Is Class IIa Becoming the New Competitive Baseline for NHS AI Scribes?

Featured image for Is Class IIa Becoming the New Competitive Baseline for NHS AI Scribes?

Dr Kola Tytler (MBBS MBA MRCGP) | 15 July 2026 | 7 min read

Tandem reached EU MDR Class IIa certification for its AI Scribe on 21 May 2026, and its Coding Assistant had already reached the same milestone in February 2026. TORTUS followed with UKCA Class IIa certification on 24 June 2026, the first ambient voice technology to achieve that status under the UK's own regime. Heidi has stated it is progressing towards Class IIa. That is three of the market's more prominent suppliers converging on the same regulatory tier within a matter of months, which raises a fair question: is Class IIa moving from a differentiator to an expected standard for enterprise-scale NHS deployment.

Why Class I has been sufficient for some scribes

Lower-risk intended purposes, mandatory clinician review before anything reaches the record, draft-only outputs rather than autonomous clinical actions, and relatively limited functionality have all supported Class I self-declaration for parts of this market. NHS England's own AVT supplier registry currently requires at least Class I registration to be listed, which has kept that route viable for suppliers with narrower functionality.

Why the market may be moving upward regardless

Modern AVT products increasingly do more than transcribe. Generated summaries now shape the longitudinal clinical record directly. Products are extending into clinical coding, draft referrals, order suggestions and, in some cases, decision support, each of which introduces a different order of clinical risk than a draft note awaiting review. Larger deployments, of the kind Heidi's Midlands framework and Tandem's reported multi-trust agreement represent, also increase the potential scale of any single systemic error. NHS procurement and clinical safety teams are, understandably, asking for more external assurance as deployments grow.

Tandem's regulatory argument

Tandem's public position is that a clinical note becomes the basis for future clinical decisions about that patient, which is precisely why EU MDR Rule 11 pushes software generating clinical information into at least Class IIa. The company frames independent notified body review, rather than self-assessment, as infrastructure rather than a compliance formality, and has extended the same logic to its Coding Assistant on the basis that NHS guidance treats clinical coding as Class IIa functionality in its own right.

TORTUS's regulatory argument

TORTUS makes a related but distinct case: that ambient voice technology composes clinical information rather than merely recording it, and that composing information which materially influences clinical workflow cannot reasonably sit on an unregulated or self-declared foundation once functionality expands. Its certification introduces an independently audited quality management system, formal risk management documentation and ongoing post-market surveillance obligations, assessed under ISO 13485-aligned processes.

Where this leaves Heidi

Heidi has already achieved substantial clinician-level adoption across the UK, and the Midlands framework, if it proceeds as reported, increases the institutional weight riding on that adoption. Achieving Class IIa would remove a visible point of regulatory difference between Heidi and two significant competitors, and would plausibly support the kind of broader clinical-assistant platform ambitions the company has signalled publicly.

Could Class IIa become a procurement requirement?

There are reasonable arguments on both sides. In favour: greater external assurance, more sophisticated functionality across the category, regional-scale risk exposure, easier like-for-like comparison during tenders, and increasing pressure from clinical safety teams who now understand the distinction between registration and certification. Against: classification should remain proportional to actual intended purpose rather than imposed uniformly for marketing parity, not every scribe has identical functionality or risk profile, and NHS England's registry currently accepts Class I registration as sufficient for listing, which means there is no formal requirement forcing the issue yet.

Conclusion

Class IIa is likely to become increasingly common among suppliers competing for enterprise and regional NHS contracts, simply because functionality across the category keeps expanding into higher-risk territory. It has not become a legal requirement for every documentation product, and functionality and intended purpose, not competitive pressure alone, should continue to determine the appropriate class.

Read the AI Scribe Regulatory Status Tracker →

Share this insight