A credit analyst who opens an incomplete file — missing a document, containing a document whose date is outside the acceptance window, or whose income figures across different documents do not reconcile — cannot complete their assessment. They return the file to the operations team. The operations team contacts the borrower. The borrower resubmits. The analyst reopens the file. The process has consumed three days and produced no credit decision. The Document Ops Agent AI runs completeness and consistency checks before the file reaches the credit analyst — so the file that arrives on their desk is complete, current, and internally consistent.
A credit analyst who opens an incomplete file — missing a document, containing a document whose date is outside the acceptance window, or whose income figures across different documents do not reconcile — cannot complete their assessment. They return the file to the operations team. The operations team contacts the borrower. The borrower resubmits. The analyst reopens the file. The process has consumed three days and produced no credit decision. The Document Ops Agent AI runs completeness and consistency checks before the file reaches the credit analyst — so the file that arrives on their desk is complete, current, and internally consistent.
The two types of document failure — and why both must be caught before credit review
Document failures in lending operations fall into two categories. The first is incompleteness — a required document is missing, the wrong document was submitted (a P&L statement when a balance sheet was required), or the document exists but cannot be read (a photograph taken at an angle, a PDF with a corrupt page, a bank statement whose last two months were not included). The second is inconsistency — documents that individually appear valid but whose key figures do not reconcile with each other. A payslip showing ₹95,000 net salary and a bank statement showing ₹68,000 average monthly credits is an inconsistency. It may have an innocent explanation (the borrower has a second bank account), but it cannot be ignored — and it certainly cannot be presented to the credit analyst as a clean file.
Both categories create work. An incomplete file must be returned to the borrower for resubmission. An inconsistent file must be escalated to the credit analyst for manual reconciliation. The Document Ops Agent AI catches both before the file crosses the operations-to-credit boundary — and returns it to the borrower once, with a specific list of what is missing or inconsistent, rather than discovering the gap after the credit analyst has already opened the file.
"A file that reaches the credit analyst with a missing document has already wasted two people's time — the analyst who opened it and the borrower who submitted it thinking it was complete. The Document AI catches it before either of them is involved."
The completeness check framework: four gate categories
Gate 1 — Document Presence
Is every required document present? · Based on applicant type and product
✓
KYC identity documents — PAN and Aadhaar present and matchPAN card PDF received. Aadhaar PDF received. Name, DOB, and address cross-referenced across both documents. Name variation flagged: "Ananya Krishnamurthy" vs "A. Krishnamurthy" — within acceptable variation (initial abbreviation).
Present
✓
Income documents — all 3 required for salaried applicantLast 3 months' payslips: received (Oct, Sep, Aug 2025). Form 16 AY 2024-25: received. 6 months' bank statement: received (May–Oct 2025).
Present
✗
Property documents — registered sale agreement required, not receivedApplication is for a home loan on an identified property. The registered sale agreement (or booking receipt with builder) is a mandatory document for title verification. The borrower submitted a builder's brochure — this is not a substitute for the sale agreement. Nudge triggered: "To proceed, please upload the registered sale agreement or your booking receipt with your builder."
Missing
⚑
Rental income proof — bank statement corroboration required but not submittedA rental agreement was submitted claiming ₹18,000/month. Policy requires this to be corroborated by bank statements showing regular rental credits. The 6-month bank statement submitted is for the primary salary account — rental credits are not visible. A second bank statement or separate rent receipt trail is needed.
Flag — needs corroboration
Gate 2 — Document Currency
Is every document within the acceptance window? · Dates checked against today
✓
Payslips within 3-month acceptance windowAcceptance window: documents dated within 3 months of application date (application: Nov 14, 2025). Received payslips: Oct 2025 (1 month old), Sep 2025 (2 months), Aug 2025 (3 months). All within window.
Current
✓
Form 16 — current or prior assessment yearForm 16 received is for AY 2024-25. Application date is Nov 2025 — AY 2025-26 Form 16 is not yet available (issued at year-end). AY 2024-25 is accepted as the most recent available. Flag: if application is after April 2026, AY 2025-26 Form 16 will be required.
Current
⚑
Bank statement — 6-month window, last statement must be within 30 daysAcceptance requirement: bank statement must cover the most recent 6 months, with the most recent statement dated within 30 days of the application date. Bank statement received covers May–October 2025. Last statement: October 31, 2025. Application date: November 14, 2025. Gap: 14 days — within 30-day window. Pass. However: the October statement is "partially downloaded" — pages 1–3 of 4 received. Page 4 (last week of October) is missing. Nudge triggered for complete October statement.
