We describe a Bayesian Reasoning Framework (BRF) that supports business rule operations for on-line information systems. BRF comprises a three-layer environment with business information systems at the top, a middle-ware Bayesian reasoning server, and a Bayesian reasoning engine at the bottom. The top and middle-ware layers communicate via SOAP/XML protocol, while the middle-ware and bottom layers communicate via a Tag-value protocol that fetches business rules from a central repository. BRF is built as a Bayesian Reasoning Agent and tested in a helpdesk system for assigning advisors to users for trouble-shooting in the operation of business information systems. BRF is modeled following a use-case methodology as well as an inference modeling that uses an assignation template from Common-KADS. The concept, design and implementation of BRF for real-world, on-line business information systems are themain contribution of this research project.