Foreclosure amount calculation: Kallix pulls the live outstanding principal from the LMS and applies the applicable foreclosure charge. RBI permits foreclosure charges on fixed-rate personal loans (typically 2–4% of outstanding principal); floating-rate personal loans to individuals have no foreclosure charge under RBI guidelines.

Step 2 — Payment facilitation: the agent dispatches a pre-filled UPI payment link for amounts up to Rs 1 lakh (UPI limit), or provides the NEFT/RTGS beneficiary details for higher amounts. The agent confirms the payment reference number and sets the LMS to monitor for incoming credit.

Step 3 — NOC generation: once the LMS confirms full credit, the Kallix agent triggers the NOC generation workflow. The NOC is generated by the lender's system (not the AI agent), reviewed by a credit officer, and dispatched to the borrower's registered email and WhatsApp within 2–3 business days.

Kallix also handles CIBIL closure update advisory: after NOC receipt, the borrower is informed that the bureau update takes 30–45 days and they should check their credit report after that period to confirm the loan is marked closed.

Lenders using Kallix for foreclosure advisory see branch walk-in volume for NOC requests drop by 65%.

RBI Digital Lending Guidelines (DLG) 2022 introduced several borrower protection measures that directly affect how personal loan pre-qualification calls must be structured:

Key Fact Statement (KFS): a standardised one-page document showing APR (Annual Percentage Rate including all fees and charges), total repayment amount, processing fee, insurance charges if bundled, and grievance redressal contact. KFS must be delivered before loan execution and accepted by the borrower. Kallix agents dispatch the KFS via SMS and WhatsApp at the end of the pre-qualification call and confirm the borrower has received it before the call ends.

Cooling-off period: borrowers have 3 business days (for loans up to 7-year tenure) to cancel the loan after disbursement. Cancellation during this period requires the borrower to repay the principal received (with interest for the days it was held) but no foreclosure charge, no processing fee retention.

Kallix cooling-off advisory: during the pre-qualification call, the agent discloses: 'Once disbursed, you have 3 business days to return the funds and cancel the loan at no additional cost — just the interest for the days held.' This disclosure is mandatory under DLG 2022.

For borrowers who call within the cooling-off window to cancel, Kallix agents initiate the cancellation workflow — no retention pitch, no collections pressure, no attempt to persuade the borrower to retain the loan. The DLG explicitly prohibits such tactics.

Lenders using Kallix for DLG-compliant pre-qualification report a 22% reduction in RBI Ombudsman complaints related to undisclosed charges.

Bundled personal loan insurance — typically credit life cover that pays outstanding principal if the borrower dies or becomes permanently disabled — has historically been sold aggressively as a de-facto condition of loan disbursement. RBI DLG 2022 explicitly addressed this: insurance must be optional, the premium must be disclosed in the KFS, and the loan cannot be made conditional on insurance purchase.

Kallix insurance disclosure script:
1. The agent discloses the insurance product name, insurer, and annual premium: 'With this loan, there is an optional credit life insurance — monthly premium of Rs 450, from [insurer name]. This is completely optional and does not affect your loan approval.'
2. The agent confirms the premium is not bundled in the loan EMI unless the borrower explicitly opts in: 'If you choose the insurance, the premium is added to your loan amount. If you decline, your EMI remains Rs X.'
3. The borrower's decision (opt in / decline) is recorded with timestamp and call recording reference

Why higher voluntary take-up? When borrowers know they can say no without losing the loan, the offer feels fair rather than coercive. Borrowers who choose to take the insurance in this context make an informed decision — reducing mis-selling complaints, IRDAI grievances, and bank ombudsman cases.

For bundled insurance above Rs 500/month premium, the agent also confirms that the borrower has received the insurance policy terms and has the right to cancel within the free-look period (15 days for online policies under IRDAI regulations).

Personal loan balance transfers are economically attractive for borrowers who took loans when rates were high (12–18% pa) and can now refinance at lower rates (10–14% pa for prime borrowers). The decision, however, requires a net benefit calculation that most borrowers cannot perform — leading to either missed savings opportunities or transfers that turn out not to be beneficial after processing fees.

Kallix BT pre-qualification flow:

1. Current loan details: existing lender, outstanding principal, current ROI, remaining tenure, and monthly EMI — pulled from Account Aggregator data if available, or captured verbally.

2. Net saving calculation: Kallix computes total interest remaining on the current loan versus total interest on the BT offer, minus the BT processing fee (typically 1–2%) and foreclosure charge at the current lender (if applicable). 'If you transfer Rs 3.2 lakh outstanding at 16.5% to our 12.9% offer, you save approximately Rs 42,000 in interest over the remaining 24 months, after a processing fee of Rs 6,400.'

3. Break-even check: if the net saving is negative or less than 3 months of EMI saving, the agent advises against the transfer.

4. Bureau pull: a bureau check confirms the borrower still qualifies for the offered rate at the time of BT — rate offers are subject to final credit assessment.

5. Top-up at BT: the agent checks whether the borrower qualifies for an additional top-up at the time of BT — many lenders allow top-ups of 20–30% of the outstanding principal at BT, which the borrower can use for home renovation or other needs.

Lenders using Kallix for BT pre-qualification see a 3.2x improvement in BT conversion versus self-service online calculators.