Evaluation of ambient voice technology in the NHS

What is AVT?

Ambient voice technology (AVT), sometimes called an “AI scribe”, uses conversational artificial intelligence to capture clinical consultations and generate structured documentation for clinical review. It aims to reduce the documentation burden on clinicians while maintaining or improving the quality of patient records.

Why are we evaluating AVT?

Clinical documentation takes up a substantial share of clinicians’ time: estimates suggest up to a quarter of a clinician’s working day can go on record-keeping and admin, with clinicians spending an average of two hours outside clinical hours on documentation. AVT is being adopted across the NHS as a potential way to reduce that burden, but evidence of its real-world impact remains fragmented. Much of the existing evidence comes from small, short-term pilots and is sometimes generated or commissioned by the technology vendors. National regulatory and procurement frameworks for AVT are evolving quickly, increasing the need for robust, independent evidence on impact, implementation and value for money.

NIHR RSET is carrying out an independent evaluation to help the NHS make evidence-based decisions about the adoption and use of AVT.

What are the aims of the evaluation?

Phase 1 (September 2025 to January 2026) aimed to establish the evidence base and evaluation foundations for the study through a scoping review of existing evidence, an AVT product taxonomy, a market map of NHS adoption, and a logic model and outcome framework to guide Phase 2.

Phase 2 (data collection August 2026 to January 2027; results expected February 2027) tests that logic model in a multi-site, real-world evaluation across four NHS trusts, focused on:

• Productivity: does AVT generate measurable time savings in documentation, and how are those time savings used in practice (e.g. more time with patients, absorbed into existing workload)?
• Value for money: what does AVT cost to implement, integrate and run, and what are the economic consequences for NHS organisations, including a return-on-investment analysis?
• Staff experience: how do staff experience using AVT, and what factors help or hinder its implementation and impact?

What did we do in the Phase 1 evaluation?

Scoping review

We carried out a rapid systematic scoping review of evidence on the use and impact of AVT for clinician-patient documentation, including how outcomes have been measured. We searched academic

databases and grey literature in September 2025. More than 500 records were screened, and 21 studies met our inclusion criteria.

Most studies were short-term pilots or quality improvement projects (often lasting less than six months). Evidence was most consistent for reductions in documentation time, including after-hours documentation, and some studies reported improved staff wellbeing and more patient-centred consultations. Evidence on patient experience, safety and costs was limited and used inconsistent metrics, while findings on note accuracy and quality were mixed. Overall, the evidence base is limited and fragmented, supporting the need for longer-term, more consistent evaluation across settings.

This published review is available: "The use of ambient voice technology (AVT) for clinician-patient consultations in healthcare practice: A rapid systematic scoping review", BMJ Digital Health and AI (July 2026); see Outputs below.

AVT taxonomy and market map

We developed a taxonomy of AVT products to provide a shared language for comparing tools with different functionality and to support evaluators, services and policymakers. Developed through document and product reviews, the scoping review, interviews, and a workshop with NHS England, it sets out 10 essential elements that define a product as AVT, and 40 additional elements covering transcription, summarisation, equity, accessibility and other functionality.

Using the taxonomy, we reviewed the AVT market in England and the wider UK, supplemented by an NHS England overview of AVT use across trusts and other services. We identified 32 products, 14 with some claimed use in NHS or social care settings; of these, 11 were MHRA-registered devices and 11 appeared on the NHS England AVT supplier registry. The market map is a snapshot as of 30 January 2026, uptake is changing rapidly and claimed product use was often difficult to verify independently.

Logic model and outcome framework

We developed a logic model and outcomes framework to link AVT inputs and activities to outputs and to short, medium and longer-term outcomes for staff, patients and NHS organisations. Designed to apply across clinical settings, it underpins Phase 2 and provides a consistent set of outcome measures in response to the wide variation identified in the scoping review.

How are we doing the Phase 2 evaluation?

Phase 2 is a rapid, multi-site, mixed-methods evaluation running from August 2026 to January 2027, with results expected in February 2027. It covers four NHS trusts in England:

• Two acute trusts, using AVT in hospital outpatient services and Accident & Emergency departments; and
• Two mental health trusts, using AVT in outpatient services.

Sites were purposively selected to vary by care setting, AVT product or supplier, stage of deployment, and organisational and digital maturity, compliance with MHRA and NHS England requirements, while ensuring access to the routine data required for evaluation.

The evaluation has three parts:

· Quantitative analysis: quasi-experimental analysis of routine NHS data including electronic health records, and where relevant Hospital Episode Statistics and the Emergency Care Data Set. We will estimate the impact of AVT on documentation time, clinician activity and service outcomes, comparing outcomes before and after implementation and, where possible, AVT with non-AVT consultations.

· Health economic evaluation: cost-consequence and cost–benefit analysis , plus budget impact modelling of the short- to medium-term financial implications for the NHS. This will produce a freely available return-on-investment decision-support tool that NHS organisations can apply using local data.

· Qualitative interviews: up to 36 semi-structured interviews with staff across settings, exploring implementation, staff experience, and the factors that help or hinder use and impact.

Findings from the three strands will be integrated to show how AVT is implemented, whether it generates time savings, how any released time is used, and what thist means for NHS organisations considering wider adoption. The RSET Patient and Public Involvement and Engagement (PPIE) Panel and a public contributor are involved throughout, contributing to study design, interpretation and communications of findings.

Full detail is in our Phase 2 study protocol, published as a preprint on JMIR (June 2026); see Outputs below.

Logic model

Overarching theory of change: if AVT systems accurately capture and summarise clinical encounters and integrate into electronic health records, then clinicians spend less time documenting and more time with patients. This improves data quality, clinician wellbeing and productivity. Over time, these translate into improved patient outcomes, better workforce retention, and system efficiency.

The model sets out how AVT inputs, including technology capability, EHR integration, governance and training, translate through implementation activities into changes in documentation workflows and service outputs. These are expected to reduce administrative burden and improve document turnaround, leading over time to better staff wellbeing and retention, improved patient experience and care, and sustained gains in productivity and care coordination.

Outputs

Published papers 

• “The use of ambient voice technology (AVT) for clinician–patient consultations in healthcare practice: A rapid systematic scoping review”, BMJ Digital Health and AI (July 2026). 
• “Evaluation of Ambient Voice Technology in the NHS in England: Phase 2 Study Protocol”, JMIR Preprints (June 2026).

Reports

Blog posts 

Forthcoming

• Phase 2 peer-reviewed papers and final NIHR report (expected from February 2027).
• A publicly available, freely usable AVT return-on-investment decision-support tool for NHS organisations.

Study team

Principal investigators

•  Professor Jenny Shand (Department of Clinical, Educational and Health Psychology, University College London)
• Professor Stephen Morris (Department of Public Health and Primary Care, University of Cambridge)

Team

• Theo Georghiou (Nuffield Trust)
• Lucina Rolewicz (Nuffield Trust)
• Dr Kevin Herbert (University of Cambridge)
• Dr Rachel Lawrence (University College London)
• Pei Li Ng (University College London)
• Holly Elphinstone (University College London)
• Dr Holly Walton (University College London)
• Raj Mehta (Public contributor, RSET PPIE co-lead)

Funding

This is independent research funded by the National Institute for Health and Care Research (NIHR) Health and Social Care Delivery Research (HSDR) programme, 2023-2028 (project reference NIHR156380). The views expressed are those of the authors and not necessarily those of the NIHR or the Department of Health and Social Care.

Who we are working with

NHS England

University of Oxford

Health Innovation Networks

For further information

To find out more, contact avtstudy@ucl.ac.uk