Kallix AI voice agents handle the full spectrum of balance and transaction queries that consume 60–70% of a bank's inbound call volume. After authenticating via MPIN, OTP, or voice biometric, a customer can retrieve their available balance, check if a specific debit has cleared, confirm a UPI transfer, or request a mini statement sent to registered WhatsApp — all in under 90 seconds.

The agent connects live to the core banking system (Finacle, Oracle FLEXCUBE, Temenos T24, BankMaster, or FinnOne) via secure APIs, so every balance figure is real-time — not cached. It distinguishes between available balance (spendable), ledger balance (including uncleared credits), and blocked/hold amounts, explaining each in plain Hindi or English based on caller preference.

Beyond savings accounts, the agent handles current accounts, OD/CC limits and utilisation, FD/RD maturity dates and interest earned, and NRE/NRO balances including last forex conversion rate applied. For UPI, NEFT, RTGS, and IMPS transactions, it gives end-to-end status: initiated, pending, cleared, returned, or failed — with the reason code if the payment failed.

Production data across 60+ banking deployments shows 85–92% of balance and statement queries fully resolved without escalation. The remaining 8–15% involve disputes, PAN-linkage issues, or suspected fraud — which the agent flags and routes to the appropriate human queue with a pre-filled case summary.

RBI's cybersecurity framework (circular RBI/2015-16/418) mandates that phone banking channels use at least two-factor authentication before disclosing account balances or transactional data. Kallix implements a three-step flow: CLI (calling line identity) match against the registered mobile number; MPIN or OTP prompt; and optional voice biometric for premium or high-value accounts, using liveness detection to prevent replay attacks.

The MPIN is never stored in plaintext — it is hashed and compared against the CBS-side credential store, with zero transmission of the raw value over the voice channel. Voice biometric templates are enrolled during onboarding and stored as mathematical representations, not raw audio, compliant with IT Act 2000 Section 43A and DPDP Act 2023 consent requirements.

Failed authentication triggers a three-attempt lockout consistent with most banks' security policies, then routes to a human agent for manual KYC verification. Failed CLI matches — where the caller is not on the registered number — trigger an additional OTP step rather than immediate denial, accommodating customers calling from a different device.

Banks running Kallix report 98.5% first-attempt MPIN success rate and under 1.2% lockout rate — lower than traditional DTMF IVR systems which average 3–4% lockout due to tone-entry errors on mobile keypads.

Most bank customers hold 3–5 product relationships with their primary bank, yet traditional IVR systems typically only expose savings balance. Kallix handles the full account hierarchy a customer holds in the CBS, presenting balances contextually.

For savings accounts, the agent reports available balance, ledger balance, and uncleared instruments. For current accounts, it adds average quarterly balance (AQB) and non-maintenance charge alerts. For overdraft and cash credit accounts — common among business owners — it reports sanctioned limit, outstanding principal, and available drawing power in plain language: 'Your OD limit is ₹5 lakh, you've used ₹1.8 lakh, and ₹3.2 lakh is available to draw.'

Fixed deposits receive specific handling: principal, current interest accrued, maturity date, maturity amount (with TDS deducted at 10%), and auto-renewal status. For recurring deposits, instalments paid, instalments remaining, and projected maturity value. Customers can request FD certificate dispatch to registered email — a common requirement during loan applications.

NRE accounts report INR balance with the last conversion rate applied; NRO accounts note TDS deduction status on interest. Joint account holders can access shared balances after authentication — the agent records which account holder initiated the inquiry in the CBS audit trail, satisfying RBI's customer service record-keeping requirements.

Transaction inquiries are the second-highest volume call type after balance checks. Kallix handles them through a multi-modal delivery model. During the call, the agent reads the last 3–5 transactions in a clear format: 'On May 15th, a debit of ₹12,500 to HDFC Credit Card via NACH. On May 14th, a credit of ₹85,000 from ACME Pvt Ltd via NEFT.' This satisfies most customers wanting to confirm a specific debit.

For customers needing a written record, the agent sends a formatted mini statement to registered WhatsApp within 30 seconds — a structured message with date, transaction type, amount, cleaned narration (CBS internal codes translated to plain language), and closing balance per row. Account numbers are masked to the last 4 digits in all shared documents per RBI's statement format guidelines.

For full account statements (1, 3, 6, or 12 months), the agent triggers PDF generation in the bank's document management system and dispatches to registered email. Banks using Kallix report 74% of statement requests fulfilled by the voice agent without human involvement — a task that previously required a branch visit or NetBanking login.

For disputed or unrecognised transactions, the agent captures the transaction date and amount, logs a pre-dispute flag in the CBS, and connects the customer to the dispute resolution queue — pre-filling the case with all transaction metadata.

UPI transaction status is one of the top three call drivers for Indian retail banks, particularly after payment failures during high-traffic periods — salary day, festival seasons, GST payment dates. Kallix handles UPI status inquiries by connecting to the bank's NPCI switch interface, providing real-time status rather than cached CBS data.

