Credit File Archiver

Solving Pre-Loan Archiving Challenges

During the pre-loan phase of credit approval, operations teams often struggle with disorganized archive packages. Manual archiving faces two core issues: mixed entities (individuals vs. corporations) and complex classification rules. For instance, a "resume" must be filed under "Other Attachments" rather than "Income Proofs," while "proposed collateral" must be strictly separated from "existing debt." Manual verification is time-consuming and prone to misclassification or missed statistics.

Core Workflow & Classification Strategy

This skill uses triple-entity identification (package name, file name, and file content) to distinguish between individual and corporate entities, creating independent top-level folders for each. To balance speed and accuracy, it employs a tiered processing strategy:
- Low-ambiguity files (e.g., clear ID cards, business licenses) are archived directly by file name.
- High-ambiguity files (e.g., purely numeric PDFs, conflicting contracts) trigger content reading for dual verification. For example, a mismatched "Loan Contract.pdf" is automatically corrected to "Bank Statements" or "Debt Contracts."
- Unreadable files (encrypted, scanned) are routed to a pending-review folder and flagged in the File Name Anomaly List.xlsx.

Additionally, the skill automatically parses bank statements to calculate total inflows, outflows, balances, and transaction counts. It groups debt contracts into specific loan folders based on bank names, amounts, and dates extracted from file names and content. The final output includes 3 core Excel dashboards (statements, contracts, and a tree-structured master list) plus the anomaly report.

Boundaries & Considerations

• Local Execution: All content parsing is strictly local; no data is uploaded to external services. Original files are copied, not moved.
• Scope Limitation: The skill is strictly for pre-loan document organization, excluding mid-loan or post-loan materials like approval opinions.
• Shared Files Handling: Multi-person documents (e.g., household registers) are only placed in the primary borrower's directory. Other related directories do not physically copy the file but reference it in the master list.
• Manual Checkpoint: The workflow includes a mandatory Phase 2 confirmation where the client only verifies the entity list (individual/corporation) and output path. All content identification and archiving are handled automatically by the AI.