EBTS reduces invoice workload by 25% with AI

EBTS, a Belgian business services company, engaged ACA Group to build a custom AI system for automating invoice processing and translation, addressing the challenge of handling technical invoices in nearly all European languages with industry-specific jargon. ACA Group built the solution on Microsoft Azure using advanced AI models that function as an add-on in EBTS's backend without requiring changes to existing systems. The project was delivered in 22 days, against an initial estimate of 28 to 39 workdays. The system uses a hybrid translation approach: whenever a technical term has been previously translated by EBTS experts, that translation is preserved as a memory; when no historical translation exists, large language models such as ChatGPT provide the translation. This combination of LLM capability with EBTS's domain expertise creates a system that is both highly accurate and consistent. The AI extracts key data from invoices and eliminates irrelevant information, reducing the invoice workload by 25%. The solution's modular design — functioning as a backend add-on — meant minimal disruption to existing workflows and allowed the team to continue working with familiar technology while significantly reducing manual effort.