Heidi's West Midlands Sweep: The NHS AI Scribe Boom Meets a Consent Reckoning

The numbers from the West Midlands are hard to argue with. An emergency consultant at Walsall Manor Hospital is saving six minutes per consultation. A Dudley clinic has cut its patient-letter backlog from six months to just 14 days. Those figures, reported by doctors to the BBC this month, come from the NHS's largest single procurement of ambient AI transcription technology to date.

The Heidi Rollout

The software at the centre of it is Heidi, an AI voice-transcription tool developed by Heidi Health. Fifteen trusts across the West Midlands, including Walsall Healthcare NHS Trust, The Dudley Group NHS Foundation Trust, The Royal Wolverhampton NHS Trust and Sandwell and West Birmingham Hospitals, jointly procured the product for use in urgent care, outpatient and GP surgery settings. Following the pilot at Dudley, deployment will reach 70,000 clinicians across 15 acute and community trusts, all Midlands integrated care boards and 1,239 GP practices, making it the largest procurement of its kind in the NHS.

The tool operates in 110 languages. Ravinder Kaur Sahota, Group Chief Information Officer at The Dudley Group and Sandwell and West Birmingham Trust, cited that multilingual capability as a way to simplify letters that many patients currently find too dense with medical terminology. Deployment has already begun in phases, with Dudley, Walsall Healthcare and Royal Wolverhampton going live in May 2026, followed by Sandwell and West Birmingham.

The West Midlands model, 15 trusts buying once through a single framework rather than 15 separate procurement exercises, is the structural story here. It offers a template that other regions are now watching closely as NHS England pushes AI scribes "at pace."

The National Picture

This regional rollout sits inside a much larger national push. In early July, NHS England set out how £10 billion of funding over the next three years will be used for a major overhaul of the health service's technology, digital and data systems. Among the headline commitments: a new AI triage tool inside the NHS App, now being rolled out following a successful Sussex trial where it produced a 29% reduction in the number of people queuing on the phone for appointments. That tool is due to reach more than 200,000 patients within the next 12 months, with full availability to all NHS App users by April 2028.

In June, NHS England announced 505,000 clinicians and support staff would receive access to Microsoft 365 Copilot, with a trial across 30,000 workers at 90 NHS organisations finding the AI-powered administrative support could save staff an average of 43 minutes a day. The rollout to the full cohort is expected to be completed by October 2026.

A major NHS England-sponsored study quantified the ambient scribing benefit at a 23.5% increase in direct patient interaction time during appointments, alongside an 8.2% reduction in overall appointment length.

The Consent Fault Line

So far so efficient. A YouGov poll of 4,039 adults found that 81% of the public want explicit consent before AI scribes are used, with strong opposition running at nearly double the level of strong support for the NHS rollout. Healthwatch England has recommended notice at the point of booking, respect for patient requests to switch scribes off during sensitive topics, and consistent national standards from regulators.

That last point matters. The findings are feeding directly into the national Commission on the Regulation of AI in Healthcare, which reports to government in late summer 2026 and may prompt the MHRA to issue new rules on AI scribes as medical devices. For now, every trust deploying "at pace" is essentially setting consent practice that the regulator may later be required to unwind.

A peer-reviewed paper in the International Journal of Medical Informatics has identified five socio-technical risks in clinical speech-to-text systems: inconsistent consent practices, weaker performance on accented and disordered speech, clinical background noise, missing human review and unclear accountability for errors. The study's author argues that full human review of AI-generated notes, not just spot-checking the opening lines, removes a fair share of the concern.

The Regulatory Horizon

The MHRA's regulatory posture is shifting in parallel. On 8 May 2026, it published the draft Medical Devices (Amendment) Regulations 2026, introducing Predetermined Change Control Plans (PCCPs) for software-as-a-medical-device. Under a PCCP, a manufacturer can pre-define certain future changes to an AI-enabled product without triggering a full re-certification process each time the model updates: a significant practical concession to the realities of iterative AI development. The draft is expected to be adopted in December 2026 and come into force in June 2027, subject to parliamentary approval.

The MHRA has also expanded its AI Airlock sandbox, which has now completed its second phase and secured renewed multi-year funding, with Phase 3 in development. Across its first two phases, the programme explored large language models in clinical decision support, synthetic data for radiology validation and real-time post-market surveillance monitoring.

What the Brand Data Reveals

The commercial enthusiasm for healthtech AI is visible in company formation and trademark filing data, but the picture is more complicated than headlines suggest. According to AI Business Dispatch analysis of Companies House and IPO data (as of July 2026), just 68 new companies were incorporated under SIC code 86.10 (hospital activities) in Q3 2026, a 78.2% fall versus the prior period. UK trademark filings in Class 5 (pharmaceuticals and medical preparations) dropped to 719 in the same quarter, down 71.9%. And 98.8% of all active SIC 86.10 companies hold no Class 5 trademark whatsoever, a striking figure for a sector in which brand protection increasingly matters as AI clinical tools multiply.

That last statistic is less paradox than warning. Companies racing to deploy ambient scribing tools, triage AI and clinical decision support at NHS scale are doing so in a market where regulatory and IP infrastructure is still being built around them. The Heidi rollout in the West Midlands shows what is possible when procurement is coordinated and evidence is generated upfront. Whether consent governance keeps pace with deployment will determine whether that story stays a success.