Background and objectives People with personality disorders are prevalent in emergency and inpatient mental health services. We examined whether implementing a stepped care model of psychological therapy reduces demand on hospital units by people with personality disorder, in a cluster randomized controlled trial. Method A total of 642 inpatients (average age 36.8, 50.5% female) with a primary ICD-10 personality disorder were recruited during 18 months baseline, then monitored during an 18 month active trial phase. In the active trial phase two equivalent sites were randomised to either treatment as usual (TAU), or a whole of service intervention that diverted people away from hospital and into stepped care psychological therapy clinics. The study design was cost neutral, with no additional staff or resources deployed between sites. A linear mixed models analysis evaluated outcomes. Results As predicted, demand on hospital services reduced significantly in the intervention compared to TAU site. The intervention site evidenced shorter bed days, from an average of 13.46 days at baseline to 4.28 days per admission, and patients were 1.3 times less likely to re-present to the emergency department compared to TAU. Direct cost savings for implementing the approach was estimated at USD$2,720 per patient per year. Limitations included not directly comparing individual symptom changes. Conclusions Using a whole of service stepped care model of treatment for personality disorder significantly reduced demand on hospital services.