Executive summary
With NHS productivity under intense scrutiny, the conversation around artificial intelligence has shifted decisively from hype to a focus on measurable efficiency gains. A landmark HSJ report on the largest NHS trial of ambient voice technology (AVT) in emergency departments has put a striking figure on this potential, modelling nearly £1bn in annual staff-time savings (£834m) if scaled nationally, alongside significant throughput benefits. This study, featuring suppliers like Tortus AI, provides a powerful signal for where value can be found.
However, the real-world evidence for NHS AI efficiency extends beyond scribing. Demonstrable savings are also emerging from patient engagement platforms that reduce DNAs and from diagnostic pathways with proven cost models, like the NICE-endorsed HeartFlow FFRCT. For NHS leaders, the key is to understand that these benefits are only realisable when underpinned by robust governance. The official NHS England guidance on AI scribes, alongside the mandatory DTAC and DCB frameworks, makes it clear that compliance is as material to ROI as the technology itself.
The efficiency mandate
The NHS productivity puzzle has become the defining challenge of 2025. Despite higher spending and staffing levels, activity has lagged, putting intense pressure on trusts and primary care networks to improve output per pound. For boards and clinical leaders, "efficiency" is no longer an abstract concept; it must translate into tangible, measurable outcomes like minutes returned to direct patient care, improved throughput, and tangible relief for clinical backlogs (House of Commons Library, nuffieldtrust.org.uk).
Tortus in ED: what today’s study signals
The recent HSJ exclusive provides the most significant data point yet on the economic impact of AVT in the NHS. The key findings from the multi-site emergency care study include:
- A national extrapolation modelling ~£834m in staff-time savings if the technology were scaled across all relevant settings.
- An associated throughput uplift of 13% in participating A&E departments (hsj.co.uk).
These headline findings are corroborated by other reports from trust pilots, which consistently highlight that AI assistants can increase the time clinicians spend with patients by approximately 25% (Yahoo Finance). For commissioners, the message is clear: these parameters can be used as credible bounds for building local business cases, but they must be validated with pre- and post-implementation measurement.
Where AI creates NHS efficiency (four archetypes)
1. Documentation & letters (ambient voice / AI scribes)
- Mechanism: The primary efficiency lever is the reduction of clinical administrative minutes spent on writing notes, coding, and drafting letters. These reclaimed minutes can then be redeployed to direct patient care, potentially leading to throughput gains.
- Adoption rulebook: The NHS England ambient-scribe guidance (April 2025) provides the definitive framework for safe and effective adoption.
2. Outpatient flow & attendance (DNA reduction)
- Mechanism: AI-powered patient engagement platforms use targeted outreach and intelligent digital communications to lower "Did Not Attend" (DNA) rates, which directly releases sessional capacity.
- Evidence signal: A case study from University Hospitals Birmingham using the DrDoctor platform reported an 18% reduction in DNAs, alongside measurable administrative time savings (drdoctor.co.uk, NHS England).
3. Diagnostic pathway optimisation
- Mechanism: Using AI to ensure the right diagnostic test is performed first time reduces the need for costly downstream procedures and follow-up appointments.
- Example with cost model: The HeartFlow FFRCT tool for cardiology has a formal NICE cost model (MTG32) which calculates a £391 saving per patient compared to the traditional invasive pathway.
4. Knowledge at the point of care (time-to-answer)
- Mechanism: AI-powered Q&A tools that provide fast, cited answers to clinical questions avert the need for time-consuming, multi-tab searching. While the minutes saved per query are small, they accumulate into significant efficiency gains across a busy clinic or ward.
- UK-specific example: iatroX is designed for this purpose, aggregating UK guidance from sources like NICE and SIGN with a conversational lookup function, plus integrated CPD capture to evidence learning.
Measurement framework
- Time-to-note (mins/encounter): Aim for a ≥20–30% reduction against your local baseline.
- Time-with-patient (% of shift): Target a measurable uplift, using the 25% figure from early trials as a benchmark.
- DNAs per clinic (%): Convert the percentage reduction into the number of clinical sessions reclaimed.
- Pathway cost per patient (£): Use NICE models as a template to calculate before-and-after costs for specific diagnostic pathways.
Governance that protects the efficiency dividend
A compelling ROI case can be completely undermined by non-compliance.
- Ambient voice adoption: Strictly follow the NHSE’s AI scribe guidance, which includes having a clinical safety case, a DPIA, and correct MHRA registration for summarisation features.
- Procurement baseline: All digital tools must have a passed DTAC assessment before purchase.
- Clinical safety: The supplier must provide a DCB0129 safety case, and your organisation must complete a DCB0160 deployment safety case.
Playbooks by setting (90-day sprints)
- Emergency & urgent care (ED): Pilot an ambient voice tool in majors or rapid assessment. Your KPIs should be time-to-note and patients-per-hour, benchmarked against the HSJ study parameters.
- Outpatients: Deploy a DNA prediction and communication tool. Your KPIs are the DNA percentage rate, sessional utilisation, and secondary savings on postage/SMS costs.
- Cardiology pathways: Implement HeartFlow FFRCT where clinically indicated. Your KPIs are the rate of invasive angiography and the pathway cost-per-patient against the NICE model.
- Primary care / cross-cutting: Introduce a rapid, UK-grounded reference search tool like iatroX. Your KPIs are time-to-answer, clicks-per-query, and the rate of CPD compliance.
Risks & Mitigation
The biggest risks to realising efficiency gains are often not technical, but organisational. Integration friction with the EPR, change fatigue among staff, and assurance gaps (missing DTAC/DCB artefacts) can each erase the apparent savings. Mitigate this by assigning clear clinical safety ownership early and budgeting for change management.
FAQs
- Is the “£1bn” AI scribe saving real?
- The HSJ has reported an ~£834m modelled staff-time saving, which it describes as "nearly £1bn." This is a national-scale extrapolation from the largest NHS ED ambient-voice trial to date. Trusts should always validate this against their own local baselines.
- What else shows hard efficiency gains?
- Documented case studies on DNA reduction and official NICE-modelled diagnostic savings provide concrete and defensible levers for building an ROI case.
- What must be in place before scaling an AI scribe?
- You must adhere to the NHSE ambient-scribe guidance, have a completed DTAC assessment, and have the full DCB0129/0160 safety cases signed off.