Ambient voice technology in health: key considerations for nursing practice

Within nursing practice, AVT has been positioned as a potential solution to support clinical documentation, streamline workflows and reduce administrative burden – issues that remain persistent challenges in the profession.

For example, studies have shown that nurses spend up to 25-50% of their working time on documentation, rather than direct patient care (Stevenson et al, 2020; Manca, 2015).

AVT, such as voice recognition systems and conversational agents, offers the promise of easing this workload by enabling real-time, hands-free recording of clinical notes and communication.

However, while the commercial scaling of AVT products has been rapid, the evidence base for their effectiveness, safety, and acceptability within nursing practice remains limited.

Early pilot studies in healthcare settings suggest potential benefits in efficiency and workflow integration (Kocaballi et al, 2020), but concerns remain regarding accuracy, privacy, data security and the potential for increased cognitive load.

Furthermore, the integration of such technologies into complex care environments requires careful consideration of ethical, professional, and organisational implications.

In this discussion piece, we aim to raise and critically explore some of the issues that may arise from the use of AVT in nursing practice, particularly in relation to clinical documentation, nurse–patient communication, and the safeguarding of patient data.

Historical development of AVT

Speech recognition has been in development since the 1950s, beginning with Bell Labs’ Audrey system, which recognised spoken digits. The technology progressed with products such as Dragon Dictate and IBM ViaVoice in the 1990s, which were among the first to be applied to healthcare contexts (Spicer, 2021; Koprowski, 1997).

These systems required structured voice commands and were prone to errors, but they laid the groundwork for the current generation of AI-enabled AVT, which integrates natural language processing and large language models.

Recent pilots demonstrate how far the technology has advanced. At Stanford, the DAX Copilot system has been trialled to generate near-real-time consultation summaries.

Meanwhile in the UK, Great Ormond Street Hospital (GOSH) has piloted TORTUS, an AVT system linked to the NHS electronic patient record (Ahmed and Wray, 2025; Sollof, 2024).

These trials suggest AVT can reduce documentation time, improve timeliness of clinical correspondence, and support more efficient communication with patients and families.

Economic claims and critical appraisal

Some AVT pilots report substantial projected efficiencies. For example, the TORTUS trial at GOSH suggested millions of pounds in potential NHS savings, largely attributed to reduced administrative load and quicker turnaround of clinical letters (Clover, 2025).

However, caution is needed in interpreting these figures. Results from specialist or short-term pilot studies may not generalise across diverse healthcare settings, such as overstretched emergency departments or community nursing services.

Factors such as training requirements, user adaptation and the Hawthorne effect (where performance improves simply because it is being observed) are rarely captured in economic models, yet they significantly influence outcomes. Rigorous, multi-site evaluations are needed before scaling.

Governance and regulatory ambiguity

The governance of AVT in the UK remains complex. NHS England has stated that only products approved by the Medicines and Healthcare products Regulatory Agency (MHRA) should be deployed (NHS England, 2025).

Yet, at present, no AVT systems are publicly listed as MHRA-approved medical devices. Some AVT products may instead be classified under MHRA frameworks as Software as a Medical Device (SaMD) or Platform as a Medical Device (PaMD) (Gov.uk, 2023).

Many hold only a Class I designation due to their limited history on the market. This creates uncertainty for organisations procuring or scaling AVT, particularly when assurances of safety and efficacy are needed.

Compliance with NHS digital clinical risk management standards provides another layer of oversight. DCB0129 places responsibility on manufacturers to assess risks within product design, while DCB0160 applies to healthcare organisations, requiring assurance that local implementation is safe (NHS Digital, 2018; 2013).

Both depend on the involvement of a clinical safety officer (CSO) – a registered healthcare professional responsible for maintaining the hazard log, leading risk assessment, and ensuring mitigations are in place. For AVT, this includes risks such as transcription inaccuracies, AI ‘hallucinations’, and bias or inequity in recognition of accents and dialects.

However, there is inconsistency in practice. Some vendors argue their products fall outside the scope of DCB0129, while clinical safety professionals typically expect DCB0160 compliance regardless. NHS England is currently reviewing these standards, but until greater clarity is provided, confusion remains across the system.

Professional accountability and clinical safety

It is important to emphasise that AVT does not transfer responsibility for documentation away from the clinician. Nurses remain accountable under the Nursing and Midwifery Council (NMC) Code to ensure records are clear, accurate and contemporaneous (NMC, 2018).

The General Medical Council (GMC) reinforces similar principles in its Good Medical Practice (GMC, 2024). Regardless of whether AVT generates a transcript or summary, the clinician must verify its accuracy.

There are also broader clinical safety considerations. AVT systems are prone to errors where speech is unclear, technical limitations exist, or the system struggles with dialects and accents. There is also risk of ‘hallucinations’, where AI generates information not present in the consultation.

If AVT outputs are used to inform clinical decision-making, this raises further questions about whether the technology should be reclassified as a medical device, with stricter governance applied. Such issues should be explicitly documented within clinical hazard logs.

Balancing opportunity with caution

AVT offers genuine opportunities to release nursing time, improve communication and streamline documentation. But benefits must be balanced with an honest appraisal of risks, regulatory gaps and professional responsibilities.

Nurses should be actively involved in codesigning, piloting and evaluating AVT systems to ensure that outputs are usable, equitable and aligned with the realities of practice. Safety should never be compromised for efficiency.

Conclusion

For nurses, who consistently report documentation as one of the greatest drains on their time, AVT systems present both promise and challenge. As outlined in the introduction, nurses currently spend between a quarter and half of their working hours on record-keeping (Stevenson et al, 2020; Manca, 2015).

AVT could help reduce this burden, streamline communication, and improve the timeliness of letters and records reaching patients and other professionals. However, these potential benefits will only be realised if systems are implemented safely, appropriately, and with careful attention to context.

This discussion has highlighted that AVT is not yet an MHRA-approved medical device, and its regulatory status remains complex.

Safe deployment within the NHS requires compliance with DCB0129 and DCB0160, with CSOs playing a critical role in identifying and managing risks. Equally, nurses and other clinicians remain accountable for the accuracy of records, regardless of whether they are handwritten, typed, or generated by digital tools such as AVT.

For nursing practice, the key recommendations are, therefore, clear. First, nurses must be actively engaged in the design, testing, and rollout of AVT to ensure systems reflect the realities of frontline work.

Second, any local implementation should be underpinned by robust clinical safety processes, with risks logged, mitigated and reviewed. Finally, professional bodies may need to revisit codes of conduct to explicitly address the accountability and governance of AI-enabled documentation.

AVT has the potential to transform nursing practice by reducing documentation burden and releasing time to care. But as with all innovations, success depends not only on the technology itself, but also on how it is embedded into practice, governed, and trusted by the profession.

Chris Richmond is principal clinical consultant, Answer Digital, and former head of delivery, NHS England, and Natasha Phillips is founder, Future Nurse, and former national chief nursing information officer, NHS England