The agent collects the UTR (Unique Transaction Reference) or, if the customer doesn't have it, the approximate transaction date, amount, and counterparty VPA. It returns a structured status: 'Your UPI payment of ₹8,500 to vendor@okicici on May 16th at 3:42 PM was successful. UTR: 425XXXXXXXXX. The recipient's bank confirmed settlement.' For pending transactions, it provides the NPCI-mandated resolution window — 48–72 hours for reversal if settlement fails.

For failed UPI transactions, the agent identifies the NPCI failure reason code and translates it: U30 (payer bank server down), U16 (transaction limit exceeded), U28 (incorrect MPIN by sender), U69 (payee VPA invalid). It confirms whether the corresponding CBS debit was reversed or is pending — RBI mandates failed UPI debit reversal within 24 hours (T+1 business day).

For UPI Collect (pull) requests — a merchant's payment request awaiting customer approval — the agent reads the pending collect details and connects the customer to the relevant channel to approve or reject, reducing confusion about unrecognised-looking debit attempts.

Fund transfer status is a significant call driver — particularly for NEFT, where customers don't understand the batch-settlement model and call immediately after initiating, expecting instant credit. Kallix retrieves exact status and proactively explains the settlement timeline.

For NEFT: the agent retrieves the UTR, current batch status (queued/dispatched/settled/returned), and explains that NEFT settles in 30-minute batches 24/7 on all calendar days per RBI's December 2019 expansion. It confirms the expected credit window based on which batch the transaction is in.

For RTGS (minimum ₹2 lakh): the agent confirms real-time gross settlement — typically credited within 30 minutes on banking days between 7 AM and 6 PM — and flags if the transaction is outside the RTGS operating window. For IMPS: settlement is instant (under 2 minutes) and the agent confirms credit or flags the NPCI return reason if failed.

For SWIFT (international transfers): the agent retrieves the SWIFT reference, identifies the correspondent bank chain, and provides the expected value date. For inward SWIFT (foreign remittances), it confirms receipt at the nostro account and estimated credit, factoring in FEMA conversion processing time — typically 24–48 hours for NRE, 24 hours for NRO.

Returned NEFT/RTGS transactions — wrong account number, account closed, name mismatch on high-value — are identified with the return reason and reversal timeline confirmed.

Unauthorized transaction calls are time-critical — RBI's circular on customer protection (RBI/2017-18/15) establishes zero liability for customers who report within 3 working days of receiving the transaction alert, making rapid triage essential. Kallix handles this as a priority escalation flow, not a standard inquiry.

The agent identifies fraud intent from natural language ('I didn't make this payment', 'someone used my card', 'I didn't authorize this debit') and shifts to fraud intake protocol. It captures: transaction date, amount, channel (UPI, NACH, card, NEFT), and counterparty details. It never asks for OTPs, card numbers, or PINs — explicitly prohibited per RBI guidelines on social engineering fraud prevention.

Simultaneously, the agent logs a provisional fraud flag in the CBS, triggering the bank's fraud management system (SAS, NICE Actimize, or ACI Worldwide) to place a review hold on further debits from the same channel. If card fraud is reported, the agent initiates a hot-listing request and confirms replacement card dispatch — typically 5–7 working days.

The customer is transferred to the fraud operations queue with a pre-filled case summary: customer name, account number, reported transaction details, and event chronology. This cuts human agent intake time from 4–6 minutes to under 90 seconds and starts the mandatory 10-working-day resolution clock at the point of reporting.

For a balance inquiry to be trustworthy, it must reflect the live CBS state. Kallix connects via the same API layer used by NetBanking and mobile apps, ensuring the balance a customer hears matches what they'd see at the same moment in internet banking.

The integration follows a secure API gateway pattern: Kallix's voice engine calls the bank's API management layer (Apigee, AWS API Gateway, or MuleSoft), which authenticates the service-to-service request via mTLS (mutual TLS) and OAuth 2.0 client credentials, then forwards the query to the CBS. No direct CBS access is granted to Kallix's systems — all data flows through the bank's existing API infrastructure, satisfying IS Audit and CISO perimeter security requirements.

For banks running older CBS versions without published APIs, Kallix uses pre-built middleware adapters for Finacle 10.x/11.x, FLEXCUBE 12.x/14.x, and FinnOne 7.x. Integration timelines are 3–6 weeks for legacy CBS and 2–3 weeks for banks with a fully published API catalogue.

Data returned to the voice engine is encrypted in transit (TLS 1.3) and never persisted on Kallix infrastructure — each call triggers a fresh API pull, and the response is used only for the duration of the call session, aligning with RBI's data localisation and storage minimisation guidance.

