Mortgage file processing (Canada)
Early access use case for extracting income, asset, property and identity data from mortgage application files for underwriting review.
Documents in
- Mortgage applications
- CRA T4
- T4A and T5 slips
- Notices of assessment
- Revenu Quebec slips
- Pay stubs
- Employment letters
- Bank statements
- Purchase and sale agreements
- Appraisals
- Credit bureau reports
- Identity documents
How it runs
-
Intake
Applicant documents are uploaded or sent via API from the broker or lender process and tracked per file.
-
Text extraction and OCR
Text is extracted from digital files and scans are read with OCR.
-
Classification and schema matching
Documents are classified and matched to a schema. The early-access mortgage pack supplies the document types and starter schemas.
-
Field extraction
Income, account, property and identity fields in the schema are extracted with confidence values.
-
Human review
Uncertain fields and unmatched documents pause for an underwriter or processor to correct and approve.
-
Summary and output
A summary is generated and structured results are delivered via API and webhooks.
What you get
- Structured file data with per-field confidence
- Applicant documents sorted by type
- Underwriter-confirmed income and property values
- Searchable file with summaries
Early access
Canada is opening gradually. The mortgage pack is built and deployed in early access. It ships document types, starter schemas, cross-document indexes, sample questions and smart folders, with English and French keywords and CAD by default. It reads CRA and Revenu Quebec slips. Social Insurance Numbers are reduced to the last 3 digits and account numbers to the last 4, and personal fields are flagged sensitive. The pack is built from official form guidance and has not been validated on production customer documents. Forms and fields are being confirmed with early customers. AI processing location is stated in the Order Form.
Where review fits
Income and identity values are reviewed by a person when confidence is low. Underwriters compare income, property and identity values using extracted data and search. The pack prepares and checks document data; underwriting, credit and lending decisions stay with the lender.