We studied the impact of a prospective, probabilistic scheduling method to replace surgeon-led booking, balancing contradictory goals of maximizing theatre utilised time while minimizing risk of over-utilised time. Implemented in an NHS Scotland hospital, the study analyzed 2420 surgical lists across multiple specialties. The Infix Schedule tool used median and median absolute deviation (MAD) of historical case times to calculate probability of finishing within allocated times, targeting a range 80-90% for utilised time. Data were compared post hoc across three groups: prior surgeon-led scheduling, surgeon over-rides of the tool, and unadjusted Infix-generated schedules. In surgeon-led scheduling, realised utilised time was 78.8%: shadow Infix-generated schedules predicted median utilised time 77.8%, significantly closer to this reality (p = 0.18). Surgeon predictions overestimated utilization (median 87.7%; p < 0.001). Surgeon over-rides were associated with lower utilised time (76.5%) than actual (77.3%) or Infix predicted (78.0%; p < 0.015). Patient cancellation rates were lower for the Infix scheduled cohort (8.51% vs 10.96% for surgeon;(p = 0.0297). However, the tool's performance was dependent on accurate surgical coding. While the "pessimistic" mathematical buffer used to ensure reliability was associated with utilization < 80% target, iterative learning and reducing human overrides should refine efficiency and waiting list management.