Singapore Management University (SMU) has modernised its administrative workflows using UiPath‘s business orchestration and AI-powered automation platform.
“This marks a shift from traditional automation to more intelligent, user-driven workflows. Document Screener can dynamically interpret inputs, adapt to different document types, and generate tailored outputs with minimal human intervention,” Cassandra Jenna Bibal, business analyst and automation developer, SMU.
Transforming business operations
Since partnering with UiPath in February 2026, SMU has adopted intelligent document processing to enable AI agents to operate in unstructured environments, tackle complex problems, and support multistep workflows. The approach addresses limitations of conventional AI agents, which often struggle with documents in varying formats or containing embedded tables, graphs, and inferred values.
The partnership initially focuses on automating the extraction of finance reports used for financial reconciliation, which has improved accuracy, reduced processing time, and minimised manual intervention.
SMU gradually transitioned to end-to-end process automation and adopted UiPath Orchestrator and additional UiPath products to power their new AI digital assistant, known as the Document Screener. Users can upload different document types, define screening criteria, and receive customised outputs automatically via email.
The next phase will involve embedding UiPath agentic AI capabilities into copilot-style agents, allowing Document Screener users to instruct AI agents to execute follow-up actions like sending calendar invites or notifications without having to switch systems or manually initiate processes.
Outcomes
By deploying these solutions, SMU has significantly improved operational efficiency, reducing manual tasks like finance report extraction from hours to minutes. The institution’s scalable automation evolved from task‑based to agentic AI, resulting in a 94.87% reduction in document screening time across 1,580 files and 9,300 pages.
