Document Intelligence
Loan document intelligence: every file split, classified, and extracted.
Kaaj ingests deal packages from email, portal, drive, zip, or API, splits combined PDFs, and classifies 40+ lending document types, including handwritten applications. Files are renamed to your convention, missing documents and duplicates are flagged, and the data lands in your CRM or LOS before anyone opens a PDF.
Book a demo40+
lending document types classified automatically
5–7 minutes
to process a typical application package end to end
5
intake channels: email, portal, API, cloud drives, and zip archives
One broker email in, one decision-ready deal file out
Real submissions are combined PDFs, phone photos, zips, and duplicates. Kaaj turns them into a clean, named, complete file and tells you exactly what is missing or needs a second look.
Broker email · 5 attachments
“New deal: Northwind Freight, see attached”
- 48 pages
- phone photo
- 6 files
- 31 pages
- 2 pages
Organized deal file
split · classified · renamed · checkedFrom intake to your CRM, step by step
Every stage runs automatically; people only see the exceptions.
| Stage | What Kaaj does |
|---|---|
| Intake | Packages arrive by email forwarding, portal upload, API, Google Drive, Dropbox, or zip archive; paper and handwritten applications can be scanned or photographed. |
| Split | Combined PDFs are broken into individual documents so one upload becomes a structured deal file. |
| Classify | Each document is identified across 40+ lending document types. |
| Rename and file | Files are renamed to your naming convention and organized into folders, so names stay clean for audit and securitization. |
| Completeness | The package is checked against your checklist for the deal type, and missing documents and unsigned applications are flagged. |
| Extract | Business, owner, EIN, address, loan amount, and document fields are extracted; handwritten fields get confidence flags for human review. |
| Deduplicate | Duplicate files, resubmissions, and duplicate applicant records are flagged instead of silently overwritten. |
| Sync | Accounts, contacts, opportunities, and asset records are created or updated in Salesforce, HubSpot, LeasePath, or your LOS. |
40+ lending document types, built for commercial and SMB credit
Not mortgage forms or generic invoices: the documents equipment-finance, MCA, and SMB lenders actually receive.
| Document | What Kaaj extracts |
|---|---|
| Credit applications | Business and owner details, EIN, address, loan request, signatures; handwritten dealer applications included |
| Bank statements | Account, period, balances, and every transaction, ready for cash-flow and fraud analysis |
| Tax returns | Business and personal returns standardized for spreading and revenue checks |
| Financial statements | Income statements and balance sheets normalized into ratios, including interim statements |
| Invoices and equipment quotes | Vendor, line items, hard vs. soft costs, VIN and vehicle class, billing and shipping addresses |
| Driver's licenses | Name, date of birth, address, license number, and dates for identity checks |
| Voided checks | Account holder, bank, and account details for funding and consistency checks |
| Insurance certificates | Insured business and coverage details |
| Credit and business reports | Key figures summarized from bureau and business credit reports |
Loan-package intelligence vs. general document AI vs. mortgage tools
General document AI platforms are toolkits. Mortgage document tools are built for consumer mortgage files. Kaaj is built for commercial and SMB loan packages, end to end.
| Kaaj | General document AI | Mortgage document tools | |
|---|---|---|---|
| Built for | Commercial, SMB, equipment-finance, and MCA loan packages | Any document workflow; you build the lending logic | Consumer mortgage files |
| Messy intake | Email, portal, API, drives, zips; combined PDFs split automatically | Depends on your integration | Mortgage LOS and borrower portals |
| Handwritten applications | Parsed, with low-confidence fields flagged and signatures checked | Varies by model | Varies |
| Invoices and equipment | Hard vs. soft costs, VIN, vehicle class, addresses | Generic invoice models | Not the focus |
| After extraction | KYB, fraud checks, cash-flow analysis, and a credit memo in the same workflow | Structured data for your own systems | Mortgage underwriting and compliance |
| Where results go | Salesforce, HubSpot, LeasePath, or your LOS | Your data pipeline | Mortgage LOS |
From chaos to organized
Ingest from email, portal, Google Drive, Dropbox, APIs
AI-powered document classification for 40+ document types
Auto-rename files from "scan_final_v3.pdf" to "2025-01_Chase_Statement_p1-12.pdf"
Detect missing required documents based on deal type and policy
Flag duplicate submissions and resubmissions
Extract key fields: business name, EIN, dates, amounts
Document intelligence FAQ
What is document intelligence for lenders?
Software that turns a messy loan package into structured data: it receives documents from email, portals, or APIs, splits combined PDFs, classifies each document, extracts the fields underwriting needs, flags missing items, and syncs the result to your CRM or LOS.
What documents can Kaaj process?
40+ lending document types, including credit applications (handwritten ones too), bank statements, tax returns, financial statements, invoices and equipment quotes, driver's licenses, voided checks, insurance certificates, and credit reports.
Can Kaaj split one large PDF into separate documents?
Yes. Combined PDFs are split into individual documents, each classified and renamed, and duplicates within or across submissions are flagged.
How does Kaaj handle handwritten applications?
Handwritten and scanned applications are parsed like digital ones. Fields read with low confidence are flagged for human review, and Kaaj checks whether the application is signed.
How are files named and organized?
Kaaj renames every file to your naming convention and organizes folders by deal, so filenames are consistent for audit, securitization, and search in Salesforce or your document store.
How long does document processing take?
A typical application package is processed in 5–7 minutes, from intake through classification, extraction, verification, and analysis.
Which intake channels are supported?
Email forwarding, a broker or vendor portal, API submissions, Google Drive, Dropbox, and zip archives. Paper, email, and digital applications can run side by side.
Is Kaaj a general document AI platform like Hyperscience or Google Document AI?
No. General document AI platforms are toolkits for any document workflow. Kaaj is built for commercial, SMB, equipment-finance, and MCA loan packages, and carries the extracted data straight into KYB, fraud checks, cash-flow analysis, and the credit memo.