How We Built an AI-Powered Media Invoice & Proof Validation Platform for the MENA Region’s Leading Media Consultancy
A Mountain of Media Invoices
The client
For one of the region’s top media consultancies — trusted by major advertisers across the MENA region — invoice validation had quietly become a nightmare.
Every month, their teams had to review thousands of invoices and accompanying proofs — more than 3,000 third-party documents and photographs — all coming from different suppliers, formats, and regions.
Each invoice needed to be verified against multiple business rules and cross-checked with evidence. What sounded simple on paper had turned into a time-consuming, error-prone, and resource-heavy process.
Complexity Hidden in the Details
There were three key players in this process:
The Owner
The advertiser or brand hiring the media agency.
The Media Company
Responsible for validating the invoices.
The Supplier
Instant financial visibility and insights for management.
Each validation involved over 1,000 conditions, spanning invoice templates, formats, amounts, dates, and supporting photographs.
The client wanted a system that could:
- Read and interpret scanned invoices in multiple formats
- Automate validations across thousands of documents
- Allow human intervention when necessary
- Provide role-based dashboards and reporting for each stakeholder
Bringing AI into the Process
When the client realised that no off-the-shelf product could handle such complexity, they partnered with us to build an AI-driven solution from scratch.
We began by mapping every stage of their validation process — from supplier submission to final approval — and then looked for opportunities to blend automation with human oversight.
The core idea was simple:
“Let AI handle the repetitive checks, and let humans make the smart decisions.”
That guiding principle shaped everything that followed.
Bello — The Smart Media Validation Engine
We developed Bello, an intelligent, end-to-end media validation platform designed to handle thousands of invoices, proofs, and validations seamlessly each month.
AI-driven invoice reading using a custom Python file reader integrated with Google Vision API
Over 1,000 built-in rules and conditions for automated validation
Custom-trained open-source LLM model for error detection and contextual data checks (reducing external API costs)
Role-based dashboards for all user levels — Owner, Media Company, and Supplier
Analytics and report generation to visualize validation progress and results
Intervention capability for manual overrides where required
API-based backend architecture built for scalability and future integrations
Accuracy, Speed, and Control
Once Bello was deployed, the difference was dramatic.
Streamlined Automation
1,000+ validations processed every month with minimal human intervention
Operational Cost Reduction
Significant reduction in manual review hours and related costs
AI-Driven Precision
Improved validation accuracy through AI-powered checks
Accelerated Cycle Times
Cycle time reduced from weeks of manual verification to a few days
Real-Time Executive Insight
Top management gained real-time visibility via mobile dashboards and analytics
The Tech Behind Bello
The tech used
- Frontend: Angular
- Backend: API-based PHP architecture
- AI & Automation: Python, Google Vision API, custom rule-based engine, and fine-tuned open-source LLM model
- Database: MySQL
- Deployment: Cloud-hosted, modular microservice approach for scalability
This combination allowed the platform to process thousands of documents efficiently while keeping operational costs low and accuracy levels high.
Human-Assisted AI at Its Best
What began as a challenge of scale and accuracy became a showcase of how AI and human insight can work together.
Bello didn’t replace people — it empowered them.
By automating the repetitive parts of validation and surfacing insights instantly, the platform helped the client save time, reduce cost, and make better-informed decisions.
It’s a perfect example of how custom AI-driven systems can deliver real, measurable impact when designed around the way people actually work.