Over the past several weeks, I designed and delivered a complete AI - powered pipeline that extracts structured data from unstructured rate confirmation PDFs. This project combines the power of OpenAI with Google AppSheet, Google Apps Script, and GCP Cloud Run services to deliver automation in logistics.

The core idea was to allow logistics managers to upload Rate Confirmation documents through an AppSheet form. The system parses the content, sends it to OpenAI with a strong prompt + schema, and returns clean structured data directly into Google Sheets, updating the AppSheet dashboard in real-time.

We migrated services to production, set up custom domains, created clean API contracts, tested parsing with real -world Rate Confirmation files, and built a fallback image-to-text solution. Every step included structured validation to minimize hallucinations and ensure cost-effective API usage (switch to GPT-4o-mini).

Our OpenAI prompt and schema were heavily iterated to ensure accuracy and flexibility — covering pickup and delivery data, contact info, reference numbers, and even dynamic delivery instructions. All work is versioned, deployed via CI/CD on Google Cloud Run, and integrated into AppSheet UI.

This is one of the most impactful automations we’ve delivered — it saves hours of manual work per day and lays the foundation for smarter logistics tools.