Dr Kola Tytler (MBBS MBA MRCGP) | 15 July 2026 | 8 min read
There are two routes an AI scribe can take into the NHS. One is bottom-up: individual clinicians find the tool, try it, like it, and word of mouth does the rest. The other is top-down: an organisation procures it centrally and rolls it out by policy. Heidi's UK story so far has been overwhelmingly the first kind. The reported Midlands framework is evidence of the second phase beginning.
Phase one: clinician-led adoption
Heidi's early UK presence rested on a low-friction product that individual GPs and hospital clinicians could try without waiting for organisational sign-off. Clinicians configured their own templates, tested it in their own consultations, and, where it worked, told colleagues. That word-of-mouth distribution is a large part of why the company has been able to describe usage among a substantial proportion of UK GPs well before any large NHS-wide contract existed. It is a genuinely different growth mechanism from an enterprise software sale, and it produced real clinical familiarity with the product ahead of formal procurement.
Phase two: structured primary care deployment
The clearest documented example of the next phase is Heidi's partnership with Modality Partnership, one of the NHS's largest GP super-partnerships. An initial 25-day evaluation involving 47 GPs and more than 2,800 consultations reported reductions in both in-consultation and after-hours documentation time, alongside improvements in clinician-reported measures such as cognitive load and rapport with patients. That evaluation supported a subsequent expansion to more than 200 clinicians across Modality's 53 GP surgeries. It is worth noting that the wider workforce in general practice includes nurses, pharmacists, physician associates and other practitioners, and that the credibility of any such rollout depends partly on whether adoption extends meaningfully beyond doctors.
Phase three: regional enterprise deployment
The Midlands framework, if it proceeds as reported, represents a further step: a central commercial agreement spanning multiple trusts, with standardised implementation, dedicated account management, and governance and reporting expectations set at a regional rather than individual-practice level. This is a different kind of relationship. It requires Heidi to operate less like a product that clinicians adopt and more like infrastructure that an NHS region depends on, with the service levels, support commitments and safety monitoring that implies.
Why the shift matters commercially
Enterprise and framework contracts tend to be more predictable than a base of individually subscribing clinicians, whose engagement can fluctuate. Deep organisational integration, once templates, workflows and EPR connections are embedded, raises the practical cost of switching supplier later. A successful regional deployment also becomes reference material other ICBs and regions can point to when considering their own procurement, which matters in an NHS environment where organisations frequently look sideways at what peers have already done. None of this guarantees renewal or expansion, but it changes the shape of the commercial relationship from transactional to structural.
Why the shift is genuinely difficult
None of this is straightforward to execute. NHS procurement cycles at regional scale are slower and more heavily governed than an individual clinician deciding to try a product. Implementation varies significantly by trust, by electronic patient record system, and by clinical setting; a workflow that works well in general practice does not automatically transfer to an emergency department or a mental health assessment. Secondary care documentation is far more heterogeneous than primary care consultation notes. Enterprise customers expect defined service levels, dedicated support and regular reporting, none of which a purely clinician-adopted product needs to provide. And safety and performance now need to be monitored at organisational scale, across specialties, rather than tracked informally through individual clinician feedback.
The competitive backdrop
Heidi is not making this transition in isolation. Tandem has built its UK and European distribution partly through its relationship with Accurx and through integrations with more than 100 electronic record systems across Europe, alongside Class IIa certification for both its AI Scribe and Coding Assistant under EU MDR. TORTUS has focused heavily on NHS institutional deployment, formal multi-site evaluation, and became the first ambient voice technology certified UKCA Class IIa in the UK's own regulatory regime, in June 2026. Heidi's regional strategy can reasonably be read as an attempt to combine what it already has, strong clinician-level recognition and primary care traction, with what enterprise NHS buyers increasingly expect: formal procurement routes and higher regulatory assurance.
What actually changed
The most important shift is not simply that more clinicians might end up using Heidi. It is that Heidi is attempting to move from being a popular application that clinicians bring into the NHS themselves, to becoming a formally procured, governed component of NHS infrastructure. Those are different products to run, even if the underlying technology looks similar from the clinician's chair.