Cipla Cuts Manual Invoice Effort by 82% Across Five Languages

Cipla is an Indian multinational pharmaceutical company headquartered in Mumbai, founded in 1935, with a presence in 80+ countries. Its shared-services accounts payable team works with 2,000+ active vendors and handles tens of thousands of invoices a month, in five languages and through four channels. To standardize that work across regions, the company implemented Neil, E42’s AI co-worker.

The Challenge

Every invoice at Cipla started the same way: someone read it and typed it in. The volume ran to tens of thousands a month, split across five languages and four entry points (CIMMRA, SAP Ariba, email and a document system), and none of it looked alike. Many files were multi-page PDFs with several invoices packed into one. Others were scans marked by noise, stamps and dot-matrix print, which caused OCR errors. Manual entry was slow and error-prone, and with no batch-level or transaction-level quality check, validation varied from one region to the next. The result was delays, data discrepancies and a shared-services operation that did not work the same way twice.

The Solution

Cipla connected Neil, powered by E42’s Intelligent Character Recognition, to CIMMRA and Ariba. Neil captures multilingual data from multi-page, multi-invoice PDFs and verifies each invoice in real time across CIMMRA, SAP Ariba and the backend systems. It catches duplicates, validates invoices again after posting, and learns from every correction a person makes. With every channel under one shared-services framework, Cipla standardized its AP process, tightened control and made its data audit-ready.