Land Record and KCC Loan Document Extraction

Read Khatauni, 7/12, Jamabandi and other land records plus KCC loan papers, extract holdings and crops, convert local area units, and flag mismatches before a credit officer sanctions.

Capability:: Document Intelligence
Industry:: Agriculture & Agritech

Book a 20-minute call WhatsApp us

Who it is for

Head of Agri Lending
Rural Branch Credit Manager
Credit Operations Lead, NBFC
Loan Processing Team Lead

The problem

A Kisan Credit Card or crop loan file depends on land records that look different in every state: a Khatauni in Uttar Pradesh, a 7/12 extract in Maharashtra, a Jamabandi in Punjab or Rajasthan, an RTC in Karnataka. Credit staff read these by hand, convert bighas and gunthas to acres, and check owner names against KYC before they can assess the loan.

State-by-state formats

Each state's record has its own layout, language and terms, so one checklist does not fit every branch.

Local area units

Holdings appear in bigha, biswa, guntha, kanal or hectare, and the size of a bigha differs between states and even districts.

Name and share mismatches

Joint holdings, transliterated names and inherited shares make it hard to confirm how much land the applicant actually holds.

Missed charges and mutations

Existing bank charges, pending mutations and other lenders' entries in the record are easy to overlook under volume.

What we build

We build a document pipeline for agri-loan files. It accepts scanned or downloaded land records and KCC application documents, identifies the state format, and extracts owner names, survey or khasra numbers, holding area, share, crop and any recorded charge or mutation. Area is converted to a standard unit using a district-level unit table your credit team approves. Names are matched against KYC with transliteration-aware comparison. The pipeline compares the holding and declared crops with the scale of finance used to size the limit, and flags gaps. Credit officers see every extracted value next to its source text, and any low-confidence field or failed check goes to them for confirmation.

Recognises common state land record formats and local-language text
Extracts owner, share, survey or khasra number, area, crop and charge entries
Converts bigha, guntha, kanal and other units using an approved district table
Transliteration-aware name matching against KYC documents
Checks holding and crops against scale of finance inputs
Source-linked review screen for credit officers

How it works

01Collect the loan fileLand records, KCC application, KYC and no-dues certificates arrive from the branch, a field agent app or a permitted portal download.
02Identify and extractEach document is classified by type and state format, and fields are extracted with a confidence score.
03Normalise units and namesArea is converted to a standard unit and names are matched across land record and KYC.
04Run credit checksThe pipeline checks holdings, shares, charges and crops against the application and scale of finance.
05Credit officer reviewLow-confidence fields and failed checks go to the credit officer with the source text highlighted.
06Send to the loan systemStructured data and the review trail are pushed to the loan origination system for sanction.

Land Record and KCC Loan Document Extraction: workflow diagram — Open full size

Architecture

Ingestion
- Branch scan and field app upload
- Permitted state portal download intake
- Loan-file grouping by application
Document understanding
- OCR for Devanagari and regional scripts
- State format classifier
- LLM field extractor with per-state schemas
Validation
- District unit conversion table
- Transliteration-aware name matcher
- Scale of finance and charge checks
Review
- Credit officer review screen with source highlights
- Query log for missing documents
Integration and audit
- Loan origination system adapter
- Audit log of values and edits

Inputs and outputs

Inputs

Land records such as Khatauni, 7/12 extract, Jamabandi, RTC or Patta
KCC or crop loan application form
KYC documents with masked Aadhaar
No-dues certificates from other lenders
Scale of finance table for the district and crop

Outputs

Structured holding record with owners, shares, survey numbers and area in a standard unit
Name match results across documents
Flags for charges, mutations and shortfalls against the requested limit
Field-level confidence with source links
Audit trail of extraction and edits

Where people stay in control

Credit officers confirm low-confidence fields, resolve name and share mismatches, and decide on every flagged charge or mutation. Sanction decisions stay entirely with the lender.

Integrations

Loan origination system via API or staging table
Core banking system
State land record portals, where access is permitted
Document management system
Field agent mobile app

Guardrails

Unit conversion only from a table approved by the credit team, with district overrides
No automated sanction or rejection
Masked Aadhaar only; full numbers never stored
Every value linked to its source text
Periodic re-check of auto-cleared files by credit audit

Compliance notes

Supports the documentation and audit trail lenders keep under RBI directions on the Kisan Credit Card scheme.
Aadhaar is handled in masked form per UIDAI rules, and personal data in line with the DPDP Act 2023.

A typical pilot

Duration: 3 weeks

KCC files using one state's land record format, drawn from a few branches and processed in parallel with current manual checks.

Deliverables

01Extraction schema and unit table agreed with credit
02Working pipeline on de-identified historic files
03Credit officer review screen
04Accuracy report against officer-entered data
05Plan for additional states

What we measure

Field extraction accuracy: Compared with the values credit officers entered for the same files.
Unit conversion correctness: Converted areas checked against manual calculations on a sample.
Flags confirmed as genuine: Credit officer feedback on each name, charge or shortfall flag.
Processing time per file: Timed on a sample of files with and without the pipeline.

FAQs

01Which states' land records can it read?

We start with the state formats that make up most of your book and add others one at a time. Each new format is tested on real samples before it is switched on.

02Can it fetch records from state portals?

Where your institution has permitted access, the pipeline can take portal downloads directly. Otherwise it works from scans or files uploaded by branches and field agents.

03How does it handle handwritten mutation entries?

Handwritten entries are extracted where legible, and anything uncertain goes to the credit officer instead of being guessed. The pilot report shows handwritten and printed accuracy separately.

04Does the AI decide the loan amount?

No. It prepares checked data and flags gaps against the scale of finance. The credit officer and your sanctioning authority decide the limit.

Related use cases

Founded by Akhilesh Mittal. 8 years building a security platform for a New Zealand company. IIT ISM Dhanbad and IMT Ghaziabad. Builds production agentic AI.

Start with one crop, one district or one document type

Tell us the workflow that costs your field, agronomy or credit team the most time. We will scope a short pilot on your own photos, calls or files.