Workday, Inc. announced the creation of Workday AI Research, a specialized technical team aimed at developing reliable and efficient artificial intelligence solutions for enterprise environments. The new unit focuses on HR, finance, and IT platforms, with findings intended to influence internal development and contribute methods to the broader academic community.
The company highlighted recent research addressing complex challenges in enterprise AI, including agent memory, explainability, multi-agent orchestration, and reward overoptimization. These studies have been accepted by major conferences such as the International Conference on Machine Learning and the ACM Web Conference.
Gerrit Kazmaier, president of product and technology at Workday, stated that evolving AI agents create privacy and auditability issues that standard models cannot resolve. He emphasized that the new research group is dedicated to solving these specific problems through rigorous science to build systems organizations can trust.
To support academic collaboration, Workday introduced the Workday AI Research PhD Fellowship. This program provides $50,000 in annual funding to doctoral students working at the intersection of AI and enterprise software. Fellows receive mentorship from Workday researchers and early access to career opportunities.
Recent findings from the research team include a selective memory approach that improved precision by 12% while running 31% faster than leading comparisons. Another study on multi-agent systems showed that splitting tasks among specialized agents increased accuracy by 5.8% and ensured all answers met defined constraints.
