← Our Work
AI Engineering
Document AI for OneTwo Home Finance.
ML-powered document classification inside a live loan approval workflow.
01/ the brief
High volumes. Inconsistent formats. A manual process that couldn't keep up.
OneTwo Home Finance is a digital-first home lender built on speed and simplicity — qualities that were being undermined by their document handling process. Payslips, bank statements, and IDs arrived in inconsistent formats, and every document required manual sorting, review, and data extraction before a loan could progress.
- Staff spent significant time manually sorting and categorising incoming loan documents
- Extracting and validating key data fields required repeated manual review
- Processing delays created bottlenecks in approvals — affecting both customer experience and compliance
- Manual workflows increased the risk of human error and inconsistent verification
02/ the solution
Automated classification, extraction, and routing — across the entire document pipeline.
Crystal Delta built an AI-powered document processing pipeline that replaced manual triage with automated classification, data extraction, and workflow routing — integrated directly into the live loan assessment process.
What's under the hood
- ML models automatically identify, classify, and route every incoming document type
- AI-driven extraction of critical data fields — payslip figures, statement balances, ID details — without manual review
- Workflow integration routes validated documents directly into the loan assessment pipeline
- Audit-ready output at each stage to support compliance and verification requirements
03/
The Impact
Manual Document Triage
Eliminated
Manual document triage across the loan approval workflow.
Faster approvals
Automated classification and extraction accelerated review cycles end-to-end.
Stronger compliance
AI-driven consistency reduced human error and improved regulatory adherence.