NRI customers represent a high-value, high-need segment — often 3–9 time zones away, unable to walk into a branch, and facing regulatory constraints (FEMA, TDS, repatriation limits) that generic IVR systems don't address. Kallix handles NRI balance inquiries with India-specific regulatory intelligence built in.

For NRE savings and fixed deposits, the agent confirms INR balance and notes that principal and interest are fully repatriable per FEMA Section 6. For NRO accounts — used for income earned in India (rent, dividends, pension) — it confirms balance, TDS deducted on interest (30% flat for NRIs under Section 195 of the IT Act, or lower per DTAA for UAE, UK, USA, Mauritius), and the USD 1 million annual repatriation cap under RBI's framework.

For FCNR (B) deposits in foreign currency (USD, GBP, EUR, AED, SGD, AUD), the agent reports both the foreign currency balance and INR equivalent using the day's RBI reference rate (published ~12:30 PM IST on business days), maturity date, and auto-renewal currency.

International routing: calls from +44, +1, +971, +65, +61 country codes are routed through a dedicated NRI IVR path that skips India-centric prompts. Voice biometric enrollment during V-CIP digital onboarding enables frictionless subsequent authentication without requiring the NRI to call from a registered Indian mobile.

NACH mandate management is a significant source of customer confusion — debit failures result in late payment charges, policy lapses, or EMI bounces, but customers rarely understand why a mandate failed or what to do. Kallix queries both the CBS and NPCI mandate registry to give a complete picture.

For each active mandate, the agent reports: the beneficiary name, mandate amount (fixed or maximum with actual debit), debit frequency, next debit date, and last execution status. For returned mandates, it provides the NPCI return reason in plain language: R01 (insufficient funds), R05 (account closed), R14 (mandate expired), R28 (account holder stop instruction) — and advises the resolution path (top-up account, contact lender for fresh mandate).

Standing instructions set within the bank — monthly SIP contributions to the bank's mutual fund arm, recurring FD creation, BBPS utility bill payments — are queried from the CBS SI module. The agent confirms execution date, amount debited, and beneficiary credited, or flags execution failures with reason.

For customers needing to cancel or modify a NACH mandate, the agent captures the request, routes it to the NACH desk with pre-filled mandate details, and sets expectations: NACH cancellations take 7–10 business days to reflect in the NPCI registry.

Cheque-related inquiries remain relevant in India, particularly for high-value transactions in real estate, business payments, and loan disbursements. Kallix handles cheque status via CBS cheque management module queries.

The customer provides the 6-digit cheque number, and the agent cross-references it with the account to return: clearing status (presented, in-clearing, cleared, returned), the presenting bank and branch from MICR data, the amount, and date presented. For outward cheques not yet cleared, the agent confirms the current float period.

For inward cheques (deposited by the customer), the clearing timeline follows CTS grid rules: within the same CTS grid (Mumbai, Delhi, Chennai), clearing completes on T+1. Outstation cheques outside the CTS grid take T+2 to T+5.

For returned cheques, the agent provides the MICR return reason: 01 (insufficient funds), 06 (account closed), 14 (stop payment), 25 (signature differs), 40 (irregular endorsement). For commercial customers, the agent notes that a returned cheque for insufficient funds triggers NI Act Section 138 — civil/criminal liability of up to twice the cheque amount — and advises seeking legal counsel. Stop-payment instructions for unpresented cheques can be initiated and confirmed by the agent.

India's PMJDY accounts — over 53 crore as of 2024 — represent the most financially underserved banking customers. These account holders rarely use NetBanking or mobile apps but call the helpline to check if government benefit transfers have arrived. Kallix's AI is specifically calibrated for this segment.

The agent defaults to Hindi for calls from STD codes associated with tier-2 and tier-3 cities and rural areas, using simplified vocabulary: 'aapke khate mein' rather than financial jargon, and reads numbers in the Indian system at a slightly slower pace. For illiterate or low-literacy callers who struggle with MPIN entry, an alternative authentication path using Aadhaar last 4 digits plus date of birth is offered where bank security policy permits.

For DBT inquiries — the primary reason PMJDY holders call — the agent queries the PFMS (Public Financial Management System) credit record in the CBS and confirms: 'PM-Kisan ₹2,000 was credited on April 15th' or 'MGNREGA payment of ₹3,600 has not arrived yet; government dispatch date is April 30th per PFMS records.' This reduces branch walk-ins from beneficiaries who can't access NetBanking.

For the ₹10,000 overdraft facility (available to PMJDY accounts active for over 6 months with consistent transaction history), the agent confirms the sanctioned OD limit, current outstanding, available amount, and the simple interest rate applicable — currently 4% per annum for qualifying accounts under the government subsidy scheme.

Self-service containment is the primary metric banks track for AI automation ROI. Kallix's production benchmark across 60+ banking deployments shows 85–92% containment on the balance/transaction inquiry category — significantly above the 45–60% containment typical of legacy DTMF IVR systems, which require customers to navigate numeric menus and fail on non-standard queries.

