Mauritius Commercial Bank cuts customer onboarding time 64% with Camunda, IDP and AI

Mauritius Commercial Bank (MCB) was constrained by disconnected legacy systems and highly manual processes that added friction for customers, limited visibility and prevented end-to-end control over operations. Onboarding of new customers, particularly Know Your Customer (KYC) identity verification, was time-consuming, error-prone and operationally inefficient, and the lack of visibility made it hard to identify bottlenecks. Manual processes also made traceability and compliance harder and risked introducing errors and exposure to risk.

MCB began its transformation in 2018 by adopting Camunda to automate and orchestrate business processes and manual tasks. For customer identity verification it built reusable components on Camunda and added intelligent document processing (IDP) to extract structured data from unstructured uploads such as proofs of address and identity documents, AI/ML-powered models that assign confidence scores to scanned documents and classify them, and RPA to fill gaps where APIs were not available and to connect with third-party providers such as adverse media screening. Data from documents with high confidence scores is extracted automatically into the BPMN process, while low-confidence documents are routed to a human-in-the-loop for review and validation. MCB also uses Camunda to orchestrate KYC recertification updates every two years, again using IDP and automation.

What used to be a 70-minute onboarding process now takes approximately 25 minutes and is more traceable because of BPMN. The case study headlines a 64% reduction in customer onboarding time, and MCB's technical lead says employees see a 64% reduction in processing time per application, freeing them to focus on more productive tasks. It also lists reduced operational costs through decreased manual labor and enhanced compliance and auditability, without giving figures.

• Build identity-verification steps as reusable components on an orchestration layer so new capabilities such as document AI and RPA can be added over time.
• Use confidence scores to automate high-confidence documents and route low-confidence ones to a human reviewer.
• Make the process traceable end to end, since poor visibility into bottlenecks was part of the original problem.