In his next column for Healthcare Leader, digital expert Dr Tom Micklewright explores ambient voice technology and its promise as the next great leap in digital healthcare
It feels like everyone is talking about ambient voice technology (AVT). The UK government has declared it a ‘ground breaking‘ innovation that will ‘bulldoze bureaucracy’. And clinicians are thrilled to have a device that finally listens to them better than the NHS does.
AVT, for those unfamiliar, uses AI to listen to conversations in the background and then to generate content. In a clinical setting, this includes listening to consultations, summarising key points and generating referral letters or consultation notes. Still typing whilst your patient speaks? Not anymore. Letter dictation eating into your lunchbreak? Hasa la vista.
For a workforce drowning in documentation at the expense of direct patient care, AVT is being hailed as the next great leap in digital healthcare. Over 100 companies worldwide are now offering ‘ambient scribes’, promising radical efficiency gains. But amid the hype, questions are multiplying just as quickly as the pilots.
A revolution or a wellbeing tool?
Several high-profile case studies have accelerated the AVT buzz. Earlier this year, Great Ormand Street Hospital (GOSH) published their evaluation of the AVT Tortus, which was trailed across primary, secondary, and tertiary care settings. GOSH reported an average time saving of 8% across care settings.
East Lancashire Hospitals NHS Trust and The Royal Wolverhampton NHS Trust also shared their findings from trials of CLEARnotes, demonstrating 8-14% productivity savings. Importantly though, clinicians in these trials reported a shift in their work, away from screens and back to patients, along with a tangible reduction in cognitive load.
Behind the headlines though, lies an uncomfortable truth. Productivity gains is NHS-speak for see more patients. In their evaluation, GOSH reported a ‘potential’ to see 10+% more patients.
This narrative has always been politically attractive but unless it’s matched by other initiatives to manage the wellbeing of the NHS workforce, it risks being seen as another way to squeeze more from an exhausted and burnt out system. Long term evaluations will be needed to determine whether AVT genuinely increases consulting activity or whether its true worth lies in reducing cognitive strain in the workforce.
From author to editor
Then there is the thorny question of what happens when clinicians become editors of the healthcare record, instead of its authors?
With AVT, clinicians are being asked to review and approve AI-generated summaries instead of writing them, and this carries consequences. Medical defence organisations have been clear that regardless of who (or what) writes the consultation note, the clinician remains legally responsible for its content.
Equally thorny is the question: what makes a good consultation summary? Only about 20% of consultation dialogue currently makes it into the clinical record. ‘That’s rich data lost’ I hear you say, but it also results in a concise and digestible summary that other clinicians can quickly review. If AVT floods records with accurate but superfluous detail, might they become harder to read in those minutes before the patient walks in, creating delays elsewhere? What’s the point of a faster consultation if bloated health records start to slow everything else down?
A digital governance reckoning for the NHS
The speed at which this technology is advancing has also exposed deep structural gaps in digital governance across the health system. The high-level guidance that has emerged from the centre has highlighted the multi-layered and continuous governance and assurance needed to implement AI safely but left many providers wondering how they can adequately resource this.
Do our current frameworks for risk, safety and ethics hold up in the age of adaptive AI? How do we interrogate training data for bias, evaluate models that learn and evolve, or assess systems that might hallucinate or omit clinical information? And when balancing harms – an inaccurate or erroneous consultation note against clinician burnout and bureaucratic overload – do we need a more sophisticated debate about how we weigh competing risks in healthcare?
Nowhere is this tension clearer than in general practice, where the rising prominence of digital clinical safety have caught practices, PCNs and federations out.
The clinical safety standards DCB0129 and DCB0160 have been in place since 2013, but in truth, compliance has been patchy and lightly enforced.
As AI embeds itself in the consultation room, these gaps have become glaring. ICBs are now scrambling to ensure their practices follow DCB0160 when implementing AVT, CQC are tuning in, and general practices now face a regulatory and governance burden for which many are ill-prepared and unfunded. Who is responsible for this – and who pays for it – remain unresolved questions.
NHS England have worked hard to produce guidance to simplify the process of safe adoption but have added to a mountain of AI policy and guidelines that leaders must now keep abreast of. And no matter how well intentioned, some of the messaging has created more uncertainty.
NHS England’s priority notification earlier this year on the use of AVT was widely misinterpreted as a direction to halt all activity. Confusion then deepened when, unexpectedly, the summarising capability of AVT was deemed to make it a medical device and subject to MHRA registration.
So what should healthcare leaders do now?
- Get governance-ready
Clear digital governance will be the foundation you build on for implementing AVT and other AI solutions safely. Identify your clinical safety officer or team, your DPO and your medical device compliance lead and understand if and how they are adapting their approach to AI. For smaller providers, including general practice, collaborative working may be the only way to deliver this sustainably.
- Start with evaluation, not procurement
Before jumping into a pilot, set clear evaluation metrics, including the metrics that matter to your end users, clinicians. A report that demonstrates efficiency savings and the ability to see 15 more patients a day is unlikely to win over your workforce! Also set metrics for clinical quality, safety, and staff wellbeing. The goal is not just to prove that AVT works, but to understand how it works, for whom, and at what cost.
- Play the long game
This is not quick win. AVT might be your organisation’s first foray into AI and with it, will come a cultural shift in documentation, communication, accountability, and role identity. Build an integrative model, with continuous feedback loops, that can grow at a pace that your workforce is comfortable with.
- Reach out to neighbours
One of the great strengths of integrated care systems is their ability to share learning and spread best practice. Connect with your ICB or your neighbouring healthcare providers, share what you are doing and ask for support. It is rare that we find ourselves at the starting line together but if we connect and collaborate, we can avoid so much of the duplicated effort, work and cost that often comes from introducing digital innovation in silos.
Perhaps the real opportunity of ambient voice technology isn’t in how it listens to patients, but in how it forces the system to listen to itself. To its inefficiencies, its governance gaps, its exhaustion. If we pay attention, this technology might just teach us how to become better listeners before we become better record-keepers and in doing so, build a digital future that serves the people who need it most.
Dr Micklewright works as a GP and a Clinical Lead with Cheshire and Merseyside ICB. He also works as Medical Director ORCHA Health, a private digital health company.