Flag — page 4 missing
Gate 3 — Income Consistency
Do income figures across documents reconcile? · Cross-document comparison
✓
Payslip vs Form 16 consistency — within 2% tolerance3-month average gross salary from payslips: ₹1,18,400/month → annualised ₹14,20,800. Form 16 gross salary: ₹13,98,000/year (₹1,16,500/month). Discrepancy: ₹1,900/month (1.6%). Within 2% tolerance — explanation: one payslip month included a ₹5,000 ad-hoc allowance not present in all months. Reconciled.
Reconciled
✓
Payslip employer vs EPFO contribution recordPayslip employer: Infosys BPM Limited. EPFO contribution record pulled via UAN: employer is "Infosys BPM Ltd" (abbreviated). EPFO contributions present for 36 consecutive months from the same employer. Salary band in EPFO: consistent with payslip amount. Pass.
Verified
⚑
Bank statement salary credit vs payslip net — 22% gap requires explanationNet take-home per payslip: ₹88,200/month. Average salary credits in primary bank account (6-month average): ₹68,400/month. Gap: ₹19,800 (22%). This exceeds the 10% tolerance for unexplained discrepancy. Possible explanations: split salary between two accounts, partial salary advance repayment, salary held by employer and paid in arrears. Requires borrower clarification before credit review. Nudge triggered: "We see a difference between your take-home salary and the salary credits in your bank statement. Please let us know if your salary goes into more than one account, or upload a second bank statement."
Flag — clarification needed
Gate 4 — Identity and Authenticity
Do documents belong to this applicant? Are there authenticity signals to note?
✓
PAN on all income documents matches application PANApplication PAN: ABCDE1234F. PAN visible on: Form 16 (ABCDE1234F — match), ITR acknowledgement (ABCDE1234F — match), bank statement header (ABCDE1234F — match). No PAN mismatch across any document.
Consistent
✓
Document digital signatures / watermarks where expectedBank statement: HDFC Bank official PDF with digital watermark detected. Form 16: TAN-linked digital signature present. Payslips: Infosys letterhead detected. No watermark or signature anomalies flagged. Payslip font and layout consistent across all three months — no formatting change flag.
Authentic signals
⚑
Address consistency — bank statement address differs from application addressApplication address: 42, 3rd Cross, Indiranagar, Bengaluru. Bank statement registered address: 14, Koramangala, Bengaluru (different area). Note: borrower may have recently moved and not updated bank KYC. Not a fraud signal — flagged as a consistency note for the credit analyst. Address proof document (utility bill) confirms current address.
Note — possible address update
Today's completeness check outcomes across the full application pipeline
Completeness Dashboard — Nov 14, 2025 · 84 applications in pipeline
All documents checked on receipt · Status current as of 09:00
31Complete — cleared for credit review
28Awaiting borrower action (nudge sent)
18Flagged — analyst clarification needed
7Incomplete — critical document missing
Most common gate failures today
Property document missing (home loan applicants)11 applications
Bank statement income vs payslip gap >10%9 applications
Bank statement page incomplete / partially uploaded8 applications
ITR acknowledgement older than 18 months6 applications
Rental income not corroborated in bank statement4 applications
31Files cleared for credit review today — every one complete, current, and internally consistent
4Gate categories — document presence, currency, income consistency, identity and authenticity
28Applications awaiting borrower action — nudge sent specifying exactly what is needed, why, and how to submit
ZeroIncomplete files reaching the credit analyst desk — the completeness gate is upstream of credit review, not parallel to it
A complete file at the credit desk is not an operations achievement — it is a credit quality decision
Every incomplete file that reaches the credit analyst adds latency: the analyst stops, flags the gap, the ops team contacts the borrower, the file goes back, it comes forward again. On a 50-application-a-day pipeline, if 30% of files have a completeness issue that the credit analyst discovers rather than the document ops team, that is 15 interruptions per analyst per day — each taking 15 minutes to document and manage. That is nearly 4 hours of analyst time daily that produces no credit decisions. The Document Ops Agent AI's completeness check is not a quality filter — it is the recovery of 4 hours of credit analyst time per day that can be spent on actual underwriting.