The containment gap is driven by three factors: natural language understanding (customers can ask 'did my salary come in?' rather than pressing 3 for balance), real-time CBS integration (no 'please check NetBanking for an up-to-date balance' deflections), and multi-modal delivery (WhatsApp dispatch eliminates the 'I couldn't hear the balance' repeat-call driver).

Average handle time for a standard balance inquiry is 55–80 seconds: 18–22 seconds authentication, 5–10 seconds CBS query, 20–35 seconds response and optional statement dispatch confirmation. Human agents handling the same query average 3.5–4.5 minutes. Applied to a bank receiving 50,000 balance-related calls per month, the AHT reduction frees over 2,000 human-agent hours per month.

First-call resolution runs above 91%. The 9% requiring escalation: disputed transactions (2%), PAN-Aadhaar linkage requiring human review (3%), and edge cases like deceased account inheritance queries (4%) — all correctly identified, flagged, and routed by the AI.

A human agent handling balance inquiries costs the bank Rs 65–90 per call — Rs 30–40 base labour cost, Rs 15–20 infrastructure (seat, telephony, HRMS), and Rs 10–15 supervision and QA overhead. At typical contact centre utilisation of 70–75%, the effective cost climbs further.

Kallix's AI costs Rs 12–18 per resolved interaction, covering telephony, API compute, CBS query charges (typically Rs 0.20–0.50 per CBS API call under the bank's existing agreement), and Kallix platform fees. The cost is fixed per interaction regardless of duration within the normal range.

For a mid-size bank handling 50,000 balance and transaction inquiry calls monthly: at 88% containment, 44,000 calls resolved by AI at Rs 15 average = Rs 6.6 lakh/month. Without automation at Rs 77 average = Rs 33.9 lakh/month. Monthly saving: Rs 27.3 lakh. Annual saving: Rs 3.27 crore. Against a typical implementation cost of Rs 30–50 lakh, break-even occurs at month 4–6 and 12-month ROI is 5–7x.

Additional ROI levers: 24/7 availability eliminates Rs 15–25 lakh/year in after-hours staffing; FCR improvement from 78% to 91% reduces total call volume by 12–15%; and CSAT improvement from 3.4 to 4.2 on a 5-point scale has documented downstream impact on product cross-sell acceptance rates.

Banking is one of India's most heavily regulated sectors for customer communication, and Kallix's platform is built with these frameworks as design constraints, not afterthoughts.

RBI cybersecurity framework (RBI/2015-16/418): mandates two-factor authentication for phone banking balance disclosure; prohibits IVR systems from reading out full account numbers; requires call recording and retention for 5 years; and mandates 24/7 helpline availability per the Damodaran Committee customer service recommendations (2014). Kallix satisfies all four through its auth design, masked-number policy, recording infrastructure, and always-on architecture.

TRAI TCCCPR 2018 governs outbound proactive alerts — banks sending balance alerts, transaction confirmations, or fraud warnings via voice call must be registered as transactional communicators, maintain NDNC scrub records, and limit calls to 9 AM–9 PM IST. Inbound customer-initiated calls are not subject to TCCCPR restrictions.

IT Act 2000 Section 43A and DPDP Act 2023: account balance data is 'sensitive personal information' under both frameworks. Kallix processes this data solely for the duration of the call session with zero persistent storage on Kallix infrastructure. Voice biometric consent is captured in writing during onboarding per DPDP Section 6.

For co-branded and partnership banking products (embedded finance, NBFC-bank partnerships), RBI's digital lending guidelines (September 2022) additionally require disclosure of the LSP identity at the start of the call and 3-day cooling-off period communication.

India's linguistic diversity is a practical barrier for centralised banking customer service. Kallix's multilingual pipeline goes beyond simple translation — it handles the full linguistic experience contextually.

Language detection operates in the first 4–5 seconds using a speaker language identification model that distinguishes 12 languages without requiring a menu selection. Once detected, the entire session — authentication prompts, balance readout, number formatting, and WhatsApp messages — switches to that language. For Hinglish (the dominant register among urban young customers), the agent responds in the same code-switched register rather than forcing pure Hindi or English.

Language-specific number formatting is handled correctly: in Hindi, ₹15,75,000 is read as 'pandrah lakh pchhattar hazaar rupaye', not the English-accented 'fifteen lakh seventy-five thousand'. For Tamil, the Tamil number system is used. This detail significantly reduces caller confusion and repeat inquiries.

For PMJDY and rural callers, the agent is tuned to dialects common in Bhojpuri-influenced Hindi (UP/Bihar), Rajasthani-influenced Hindi, and non-standard Tamil. Banks with geographic concentration (cooperative banks serving a single state) can request single-language deployments with dialect tuning.

