Automating the Lending Document Stack: From Upload to Credit Decision

The lending document stack is a collection of different document types, each carrying different kinds of information about the borrower. Salary slips establish income. Bank statements reveal cash flow and obligations. KYC documents confirm identity. ITR filings show annual declared income. Property documents establish collateral. Each document type is processed separately, often by different people, before a credit analyst has the complete picture needed for a decision.

Automating this stack means each document type is processed the moment it is submitted, the data is extracted, and the credit analyst arrives at the application with a structured summary ready — not a pile of PDFs.

The Complete Lending Document Stack

For a standard retail loan application in India, the document stack typically includes:

Bank statements

Three to six months of statements from one or more accounts. Key extracted fields: salary credits (amount, frequency, source), EMI debits (count, total amount), inward bounces, average monthly balance, end-of-month balance. This data feeds directly into FOIR calculation and cash flow analysis.

Salary slips

Two to three months of slips. Key extracted fields: employer name, gross salary, net salary (actual take-home), deductions breakdown. Cross-referenced with bank statement salary credits to verify consistency.

ITR documents

Income Tax Return filings for the last one to two years. Key extracted fields: gross total income, net taxable income, tax paid, assessment year, PAN number. Used for income verification, especially for self-employed applicants where salary slips are unavailable.

KYC documents

Aadhaar card and PAN card. Key extracted fields: name, date of birth, address, Aadhaar number, PAN number. Used for identity verification and address proof. Cross-checked against application form entries to flag discrepancies.

Property documents (secured loans)

Sale agreements, property tax receipts, title documents. Key extracted fields: property address, survey number, owner name, property type, registered value. Used for collateral assessment and LTV calculation.

Form 16

Employer-issued tax certificate. Key extracted fields: gross salary, TDS deducted, employer TAN, assessment year. Corroborates salary slip and ITR data.

Why Processing Each Type Separately Creates Delays

When document types are processed by different teams or at different stages of the workflow, information sits in silos until it is manually consolidated. An analyst computing FOIR needs salary slip data, bank statement EMI data, and sometimes ITR income data simultaneously. If these are processed sequentially, the analyst waits for each stage before the calculation is possible.

Parallel processing — all document types extracted simultaneously the moment the application is submitted — eliminates this sequential delay. By the time the credit analyst opens the application, all document data is structured and available.

Automated FOIR Calculation

FOIR (Fixed Obligation to Income Ratio) requires two numbers: the applicant's net monthly income and their total monthly fixed obligations (existing EMIs plus the proposed EMI). Both numbers can be derived directly from extracted document data:

• Net income: extracted from salary slip (net_salary field) and verified against bank statement salary credits
• Existing obligations: extracted from bank statement (total_emi_debits field), representing all existing EMI payments

When both numbers are extracted automatically, FOIR calculation happens without analyst input. The analyst sees the computed FOIR, the underlying data it was calculated from, and the confidence scores on the input fields — so they know whether the calculation is based on high-confidence extractions or values that need verification.

Cross-Document Validation

A significant advantage of automated extraction across the full document stack is automated cross-document validation. Rules that previously required an analyst to manually compare documents can be applied automatically:

• Salary slip net_salary vs. bank statement salary credit amount — do they match within tolerance?
• ITR declared income vs. salary slip gross salary — consistent across years?
• KYC name vs. application form name — exact match or discrepancy requiring review?
• PAN number on KYC vs. PAN number on ITR — consistent?

Discrepancies are flagged automatically. The analyst's attention is directed to the specific inconsistency, not to reading every document in full to find it.

Processing Time Impact

Manual document processing in a mid-size NBFC typically takes one to two business days from document submission to having a complete, verified data set ready for the credit analyst. With the full document stack automated, the same data set is available in under five minutes from submission.

The credit analyst's time shifts from document reading to decision-making. Application throughput increases without adding headcount.