Automated Parsing, Routing & Management for Corporate Documents

Context

The company faced operational chaos due to a high volume of incoming documents (resumes, invoices, contracts). Managers, HR, and legal teams spent hours manually reading files, transferring data across spreadsheets, and renaming scans. Without automation, critical contracts with hidden penalties slipped through, top-tier candidates were lost, and the file system devolved into a dump of poorly named documents like "scan_123.pdf".

Problem

• Manual Data Entry: Spending 15–20 minutes manually extracting and transferring information from each PDF into the CRM.
• Human Error & Blind Spots: Missing critical terms or penalties in multi-page contracts due to fatigue or oversight.
• File System Chaos: Lack of standardized document naming conventions, making it nearly impossible to locate files on Google Drive.
• Delayed Response Times: Bottlenecks in processing top candidates and routing non-standard requests.

Solution

Developed an intelligent, end-to-end agent designed to eliminate manual handling at every stage of the document lifecycle:

• AI-Powered Extraction & Qualification: An LLM automatically reads incoming files, identifies the document type, and extracts key entities (e.g., candidate scores out of 10, invoice totals, arrays of legal risks).
• Aesthetic File Management: The system automatically renames files in Google Drive using a standardized, visually clean format (e.g., [CV] - Name - Position.pdf).
• Visual Command Center in Notion: Built a centralized dashboard where resumes are displayed on a Kanban board, invoices on a payment calendar, and contracts grouped strictly by severity with red "High Risk" tags.
• Smart Routing & Escalation: Built-in deduplication logic prevents redundant records. A dedicated Gatekeeper instantly sends push notifications via Slack to the responsible managers (e.g., alerting HR to candidates scoring 8+, and warning legal about high-risk contracts).

Process

• System Architecture: Designed a branching workflow that categorizes three primary document types and routes non-standard files ("Other") for manual review.
• Prompt Engineering: Crafted strict LLM prompts to guarantee stable JSON data outputs, specifically handling complex risk arrays and boolean values.
• Data Transformation: Configured precise data mapping for complex types (stripping quotes from numbers, mapping booleans to checkboxes) for seamless Notion integration.
• Workflow Orchestration: Built and automated the entire end-to-end pipeline in n8n—from the Google Drive trigger to deduplication filters and Slack alerts.

Tech Stack