In a private hospital in Malaysia, a doctor finishes a consultation without touching a keyboard. A microphone has listened, a note has been drafted, and the next patient is already sitting down. Repeat that across the 27 hospitals of KPJ Healthcare and you have the kind of deployment British digital leaders have described in strategy documents for two years. The Melbourne company behind it, Heidi, has now raised $340 million at a valuation of roughly $900 million, and says the money will fund expansion across Asia-Pacific along with regulatory and clinical safety capability. The lessons for Britain are close to hand.
The NHS has made ambient voice technology its most visible AI commitment. The 10 Year Health Plan promised wide adoption, and national guidance has told providers that these tools need clinical safety assurance and may qualify as medical devices once they summarise rather than merely transcribe. The ambition is settled. What remains unsettled is who does the assuring, who pays, and who is accountable when a fluent, confident note contains an error.
Heidi’s expansion strategy illustrates the difficulty. To sell in Hong Kong, Malaysia, Japan, South Korea, Indonesia and India, the company has to adapt to different languages, clinical habits and regulators. It has hosted data in Japan because a health system there expects it. Data residency has become a product feature, decided by the buyer’s political sensitivities. The NHS is a buyer with strong views on data and, in principle, considerable leverage as a near-monopoly purchaser. In practice, practices, trusts and integrated care boards have often contracted separately, with uneven technical scrutiny and little bargaining power against a supplier that now has a compliance department larger than most trusts’ digital teams.
That imbalance matters more as the product widens. Heidi’s portfolio now includes a research tool for clinicians, a wearable microphone for capturing consultations in different care settings, and dictation software. A scribe that only transcribes sits beside the record. A scribe that also surfaces evidence during the consultation sits inside the clinical decision. The regulatory question shifts from whether the note is accurate to whether the suggestion influenced care, and the answer determines who carries liability. Trusts buying a cheap scribe today may find themselves adopting a decision-support platform by software update.
Then there is the case for the technology, which is real and should not be buried under caution. Clinicians who finish notes at home at eight in the evening are a retention problem as much as a productivity one, and any tool that returns that time deserves a hearing. Yet the benefit to waiting lists is not automatic. Minutes saved per consultation become extra appointments only if rotas, estates and financial rules allow it. Trusts under deficit pressure will be tempted to bank the saving as headcount reduction, and clinicians who see that happen will stop volunteering for the next pilot. Industrial relations in the NHS remain fragile enough that a tool introduced as relief could be received as surveillance or as a cost-cutting device.
Heidi’s agreement with a Hong Kong university to support nurse training points to something British leaders have underweighted. Adoption follows education. A technology taught during training becomes habit before it becomes policy. The NHS has no comparable route by which new clinicians learn to supervise, correct and challenge AI-generated records, and the skill of noticing a plausible but wrong summary is not one people acquire by accident.
For policymakers, the practical implications are unglamorous. A national assurance route that trusts can rely on without repeating due diligence would cut duplicated effort and shift power towards the buyer. Contracts should protect data portability, so that a scribe sitting between clinician and record cannot become a lock-in mechanism. Evaluation should measure what happens to appointments, error rates and staff attitudes over time, not only minutes saved in a pilot. For UK health-tech firms, the message is that a funded international competitor is arriving with regulatory depth they will struggle to match, and the differentiator will be evidence of safety in NHS conditions.
Suppliers are learning to sell across many health systems at once. The NHS has the scale to buy as one and has so far mostly declined to. The gap between those two positions will decide who benefits from the technology.