Earlier work by Camp et. al (2018) on the preliminary hindcast of the Bureau's new seasonal model ACCESS-S1 showed it had multi-week skill in forecasting cyclone formation in the Southern Hemisphere. This was attributed to the model correctly simulating large scale changes in the atmosphere with the phase of the MJO. Continuing this work on the full hindcast showed monthly variation in forecast cyclone biases during the cyclone season. Further investigation from 2017-18 season involved the use of monthly bias corrections and lagged ensembles. Ten day forecasts generated by the numerical weather prediction model ACCESS-GE over the same period also showed good skill in predicting cyclone formation. An operational system combining both ACCESS-S1 and GE ran in real-time during the 2018-19 season. Results for this season will be presented along with case studies, customer feedback and user experiences.