Background Adoption of intraoperative computer-vision (CV-AI) remains nascent, with current surgeon perspectives shaped predominantly by general AI discourse rather than clinical experience. This study evaluated evolving surgeon perspectives toward real-time intraoperative CV-AI following clinical deployment within a tertiary surgical referral unit in Aotearoa New Zealand. Methods Two independent surveys (baseline and follow-up) were administered to surgeons performing laparoscopic cholecystectomy (LC) at Christchurch Hospital over six-months (May-November 2025), spanning deployment of a novel real-time multi-task CV-AI system for LC. Likert-scale (1-5) responses were summarised using median and interquartile ranges. Prespecified thematic composite scores were calculated at baseline, with anchor items compared pre- vs. post-deployment. Results Baseline surveys were completed by 22 surgeons (64.7%); follow-up by 15 (44.1%) post-deployment. Baseline attitudes were cautiously positive, with composite medians for usefulness/readiness of 3.8 [3.5-4.1] and risk/governance 3.2 [2.8-3.7]. Consultants perceived significantly higher usefulness/readiness than non-consultants (trainees/fellows) (4.0 [3.8-4.4] vs 3.6 [3.0-3.8]; p = 0.008). Across phases, willingness to use the CV-AI system remained high and unchanged (pre 4.0 [4.0-4.3] vs post 4.0 [4.0-4.0]; p = 0.929). In contrast, accountability/liability concern was significantly lower post-deployment (pre 4.0 [3.0-4.0] vs post 3.0 [2.0-4.0]; p = 0.020). Conclusions Surgeons are conditionally receptive toward surgical CV-AI, although governance and liability remain pivotal barriers to clinical implementation. Real-world exposure may attenuate perceived accountability risk, supporting the role of staged, safety-first deployments and transparent governance frameworks. Translation efforts should prioritise surgeon autonomy, clear accountability pathways, and collaborative implementation models that align technical performance with clinician trust.