From Checking Documents One by One to Automatic Verification with Gemini AI

Every registration batch arrives with a stack of documents: forms, CVs, ID documents, and more. Every one had to be checked one by one, its data retyped, and the files then organized in Google Drive.

I built an automation system on Google Apps Script integrated with Gemini 2.5 Flash on Vertex AI. The AI validates the documents and extracts their data, then the script moves every file into organized folders automatically.

The problem: manual verification that was slow and error-prone

Verifying registration documents was done by hand, took a long time, and was prone to human error. The biggest challenges came down to two things: making sure the entered data matched the original documents, and organizing thousands of files in Google Drive.

In the manual era, the process needed four admin roles:

1. Admin 1 (Sorting): receives incoming data and sorts documents by hand.
2. Admin 2 (Validator): opens each file one by one to make sure the uploaded document is correct, for example a CV file that turns out to contain a passport photo instead.
3. Admin 3 (Collector & Editor): downloads the documents, renames each file one by one, then arranges them into folders.
4. Admin 4 (Data Entry): retypes document numbers, like NIK (Indonesia's national ID number) and diploma numbers, into the report format the relevant institution requires.

The approach: AI that reads, a script that organizes

I built the system on the Google Workspace the team already used, with Gemini 2.5 Flash as the "eyes" reading the documents. Gemini runs computer vision analysis to validate the documents and extract text from the images, while Google Apps Script handles migrating the files into an organized folder structure.

1. 1AdminRuns the system from a menu in Google Sheets
2. 2AI ValidationGemini checks that each document matches
3. 3AI ExtractionDocument numbers and personal data are pulled automatically
4. 4Auto MigrationFiles are renamed and moved to their destination folder

What I built

A custom menu in Google Sheets

I made sure admins wouldn't need to learn a new app. Every control lives in the AI Verifikator menu, right inside Google Sheets: verify or extract only the checked rows, verify or extract everything at once, and reset notes and extraction results.

AI that understands context, not just OCR

This is the core of the system. Gemini doesn't just read text, it understands what's in the document. If an uploaded file turns out not to be the document requested, the AI flags it right away with a note in the relevant cell.

Automatic data extraction into Google Sheets

Important data like document numbers and personal details is extracted automatically with high accuracy, then shows up on the Google Sheets dashboard without anyone retyping it.

Automatic file migration

Once verification is done, the script I built bulk-renames the files and moves them into folders organized by batch and document type, then compiles a summary table for reporting. No more files with random names straight out of Google Forms.

Results

Reduction in document verification time

70%

Document numbers retyped by hand

0

Registration batches handled systematically

Hundreds

• Processing time dropped 70%. Verification work that used to be done manually, one by one, now runs automatically.
• Accurate data extraction. Important data like document numbers and personal details is extracted automatically with high accuracy, thanks to the latest multimodal AI models.
• Scalable file management. Data clutter disappeared thanks to automatic file migration that can handle hundreds of registration batches systematically.