Ambient voice technologies in the NHS: all talk?

Is it time to shout about the benefits of ambient voice technologies, or do we need to consider how to unlock the long-term benefits first?

5 years on from the COVID-19 pandemic, burnout continues to affect many NHS staff (30% according to the 2024 NHS staff survey), while pressures on the system continue. Relieving the administrative burden on clinicians could help to improve working hours and boost NHS productivity and is an area where AI could offer solutions. We emphasised non-clinical AI applications as a priority in our call for an NHS AI strategy, and it’s why we’re exploring the potential of non-clinical applications of AI.

Ambient voice technologies (AVTs) could be particularly transformative given their potential to support more meaningful face-to-face consultations with patients, while capturing and converting information more quickly and comprehensively than most people can. This potential has left many NHS organisations excited about AVTs. 

Also known as ‘digital scribes’, these tools record conversations between people (in health care, mostly between clinicians and patients) and can then convert that recording into patient note summaries. Over time, tools are also expected to integrate directly into electronic patient records, draft referral or patient letters, or convert speech into other structured medical documentation.  

‘Leaping and learning’ vs ‘watching and waiting’

The functionality of these tools is still evolving and depends on the specific product and how it’s implemented. The evidence base for AVTs’ impact is also still emergent, so where and how these tools will have greatest benefit will require careful thinking across the NHS.

It's against this backdrop that NHS England recently released guidance on the use of ambient scribes. To date there’s been divergent thinking on if, when and how to deploy these technologies. The overview of AVTs, published alongside the guidance, majors on the benefits. 

There are also those who counsel caution, noting we need better evidence on the impact of these tools and confidence that data and cyber security, patient consent and environmental considerations are well understood and navigated. 

This doesn’t imply earlier adopters haven’t been conscious of these considerations. But perhaps they have been ‘leaping and learning’ rather than ‘watching and waiting’ to understand the benefits and disadvantages. There’s also been much confusion over whether these tools should be classified as ‘medical devices’, with implications for approvals and their use.  

Where next?

The NHS England guidance is welcome to help make sure AI is used responsibly and impactfully in the health service. It will provide much-needed insight for many NHS organisations, but it’s still only one part of the suite of skills, support and tools needed to use AVTs responsibly and effectively.

Support for NHS organisations needs to be given relative to their current knowledge, capabilities and headroom for change alongside the guidance. There’s a need for:

• further evidence of impact. The London-wide AVTs trial has promising early data, with insights highlighted in the government press release. Understanding the output from the full evaluation will help build our evidence base of these tools. Recent international qualitative evidence cites clinicians' benefits and where we need more information, including in understanding how these tools are perceived by patients. Research we have commissioned from The Healthcare Improvement Studies (THIS) Institute will contribute to our understanding of what staff do with their time freed up, of the patient and clinician experience, and of implementation barriers.
• information parity. This will enable NHS organisations to have the information they need from the developers of AVTs for comparison and assessment of how well solutions might perform in their own environments and for their patients, compared with the data and environments these tools were trained and validated on.
• capability building. NHS organisations need to be able to assess and continually monitor tools, given we know the skills and knowledge needed to reasonably assess tools doesn’t exist universally across NHS settings. Time-poor NHS organisations also need easy tools and templates to support each step, from scoping to post-implementation monitoring.
• resources and capabilities to support change alongside the technology. The guidance identifies the need for organisations to consider the impacts of AVTs within the workflow and care pathways. This will require change management capabilities, including planning for any indirect impacts of AVTs on non-clinical roles. Take-up to date has also largely relied on free licences and the enthusiasm of staff.
• ensuring informed consent from patients about how their data is used. Alternatives to AVTs should be also offered to patients who don’t want consultations recorded. Our attitudes tracking identifies that the public do not, on balance, support the use of AI if outputs are not checked by staff. Patients say recording appointment notes and creating letters are areas where it's important to have a ‘human in the loop’, raising questions about what this means for AVTs. 

Vitally, we also need space for those using AVTs to learn from fellow practitioners, so we get the most out of the technology. We’re excited that THIS Institute, funded by the Health Foundation, will be developing an AVTs Innovation Community on the Thiscovery platform to enable cooperation, collaboration and coordination for AVTs in health care. One feature of the community will be providing at-scale service evaluation capability using an ‘evaluation in a box’ model. You can engage with the community via email.

There’s a lot to be excited about when it comes to what these tools might offer and it’s great to see early benefits. With further support for adoption, more insight and evidence, and learning among practitioners, we can hopefully sing AVTs’ praises soon.