Give loan operations consistent records as package documents arrive
For Loan operations teams coordinating document collection
Prepare each incoming document for a loan package register. A small intake schema can capture the document’s visible identity and dates before your application joins the result to the case.
25 free credits/month · No card needed
See the workflow and review checks{
"document_label": "Monthly account statement — synthetic",
"subject_name": "Example Borrower LLC",
"issuer_name": "Example Bank",
"document_date": "2026-08-01",
"period_label": "July 2026",
"visible_reference": "DEMO-STATEMENT-07"
}Synthetic data, not a live extraction. Real responses also include confidence, validation, and result status.
When this work lands on your desk
Documents arrive in several uploads for the same loan, and operations must determine which files are useful, duplicated or still outstanding.
A full inbox is not the same as a complete loan package. Filenames may be vague, dates may refer to different reporting periods and the same statement may arrive twice. Structured per-file records support a customer-owned checklist, while completeness still depends on your loan requirements and source review.
Start with the documents you already collect
- Individually prepared income documents
- Bank statements
- Application or supporting document PDFs
PDF, JPEG, PNG, TIFF, or WEBP. Send one file per extraction call; create separate schemas when the document types need different fields.
Define the fields
Open the starter schema. Edit field descriptions, mark the values your process requires, and save it in your account.
Extract at intake
Send the document and saved schema slug to
POST /api/v1/extract, or collect the file through a branded upload site.Hand off reviewed data
Your system adds the internal case ID, file identifier and received timestamp. It compares records with your required-document checklist, detects possible duplicates and sends uncertain cases to operations.
A schema shaped around the next step.
Start with these fields, then adjust them to match your documents and destination system. The example above shows the data object you are asking Dokyumi to extract.
Download editable starter schemaStarter fields are optional until you choose which are required. Nested line items need a defined structure if you want validation of their individual values.
| Field | Type | What to capture |
|---|---|---|
document_label | string | Visible document heading or title, not a verified document classification. |
subject_name | string | Primary borrower or business named in the file. |
issuer_name | string | Named bank, employer or other issuing organization. |
document_date | date | Issue or preparation date when explicitly printed. |
period_label | string | Stated reporting period without inventing missing dates. |
visible_reference | string | Document reference as printed; avoid unnecessary identifiers. |
Give exceptions a clear path to review.
Dokyumi returns a model-reported confidence map, validation details, and a completed or review status for produced results. Check required fields and your business rules before passing data onward.
A completed extraction describes the extraction result. Your application and reviewers own matching, approvals, routing, and updates to other systems.
Read the response contractChecks for this process
- A document names more than one borrower
- Period label does not cover the requested month
- A combined file contains several document types
- An absent date or heading requires source inspection
Prove the fit on a representative batch.
Choose a few authorized packages with known missing and duplicate files. Test the per-file extraction separately from your checklist logic and verify the final exceptions with operations.
Compare extracted values with a reviewer’s reference set. Record corrections, unresolved fields, review time, and credits used before connecting downstream actions.
The free plan’s credit limits apply. Assisted pilots are scoped with Dokyumi before work begins.
Expand when the next process is ready
Expand to paid volume when the register receives regular submissions. Add specialized schemas for the fields analysts need after the initial intake review.
For a tailored implementation
Discuss custom capacity and schema assistance when several loan products require different document sets and intake portals.
Talk through document mix and volumeStart small. Add capacity as intake grows.
All self-service plans include the REST API, upload-site webhooks, custom confidence thresholds, and schema inference. Choose by volume, schema count, and upload sites.
Each document uses one credit per 5 pages, rounded up. An 8-page file uses 2 credits. Self-service files can be up to 50 pages; larger files, custom-built schemas, and additional volume are quoted through Enterprise. Full pricing details
Questions about loan package intake
Does this automatically certify a complete package?
No. Your application owns the checklist and your reviewers confirm sufficiency. Extraction provides document-level facts to support that process.
Will the API split a mixed document bundle?
Do not assume automatic splitting. Prepare one intended file per extraction and select an appropriate schema for its contents.
Where does the internal loan ID come from?
Your application attaches its own case metadata. An internal loan ID should not be invented from the document when it is absent.
How do I test this workflow on my own documents?
Start free with 25 credits per month, 2 schemas, and 1 branded upload site. The starter schema opens in the schema builder after sign-up. Edit the fields, save it, and upload a representative file through your account or the extraction API. Account documents are stored according to Dokyumi’s privacy policy; contact hello@dokyumi.com for deletion requests and retention requirements.