Escalation design is where most AI banking deployments fail — an abrupt transfer requiring re-authentication and re-explanation destroys trust and negates efficiency gains. Kallix's escalation model uses warm transfer with pre-filled context as the default.

The AI identifies escalation triggers in two categories: explicit (customer says 'speak to a human', 'I want to talk to someone', 'agent please') and implicit (three failed authentication attempts, fraud keywords detected, 'this is wrong', complex regulatory query). For implicit triggers, the agent proactively offers: 'This looks like something I should connect you with a specialist for — shall I transfer you now?'

On transfer, the human agent's screen receives: caller's authenticated identity (name, masked account number), the query being handled, information already provided, any fraud or dispute flags raised, and suggested resolution path. This pre-fill — delivered via integrations with Avaya, Genesys, or Cisco Finesse telephony platforms — eliminates the re-authentication and re-explanation overhead that accounts for 40–60% of human agent handle time on escalated calls.

For after-hours escalation, the agent offers callback booking: 'Our specialists are available 8 AM–8 PM IST. Shall I book a callback at a time that works?' Callback slots are written to the contact centre's scheduling system (Verint, NICE, Genesys WFM) with call context pre-attached, eliminating after-hours IVR dead-ends that typically cause 23–31% of callers to abandon without resolution.

Banking deployments require more rigorous security and integration review than most verticals, and Kallix's deployment process is designed around the bank's IT governance requirements.

Week 1–2: CBS API integration. Kallix's integration team connects to the bank's API gateway (Apigee/MuleSoft/AWS), establishes mTLS certificates, and conducts API contract testing across all account types and balance scenarios. The bank's IS Audit team reviews Kallix's SOC 2 Type II certification and security questionnaire. Data flow diagrams showing that no balance data is persisted outside the bank's perimeter are provided for CISO sign-off.

Week 2–4: Language model tuning. Kallix receives 2,000–5,000 anonymised inbound call transcripts from the bank, used to fine-tune ASR for the bank's customer demographic — accent patterns, product-specific jargon, and common query phrasing. IVR routing logic is configured in the bank's telephony platform.

Week 4–7: UAT and go-live. UAT is conducted on a staging CBS environment — Kallix runs 300+ test cases covering all account types, edge cases (joint accounts, frozen accounts, PMJDY limits), and failure modes (CBS timeout, network interruption). After UAT sign-off, production cut-over happens during a low-traffic maintenance window. Kallix provides a 30-day hyper-care period with a dedicated customer success manager monitoring containment rates and escalation patterns daily.

Most Indian banks run DTMF IVR systems deployed 10–15 years ago. These require customers to navigate a menu tree (Press 1 for balance, Press 2 for statement, Press 3 for transaction), which fails whenever the customer's need doesn't map to the available menu — or when the customer simply prefers to speak.

Key performance comparison based on production data: Containment — DTMF IVR 45–60% vs Kallix AI 85–92%. The gap comes from natural language understanding: a customer saying 'my salary hasn't come in' is resolved by the AI, whereas the DTMF IVR would require navigating to 'inward credits' or 'recent transactions', which many customers don't do and instead press 0 for an agent.

AHT: DTMF IVR 2.5–3 minutes (menu navigation, re-prompts for invalid entries) vs Kallix AI 55–80 seconds. The IVR generates 18–24% repeat calls from customers who couldn't hear or note the balance aloud. Kallix's WhatsApp dispatch reduces this to under 4%.

Language: most DTMF IVRs support Hindi and English via pre-recorded audio. Kallix supports 12 languages with live synthesis, allowing callers to switch mid-call and ask follow-up questions not anticipated in the IVR script.

Maintenance: DTMF IVR requires re-recording and re-publishing audio files for every product, rate, or policy change. Kallix updates via configuration change. Migration from existing IVR is done in parallel — routing 20% of balance calls initially, validating containment and CSAT, then expanding to 100% over 8–12 weeks.

FD and RD inquiries spike at two predictable moments: the day after issuance (customers want to confirm the deposit is reflecting correctly) and 30–45 days before maturity (customers planning the reinvestment or withdrawal). Kallix's CBS integration surfaces both scenarios with proactive outreach — not just reactive inquiry response.

For FD maturity inquiries: the AI provides the exact maturity date and amount, confirms the current auto-renewal instruction ('Your FD is set to auto-renew at the prevailing rate at maturity — would you like to change this to a payout to your savings account instead?'), and advises on the TDS deduction if applicable. FDs above Rs 40,000/year in interest trigger TDS at 10% (20% without PAN). The AI confirms whether Form 15G/15H has been submitted for the current financial year.

Pre-maturity withdrawal advisory: when a customer asks about breaking an FD before maturity, the AI calculates the penalty (typically 0.5–1% below the contracted rate for the actual tenure) and presents the net interest receivable after penalty. This calculation is done in real time from the CBS — giving the customer an accurate figure rather than a generic penalty statement.

