Approach
NSquare Gen AI deployed a production-grade invoice extraction engine with dual-OCR strategy, LLM validation, and spatial intelligence. The system preserves layout context through bounding-box alignment, enabling precise extraction.deployed a production-grade invoice extraction engine with dual-OCR strategy, LLM validation, and spatial intelligence. The system preserves layout context through bounding-box alignment, enabling precise extraction.deployed a production-grade invoice extraction engine with dual-OCR strategy, LLM validation, and spatial intelligence. The system preserves layout context through bounding-box alignment, enabling precise extraction.deployed a production-grade invoice extraction engine with dual-OCR strategy, LLM validation, and spatial intelligence. The system preserves layout context through bounding-box alignment, enabling precise extraction.
Over 250 finance hours reclaimed
99.9% field-level extraction accuracy
Zero human extraction errors
Processing reduced from days to minutes
Iyuno: Turning Manual Invoice Chaos into Zero‑Touch Finance Operations
Iyuno is a global leader in localization, dubbing, and media services, supporting the world's largest entertainment companies across film, television, and streaming.
With 30+ offices worldwide, Iyuno manages complex, high-volume intercompany transactions that span regions, currencies, formats, and internal systems. As the company scaled through mergers and global expansion, finance operations inherited a fragmented, manual workload — especially around invoice processing and validation.
The Problem: The Operational Reality of Finance teams behind Invoice Processing
Iyuno’s finance team processed approximately 400 invoices per month, primarily received as PDFs via email from multiple global entities. Each invoice followed a different structure and layout, requiring manual extraction of invoice numbers, PO numbers, line-item amounts, and currency details.
Despite the apparent simplicity of the task, invoice processing required significant manual effort. Finance staff had to read each PDF, identify key fields, validate extracted values against internal systems, compile spreadsheets, and route invoices downstream for approval and posting.
At scale, this process resulted in :
Processing cycles measured in hours per batch and days end-to-end
High operational overhead for skilled finance staff
Human extraction errors entering downstream systems
Limited ability to scale without increasing headcount
Iyuno needed a way to eliminate manual invoice processing while maintaining
enterprise-grade accuracy, control, and auditability.
Why Traditional Automation Wasn’t Enough
Iyuno didn’t need a basic OCR script or a template-based solution that only worked on clean invoices. The variability of layouts, currencies, and line-item structures made brittle automation impractical.
What Iyuno required was a system that could:
Handle any invoice format
Preserve layout and positional context
Achieve accuracy levels finance teams could trust
Operate securely within enterprise infrastructure
Scale globally without adding headcount
The Solution: Designing a Production-Grade Extraction Engine
N Square AI approached the problem as a systems challenge, not a single-tool deployment.
Invoices are automatically ingested from a single monitored inbox and converted into page-level images to maximize OCR performance. Instead of flattening invoices into plain text, the system preserves bounding-box data, enabling precise alignment between PO numbers, line items, and amounts.
A dual-OCR strategy cross-validates extracted fields, while Gemini Pro performs structured reasoning to reconcile outputs and validate accuracy. This multi-stage approach addressed common OCR edge cases and ensured consistent, reliable extraction across invoice types.
Once validated, data is automatically structured and matched against internal systems. Any anomaly or missing field triggers human-in-the-loop escalation, ensuring safety without slowing standard processing.
Outcoomes
Following deployment, Iyuno achieved 99.9% field-level extraction accuracy, eliminating human extraction errors and reducing invoice processing time from days to minutes.
More importantly, the solution reclaimed over 250 finance hours, allowing teams to redirect effort toward higher-value financial work rather than repetitive manual tasks.
This was not a pilot or proof of concept. The system operates as a production-grade component of Iyuno’s finance infrastructure, delivering repeatable, auditable outcomes at scale.
Invoice processing is one of the most underestimated sources of operational friction in finance. High volume, high variance, and low strategic value make it an ideal candidate for automation when done correctly.
By partnering with TrueHorizon AI, Iyuno eliminated an entire category of manual work while maintaining the accuracy, control, and trust required by enterprise finance teams.
The result is not just faster processing, but a more resilient and scalable finance operation.