AI-powered fine processing from hours to seconds

Challenge

As a car rental operator, Poppy receives a constant stream of traffic and parking fines tied to vehicles across its fleet. Each fine arrives as a PDF document that must be opened, reviewed, and manually cross-referenced with rental records to identify which renter was driving at the time of the violation.

This manual process consumed hours of administrative time each week. Staff had to extract license plate numbers, fine amounts, and incident dates by hand, then search through booking records to find the matching renter. Errors were common. Wrong assignments, missed fines, and delayed forwarding created friction with customers and revenue leakage for the business.

Poppy needed an automated system that could handle the full pipeline: receive a fine, extract the relevant data, match it to the correct renter, and forward it, all without human intervention.

Solution

We built an AI-powered pipeline that processes fines end-to-end. Incoming PDF documents are first passed through Google Vision API for optical character recognition, converting scanned documents into structured text. OpenAI then extracts the key fields (license plate number, fine amount, and incident date) from the recognized text.

The system automatically matches extracted license plates against Poppy's rental records to identify the renter who had the vehicle at the time of the violation. Once matched, the fine is forwarded to the correct renter instantly.

Key Features

• PDF ingestion: Automated intake and processing of fine documents
• Google Vision OCR: Converts scanned PDFs into machine-readable text
• OpenAI extraction: Identifies license plates, amounts, and dates from unstructured text
• Automated renter matching: Cross-references extracted data with rental records in real time
• Instant forwarding: Matched fines are sent to the correct renter without manual steps

Results

• Hours to seconds: Fine processing time reduced from hours of manual work to near-instant automation
• High accuracy: Automated extraction and matching maintain reliable results across fine formats
• Eliminated manual bottleneck: Staff no longer spend hours reviewing PDFs and searching rental records
• Faster renter notification: Fines reach the correct renter within seconds of being received