AI FEATURE · BORROWER JOURNEY
EVERY LENDER DOCUMENT, UPDATED FROM THE LOS — AUTOMATICALLY.
Doc Annotator reads a filled sample PDF, detects field positions, and suggests the correct LOS variable template for each field — building the configuration that generates every subsequent document without manual entry.
THE PROBLEM
LENDER-ISSUED DOCUMENTS ARE CONFIGURED ONCE — AND A SINGLE MISCONFIGURED FIELD GENERATES ERRORS AT SCALE.
Every loan triggers one or more lender-issued documents — a Key Fact Statement before execution, a loan agreement at disbursement, NOC letters at closure, settlement letters when accounts are resolved. Under RBI's Digital Lending Master Directions, the KFS alone must be provided to every borrower before the loan is executed, in a prescribed format, carrying accurate per-loan data.
These documents start as blank PDF templates that an implementation team must configure: each field in the PDF has to be mapped to the right LOS variable, positioned to print within its box, and verified against real loan output before the template runs at volume. In most NBFCs, this is done manually — by operations staff who must simultaneously understand the document's layout and the full LOS variable dictionary. A wrong binding does not surface as a system error. It surfaces as an incorrect document sent to a borrower, or flagged at an RBI examination.
SILENT MISCONFIGURATION
A wrong variable binding generates without error. The document looks correct until a field value is wrong — discovered by the borrower or during audit, not at setup.
FORMAT CHANGES MEAN FULL REWORK
When RBI revises the KFS format or a lender updates their agreement template, the coordinate configuration must be rebuilt from scratch — no differential update path.
DUAL KNOWLEDGE REQUIREMENT
Configuring a template requires someone who understands both the PDF layout and the LOS variable namespace. That combination is rarely one person's skill set.
LANGUAGE VARIANTS DOUBLE THE EFFORT
RBI requires KFS in the borrower's preferred language. Each language version is a separate template, multiplying the configuration effort with each language added.
HOW IT WORKS
UPLOAD A SAMPLE. THE SYSTEM MAPS THE FIELDS.
The annotation workflow runs once per document type. After setup, every loan generates the document automatically from the confirmed template — no manual entry per loan.
1
SET UP THE DOCUMENT
Name the document, upload the blank template, and set the language — including any regional variants that need to be linked to it.
DOCUMENT SETUP
2
DETECT THE FIELDS
Upload a filled sample and the system locates every field on the page automatically, with no manual measurement required.
AUTO-DETECTION
3
AI SUGGESTS THE MAPPING
Reading each field's label and context, the AI ranks the most likely LOS variable and populates the configuration automatically.
AI MAPPING
4
REVIEW AND CONFIRM
Accept a suggestion, pick an alternative, or enter a custom expression. Lower-confidence fields are flagged for a closer look.
HUMAN REVIEW
5
PREVIEW AND ACTIVATE
Verify the output on a sample document, then activate it. Every subsequent loan generates the document automatically.
PREVIEW & PUBLISH
PLATFORM INTEGRATION
EVERY TEMPLATE FIELD DRAWS FROM THE SAME DATA YOUR LOAN BOOK RUNS ON.
Doc Annotator's variable templates reference live OneFin LOS data objects — the same records that power underwriting, disbursement, and servicing. There is no secondary data sync, no export step, no reconciliation between the document system and the loan system.
🔗
CONNECTED TO YOUR LOS AND LMS DATA LAYER
Variable templates point directly to live loan data objects: approved_loan, loan_request, profile, payload. When a document is generated for a loan, the system reads from the live record at that moment — borrower name, account ID, and sanctioned amount are drawn directly, not from a pre-generated export or a separate document database.
LOS
LMS
DOCUMENT VAULT
KEY CAPABILITIES
WHAT THE SYSTEM DOES — AND HOW.
.webp)
COORDINATE AUTO-DETECTION
Upload a filled sample PDF and the system detects field positions automatically — capturing bounding box coordinates for every annotated field without manual measurement. No pixel-counting, no trial-and-error positioning.
.webp)
AI VARIABLE SUGGESTION
For each detected field, the AI reads the surrounding label text, table context, and document section, then ranks LOS variable templates by relevance. The top suggestion populates the configuration, with a ranked list of alternatives retained per field.
.webp)
RANKED ALTERNATIVES AND CUSTOM EXPRESSIONS
Each field carries the AI's full ranked list of alternative templates. From the annotator, the reviewer can select any alternative or enter a custom Jinja-style expression — including conditional logic for fields that vary by loan product or language.
.webp)
MULTI-LANGUAGE DOCUMENT SUPPORT
Vernacular document variants are linked to the English master via the Parent Doc field. The annotation workflow supports multiple language outputs from a single template structure — relevant for NBFCs serving regional borrower segments that require KFS in the borrower's preferred language
SAMPLE PDF PREVIEW BEFORE ACTIVATION
.webp)
Before a template goes live, generate a sample document using test values from the configuration. Verify every field populates correctly in the actual rendered PDF — not a screen preview — before activating for live loans.
.webp)
EVENT LOG AND AUDIT TRAIL
Every coordinate generation run, template generation event, and configuration change is logged with a timestamp and user ID — accessible from Event Executor Logs and per-document Audit Trails. The full configuration history is available for internal audit and RBI examination.
USE CASES
BUILT FOR EVERY DOCUMENT THE LOAN REQUIRES.
Doc Annotator can be configured for any PDF-based lender document. These are the document types used across OneFin's lender base.
REGULATORY MANDATE
KEY FACT STATEMENT (KFS)
RBI's Digital Lending Master Directions require a KFS to be provided to every borrower before loan execution — in a prescribed format, carrying accurate per-loan data including APR, repayment schedule, fees, and grievance mechanism details. Doc Annotator automates the template configuration that makes accurate, per-loan KFS generation possible at scale.
loan_account_no
borrower_name
sanctioned_amount
interest_rate
loan_account_no
processing_fee
DISBURSEMENT
LOAN AGREEMENT
The core legal document executed at disbursement. Variable bindings cover borrower identity, loan terms, applicable interest rate, repayment schedule, and lender details — drawn from the approved loan record. Conditional expressions handle loan-product-specific clauses such as stage-wise disbursement schedules.
borrower_name
loan_agreement_id
disbursement_schedule
lender_name
tenure
LOAN CLOSURE
NO-OBJECTION CERTIFICATE (NOC)
Issued at loan closure or on partial prepayment to confirm the lender has no further claim on the borrower or the collateral. Accurate generation requires the borrower's name, account reference, closure date, and collateral release details — all verifiable from the LOS record at the point of generation.
borrower_name
loan_account_no
collateral_details
closure_date
ACCOUNT RESOLUTION
SETTLEMENT LETTER
Issued when a loan is settled outside the standard repayment schedule. Requires precise figures for outstanding principal, the settlement amount agreed, and any waiver — all verifiable from the LOS record. Errors in settlement documents carry direct legal and credit bureau reporting consequences.
outstanding_principal
settlement_amount
waiver_amount
settlement_date
borrower_name