Automation agent · Operations and back office
Reads scanned forms and documents and fills the fields your system needs
The document extraction agent reads scanned KYC papers, claim forms, bills and certificates, pulls out the fields your process needs and checks them against each other. Anything it cannot read clearly, or that does not add up, goes to a named reviewer before it enters your system.
- Type
- Automation agent
- Team
- Operations and back office
- Connects to
- 6 systems, including Loan management system
- Limits
- 5 set, plus approvals
How it is done today
The work before the agent
At a two wheeler loan NBFC in Ahmedabad, dealers upload PAN cards, masked Aadhaar, salary slips and bank statements as phone photos and PDFs. A team of data entry operators types names, dates of birth, addresses and incomes into the loan system, then a second person checks the typing. A blurred photo is often found only days later, when the file is already with credit.
What the document extraction agent takes on
Sorts what arrives
Recognises each document type in a mixed upload, such as PAN, masked Aadhaar, salary slip, bank statement or hospital bill, and files it against the right application or claim.
Pulls out the fields
Reads name, date of birth, PAN number, address, employer, monthly salary, bill amounts and dates, including from photos taken at an angle, and marks how confident it is about each field.
Cross checks the documents
Compares the name and date of birth across PAN and Aadhaar, matches salary credits in the bank statement against the slip, and flags gaps such as a missing month or an expired document.
Fills the system and routes exceptions
Writes clean fields into the loan or claims system, and sends low confidence fields and mismatches to a review queue with the image and the field side by side.
How the work moves
A typical piece of work, stage by stage
- Step 1Document arrivesScan, photo, PDF
- Step 2Text readOCR and layout
- Step 3Fields extractedInto your format
- Step 4Checked against rulesLow confidence to a person
- Step 5Sent to your systemAfter approval
Where a person decides
A named reviewer, for example Rakesh in credit operations, sees every field the agent marks as uncertain and every mismatch between documents. He confirms or corrects it before the file moves to credit. The agent prepares the data; it never approves a loan or a claim.
Limits it works within
- Stores only masked Aadhaar numbers and never writes the full number anywhere
- Uses documents only for the purpose the customer gave them for, in line with the DPDP Act
- Fields below the confidence level you set always go to a person
- Can run in your private cloud or on premise so scans stay in your environment
- Keeps a log of each field, its source image and who confirmed it
Connects to
Works with the systems you already use
Through APIs, exports or a custom connector our team builds. The exact list depends on your versions and setup.
- Loan management system
- Google Drive
- Amazon S3
- Gmail
- WhatsApp Business API
- PostgreSQL
Teams that use it
Watch it work
One invoice, from inbox to Tally
Fields read, each checked against your rules, the exception sent to a person, and the entry posted only after approval. Sample data.
Questions
About the document extraction agent
Does it work with handwritten forms?
It reads clear handwriting, and handwritten fields are the ones most often sent for review. We test on a sample of your own documents before go live, so you see the results first.
Where do the scanned documents stay?
Wherever you decide. For KYC and health records, most clients keep the agent in their own cloud account or on premise, so scans never leave their environment.
Can it handle documents in Gujarati or Hindi?
Printed Gujarati and Hindi text can be read, depending on the model chosen for your project. We check this on your samples during setup and tell you plainly which documents need a person.
Keep going
Other agents and reading
Automation agent
Support agent
Answers routine order questions on WhatsApp and hands the rest to your team
Learn moreAutomation agent
Invoice matching agent
Matches supplier invoices to purchase orders and goods receipts before they reach Tally
Learn moreGuide
Document extraction accuracy: what to measure and how
Vendors quote accuracy figures for document extraction, but one number hides what matters. Measure extraction on your own documents, field by field, and design the process so the inevitable errors are caught.
Learn moreSee it on your work
Bring one slow, repetitive process. We will show you an agent doing it.
A working session with someone who builds these every week.
Or get a ballpark proposal in two minutesWhat happens next
- Step 1A short call about one process and the systems involved
- Step 2A demo of an agent working on a version of your task
- Step 3A clear view of fit, approvals and what it would take