RD missed installment: if the RD has a missed installment, the AI alerts the customer and confirms the penalty structure (typically Rs 1–2 per Rs 100 per month of delay). The AI offers a UPI payment link to clear the pending installment on the call.

FD loan (overdraft against FD): the AI advises customers considering premature closure that a loan against FD (typically 90–95% of FD value, at FD rate + 1–2%) is cheaper than breaking the FD and losing the penalty. This advisory reduces premature closures by 18–24%.

Loan outstanding balance inquiries are the second most frequent reason customers contact their bank (after account balance), yet most banks handle these via human agents or partially automated IVR without real-time LMS integration. Kallix's LMS API integration provides live, accurate outstanding balances — not the figure from the last statement.

Key details provided in a loan balance inquiry call: (1) current outstanding principal (as of today, not statement date); (2) next EMI amount (including any EMI change due to floating rate revision); (3) last three EMI payment dates and amounts (confirmation that recent payments are reflecting); (4) remaining tenure in months; (5) total interest paid to date vs total interest payable over the loan's full tenure.

Floating rate impact advisory: for home loans, the AI proactively advises customers when a recent RBI repo rate change has affected their EMI or tenure: 'The RBI reduced the repo rate by 25 bps in [month]. Your floating rate home loan rate has been revised from [X]% to [Y]%. Your revised EMI from [date] is Rs [Z] — a reduction of Rs [A] per month.' This proactive communication reduces rate-related inquiry calls significantly.

Amortisation schedule dispatch: the AI generates a PDF amortisation schedule (month-by-month EMI breakup into principal and interest) and dispatches it via WhatsApp and email within 2 minutes. Customers preparing for IT returns or financial planning frequently request this.

Prepayment/foreclosure estimate: the AI calculates the current foreclosure amount (outstanding principal + accrued interest + prepayment charges if applicable) and confirms the prepayment charge rate (RBI prohibits prepayment charges on floating rate home loans — the AI confirms nil prepayment charges for eligible loans).

Credit card balance inquiries are among the highest-frequency banking customer service calls — particularly in the 5–10 days before the payment due date. Kallix's CBS/card management system integration provides live balances, not the last statement balance.

Billed vs unbilled balance distinction: the AI explains the two components: 'Your billed amount due by [date] is Rs [X] — this is from your last statement. Your unbilled amount of Rs [Y] represents purchases since your last statement date — these will appear in next month's bill.' Customers who confuse these two figures frequently overpay or underpay; the AI's clear breakdown reduces this confusion.

Minimum due advisory: the AI explains the true cost of paying only the minimum: 'Your minimum due is Rs [X] — paying only this will result in interest of [rate]% per month (effectively [annual]% p.a.) on the outstanding balance. The full payment to avoid interest is Rs [Y].' This education reduces chronic minimum-payment behaviour and lowers revolving credit risk for the bank.

Credit limit increase: post-balance disclosure, the AI checks if the customer is eligible for a pre-approved credit limit increase (based on payment history and credit utilisation in the CMS) and presents it: 'You are eligible for a limit increase to Rs [Z] — shall I process this now?' Pre-approved limit increase conversion on balance inquiry calls: 14–20%.

Transaction dispute initiation: 'I see a Rs [X] charge from [merchant] on [date] — is this a transaction you recognise?' If not recognised, the AI initiates a dispute lodging on the call and advises on the timeline (typically 7–15 working days for dispute resolution).

Account statement requests are the most easily automatable banking inquiry — yet many banks still require customers to visit a branch or log into NetBanking for historical statements. Kallix AI automates the full flow in under 3 minutes from authentication to dispatch.

Custom date range statements: the AI accepts natural language date specification ('I need transactions from April 1 to June 30') and converts to a date-range query against the CBS. For periods beyond 12 months (common for home loan IT deductions), the AI confirms the maximum available period and generates in parts if the CBS limits single-query statement size.

Password-protected PDF: the statement PDF is protected with the customer's date of birth (DDMMYYYY format) as the password — a bank standard that does not require a separate password communication. The AI communicates the password: 'The PDF is protected with your date of birth in DDMMYYYY format.'

Form 16A for TDS on savings interest: customers who have paid TDS on their savings account interest (above Rs 10,000/year for most banks) can request their Form 16A (TDS certificate) for IT filing. The AI confirms TDS deducted for the financial year and dispatches Form 16A via email. For customers with multiple accounts at the same bank, Form 16A is consolidated across all accounts.

Statement for specific purpose (loan application, visa, tenancy): customers often need statements for a specific stated purpose — the AI notes the purpose in the request log for audit purposes but does not add any attestation without a branch visit. It advises on the bank's certification process if an attested/stamped statement is required.

Dormant and inoperative accounts are a significant customer service and compliance issue: RBI mandates that banks maintain dormant accounts without charges and transfer unclaimed deposits to the Depositor Education and Awareness (DEA) Fund after 10 years of inactivity. For customers, discovering their account is inoperative — often when trying to make a payment — is a source of significant frustration.

