Over the past few years, the telecommunications paradigm has been shifting rapidly from hardware to middleware. In particular, the traditional issues of service characteristics and network control are being replaced by the modern, customer-driven issues of network and service management (e.g., electronic commerce, one-stop shops). An area of service management which has extremely high visibility and negative impact when managed badly is that of problem handling. Problem handling is a very knowledge intensive activity, particularly nowadays with the increase in number and complexity of services becoming available. Trials at several BT support centres have already demonstrated the potential of case-based reasoning technology in improving current practice for problem detection and diagnosis. A major cost involved in implementing a case-based system is in the manual building of the initial case base and then in the subsequent maintenance of that case base over time. This paper shows how inductive machine learning can be combined with case-based reasoning to produce an intelligent system capable of both extracting knowledge from raw data automatically and reasoning from that knowledge. In addition to discovering knowledge in existing data repositories, the integrated system may be used to acquire and revise knowledge continually. Experiments with the suggested integrated approach demonstrate promise and justify the next step.