Case Study

APACS: Invoice Entry Automation with OCR

Client: Global Aviation Maintenance & Operations Firm

Problem: 10k Invoices every month for manual entry causes significant time wasted on managers.

The Solution: Zero-Training AI OCR

We deployed a "Human-in-the-Loop" automation engine.

• The Brain: We leverage n8n to perform data extraction and integration into database.

• The Extractor: Advanced AI OCR that requires zero template training—it reads any invoice format instantly.

• UI/UX: A custom React + Vite viewport that lets managers view live document next to AI extracted field to verify data in seconds, not minutes.

  
  

The Result: 600h+ Saved Every Month

We didn't just "speed up" the process; we broke the bottleneck. By shifting from a 5-minute manual slog to a 1-minute AI-assisted verification, the impact was immediate:

• The Math: 10,000 invoices x 4 minutes saved = 40,000 minutes saved per month.

• The Reclaimed Time: 600+ hours of pure management overhead returned to the business.

• Money saved: 6,000 MYR SAVED per month (Assume each hour cost 10 MYR on hires)

Features:

Flagship OCR: Works on multi-page hand writing, digital scan and e-invoice. All while no training needed, even if it is a totally unseen format or structure.

Document Viewport: A side panel that show the real document. This help the client verified the data by a human.

Security: Checkout Security Page

Stack:

• Automation: n8n

• Frontend UI/UX: React + Vite

• Storage: Microsoft Azure

  
  

  
  

Kingsman Realty: Real Estate AI Match Making

Client: Realty agency, managed properties with a total value of RM11 billion

Problem: Slow, manual property match making to their client.

The Solution: The AI Search Architect

Our system rate a score on all property based on the user prompt, with advance reasoning LLM based on Kingsman's core principle. This ranked system ensure no choices are left unconsidered.

The Result: 3 hour saved per usage.

Built one system, but shared with multiple agents. More time saved, more time for showroom.

Features:

• Contextual Matching: Understands complex human prompt, "quiet neighborhood next to a church" or "investment potential" using LLM reasoning.

• AI Recommendation: A side-by-side showing how the property matches the client's requirement or when it is not.

Stack:

• Automation: n8n (Real-time Lead Scoring)

• Frontend UI/UX: React + Vite

• Storage: Google Cloud

  
  

Atsource: B2C Manufacturer Database

Client: Multinational Manufacturing Material Sourcing Company, 130k+ follower on Instagram

Problem: Database require data cleaning, and needed to be displayed on a premium SaaS platform.

The Solution: Data Cleaning with LLM + SaaS Platform

With our custom automation, the data cleaning was done in hours instead of 100h+ manual data cleaning & entry.

• Dataset cleaning with custom automation and scripts

• Highly optimized SaaS Webapp

The Result: Enterprise Grade SaaS

• 100h+ Saved from manual data cleaning

• Custom High Converting Web App

  
  

Stack:

• Data Cleaning: n8n, Python Script

• Frontend UI/UX: React + Vite

• Storage: Supabase