Kallix AI's proactive dormancy prevention: at the 22-month mark of inactivity, the AI makes an outbound call: 'Your [account type] account has had no transactions for [N] months. Under RBI guidelines, accounts inactive for 24 months are classified as inoperative and debits are blocked. Making any transaction — even a Rs 1 transfer — before [date] will prevent this restriction.' This specific call-to-action (make any transaction) is the simplest dormancy prevention intervention and achieves 64–72% account reactivation.

For already inoperative accounts: the AI explains the two-step reactivation process — (1) submit a written request to the bank (branch visit or physical letter) with the account holder's KYC documents; (2) the bank confirms reactivation within 3–5 working days. For banks offering video KYC for reactivation: the AI books the video KYC appointment directly.

DEA Fund transfer: for accounts inactive for 10+ years, the balance is transferred to RBI's DEA Fund. The owner (or legal heir) can still claim the balance by filing with the bank, which processes the claim with RBI. Kallix AI explains this process when customers inquire about very old inactive accounts — preventing the loss of funds the customer was unaware of.

Zero-balance dormant accounts: for accounts with zero balance that have been dormant, the AI advises that reactivation is still possible but the account may be closed by the bank after extended dormancy per the account agreement terms.

Joint accounts in India serve multiple purposes: spousal savings accounts, parent-minor accounts, and business partnership current accounts. Each has a different operating instruction that determines who can transact and who can inquire — and the AI must enforce these distinctions correctly.

Operating instruction mapping: (1) Either or Survivor (E or S): most common — any joint holder can operate independently. Both holders can inquire and transact. AI authenticates whichever holder is calling. (2) Former or Survivor (F or S): only the primary (Former) holder can transact during their lifetime; the Survivor inherits. The AI confirms this restriction to the secondary holder if they attempt to initiate a transaction. (3) Jointly (J): all holders must authorise — the AI cannot process transactions for jointly operated accounts and routes these to the branch.

Secondary holder incomplete KYC: a common issue is that secondary holders were added to the account years ago with incomplete digital KYC — their mobile number is not registered, or their date of birth in the CBS does not match their ID. The AI cannot authenticate these holders over the phone and routes to the branch for KYC update.

Death of a joint holder: when a customer calls to report the death of a joint holder and inquire about account status, the AI explains the survivor's rights, the required documents (death certificate, succession certificate if the account is not E or S), and the account freezing implications during the claim process. The AI routes to the bank's NRI/estate team if the deceased was an NRI.

Minor account (parent as joint holder): for savings accounts with a minor and parent as joint holders, the AI confirms when the minor will turn 18 and the process for converting to a sole account.

Demat account inquiry via AI is primarily used for portfolio value checks (total holdings at current market prices) and confirmation of recent transactions — not investment advisory (which would require SEBI IA registration). The AI presents factual data, not recommendations.

Holdings query: the AI presents the top 5–10 holdings by current market value, total portfolio value at live prices, and the total change in value since the last business day (absolute Rs and percentage). For users who ask about specific stocks: 'Your holding in [Company Name] is [N] shares at a current market price of Rs [X] — current value Rs [Y], representing a [+/-Z]% change from your average acquisition cost of Rs [W].'

Corporate action alerts: the AI proactively notifies demat account holders of upcoming corporate actions — rights issues (acceptance deadline), bonus record dates, dividend credit dates, and AGM e-voting deadlines. These proactive alerts drive 28–34% higher corporate action participation compared to email-only notification.

Pledge status: for demat holdings pledged against a loan or margin facility, the AI confirms which holdings are pledged, the pledge value, and the loan outstanding — allowing the customer to assess whether they have unpledged holdings they could pledge for additional credit.

Mutual fund units in demat: units held in demat form (as opposed to statement of account format) are reflected in the demat holdings. The AI confirms the NAV-based current value and the XIRR (investment return) if available from the DP platform. SEBI caveat: the AI presents XIRR as a factual data point and does not interpret it as a buy/hold/sell signal.

TDS on savings and FD interest is a significant customer service category — particularly in Q1 of each financial year (April–June) when customers discover TDS was deducted on their prior-year interest income and want to understand their tax position.

Form 15G/15H submission: the AI confirms whether a Form 15G (non-senior citizen, income below exemption limit) or Form 15H (senior citizen) has been submitted for the current financial year. If not submitted and the customer is eligible: 'You may be eligible to submit Form 15G to avoid TDS deduction on your interest income this year — I can send you the form via WhatsApp and guide you through submission at the branch.' Proactive 15G/15H advisory in Q1 reduces unnecessary TDS deductions and associated IT refund processing.

