JMIR Publications

ABSTRACT

Background:

Ambient Voice Technology (AVT) is a generative artificial intelligence tool that transcribes clinical encounters into draft notes. It is being rapidly implemented across the NHS to reduce administrative burden and improve the quality of patient care and clinical documentation. Despite increasing adoption, there remains limited qualitative evidence exploring clinicians' experiences of AVT implementation, particularly within UK healthcare settings. Understanding implementation challenges and workforce engagement is essential to support safe, effective and sustainable deployment.

Objective:

To present qualitative findings from a service evaluation of AVT implementation, exploring clinicians' experiences of its use in routine practice. By identifying the perceived benefits, challenges and implementation considerations associated with AVT adoption, the evaluation aims to inform strategies that support the safe and confident implementation of AVT technologies.

Methods:

A multi-phase qualitative service evaluation was undertaken at a large NHS teaching hospital in South West England, informed by the Consolidated Framework for Implementation Research (CFIR) and reported in accordance with SQUIRE 2.0 guidance. Data collection comprised eight hours of grounding observations, one clinician focus group (n=4), six semi-structured clinician interviews and two interviews with implementation project managers. Data were analysed using a deductive thematic approach informed by CFIR, with iterative coding refinement and triangulation between researchers.

Results:

Seven overarching themes and 28 subthemes were identified across the five CFIR domains. Clinicians were generally optimistic about AVT and viewed widespread adoption as inevitable, recognising benefits including improved patient interaction, reduced cognitive burden, enhanced documentation efficiency and reduced administrative work outside clinical hours. However, user satisfaction and future use behaviours were influenced by documentation quality, including note verbosity, perceived loss of clinician voice, hallucinations, and difficulties attributing dialogue during multi-speaker consultations. Productivity gains were perceived as contextual and variable, with documentation review and editing offsetting perceived efficiencies for some users. Successful implementation depended on responsive technical support, integration with existing digital systems, personalised documentation templates, opportunities for iterative learning, and meaningful clinician involvement throughout implementation. Participants reported few concerns regarding patient acceptance or consent, while emphasising the importance of transparent communication regarding AVT use.

Conclusions:

AVT has the potential to improve clinician experience and patient-centred care, but successful implementation extends beyond software functionality alone. Workforce engagement, behavioural adaptation, system integration, implementation support and ongoing refinement are critical determinants of successful adoption. These findings provide practical implementation recommendations to support future NHS deployment of ambient AI technologies and contribute to the emerging evidence base informing national scale-up.