TDS reconciliation with Form 26AS: the most common inquiry is 'TDS was deducted but not showing in my 26AS.' The cause is typically a TRACES upload delay (banks must upload TDS data to TRACES quarterly — Q4 data uploaded in June). The AI explains: 'TDS from January to March is uploaded to TRACES by June 15 — it will be reflected in your Form 26AS by then. We have raised a reconciliation check to confirm the upload is on track.'

Senior citizen enhanced TDS exemption: from FY 2021–22, Section 194A exempts senior citizens from TDS on bank FD interest up to Rs 50,000/year (vs Rs 40,000 for others). The AI confirms the senior citizen flag in the CBS and verifies the TDS exemption threshold has been correctly applied. If incorrectly applied (TDS deducted below the Rs 50,000 threshold for a senior citizen): the AI raises an internal correction request and provides the refund timeline (typically next payment cycle within 30 days).

Sweep account inquiries are a source of customer confusion because the account statement shows only the savings balance — not the auto-created FD tranches that hold the swept funds. Customers believe their 'available balance' is lower than it actually is.

Kallix AI provides a consolidated view: 'Your savings account balance is Rs [X]. You also have [N] auto-sweep FD tranches totalling Rs [Y] — your total accessible funds are Rs [X + Y]. The next Rs [Z] you need above your savings balance will be swept back from your most recent FD tranche at [date].' This consolidated view resolves the most common complaint: 'I have money in the account but my payment was rejected.'

Sweep-back calculation: the AI explains that when the savings balance falls below the minimum (typically Rs 10,000–25,000), the oldest or newest FD tranche (depends on the bank's sweep-back rule) is broken and the funds move to savings. Interest on the broken tranche is paid pro-rata for the actual tenure. The AI calculates the expected interest on a hypothetical tranche break in real time if the customer is considering a large withdrawal.

Sweep threshold update: if the customer wants to change the sweep threshold (e.g., increase it so more funds go into FD tranches), the AI confirms whether this can be done via a verbal instruction on the recorded call or requires a branch visit per the bank's policy.

Interest rate transparency: each FD tranche earns interest at the rate applicable on the date it was created — not a blended rate. The AI lists each tranche's rate separately so the customer understands the interest income from their sweep account.

Bank locker management is a largely manual process at most banks — customers have to visit the branch for almost every interaction. Kallix AI automates the inquiry and reminder layer, reducing branch footfall for locker-related queries by 44–58%.

Locker rent payment reminder: RBI mandates that banks send locker rent reminders 30 days before the due date. If rent is not paid, the bank can restrict locker access after a defined notice period. Kallix AI manages the reminder sequence with a payment link: at Day 30, Day 14, and Day 3 before due date, the AI calls with a UPI payment link for the exact rent amount. Same-session payment rate: 68–76%.

Locker access record: under RBI's 2017 locker guidelines, banks must maintain a record of every locker access — date, time, and the identity of the person accessing. The AI can confirm the last access date and time, which is useful for estate settlement (proving when a locker was last accessed by the deceased) or dispute resolution (if the customer claims the locker was accessed without their knowledge).

RBI 2022 revised locker agreement: RBI mandated a revised standardised locker agreement that clarifies bank liability (banks are liable for losses due to their own negligence — fire, burglary due to security lapse — up to 100x annual rent; not liable for natural disasters or customer negligence). Customers who had not signed the revised agreement by January 2023 faced locker access restrictions. The AI confirms which customers have signed the revised agreement and routes unsigned customers to the branch for signing.

Locker nomination: the AI confirms whether a nominee has been registered for the locker contents and advises on the process for adding or changing the nomination. Locker nomination is distinct from account nomination and must be done separately.

PAN-Aadhaar linkage became a significant driver of banking customer service calls from 2023, with Income Tax Department restrictions on non-linked PANs affecting savings account debit limits and TDS rates. Many customers call confused about why their account is restricted.

Kallix queries the bank's PAN-Aadhaar linkage status record in the CBS or KYC management system, which syncs with NSDL daily. The agent reports clearly: 'Your PAN ending XXXX3456 is linked to Aadhaar — no restrictions apply.' Or: 'Your PAN is not yet linked to Aadhaar. Accounts with unlinked PANs face restrictions on debit transactions above ₹50,000 and TDS at 20% instead of 10%.'

For accounts with an active freeze or restriction, the agent explains the specific restriction level: no-debit (funds in but cannot be withdrawn), no-credit-or-debit (completely frozen), or partial restriction (only transactions above a threshold blocked). It guides the resolution path: NSDL portal self-service for PAN-Aadhaar linking, or Aadhaar seeding via the bank's NetBanking portal.

If the resolution requires a branch visit — such as manual KYC or biometric Aadhaar update where Aadhaar data doesn't match PAN — the agent books the appointment in the branch scheduling system, confirms required documents, and sends a WhatsApp checklist. Proactive advisory on compliance deadlines (extended to December 2025 for non-KYC PANs) is included.