Tandem and Accurx After the $100 Million Raise: How Clinical AI Reaches NHS Teams

Accurx Scribe is built on Tandem Health's technology. The companies are therefore not two unrelated scribe suppliers in this relationship: Tandem contributes the underlying capability, while Accurx provides an established route into NHS communication and record workflows. That distribution relationship helps explain the practical UK significance of Tandem's US$100 million raise on 14 September 2026.

The relationship is documented in Accurx's 27 May 2025 announcement. The new financing is documented separately in Tandem's September 2026 release. The distinction matters: the funding round did not create the partnership, and the partnership announcement did not establish how many people would ultimately use the scribe regularly.

The announcement described access, not a census of active users

In May 2025, Accurx described an access route for more than 200,000 NHS staff and an initial rollout through practices already using its software. It also described write-back to EMIS and SystmOne. Those were announced distribution and integration arrangements, not proof of 200,000 active scribe users. The original announcement is the relevant source and date.

Accurx subsequently provided a different kind of figure. In its 13 October 2025 article on using Scribe, the company reported more than 500,000 consultations in GP practices since launch. That is a company-reported historical activity count. It is neither a unique-clinician count nor an independently measured clinical-outcome result.

The two figures are not contradictory. One concerns the audience a product could reach; the other concerns activity reported after launch. Neither should be silently updated to September 2026 or divided into the other to estimate a reliable active-user rate. The periods, populations and measurement units do not match.

That distinction is especially important when comparing suppliers. A large existing communications platform may provide a substantial potential audience. The scribe's adoption within that audience is a separate question.

Why a distribution partnership may matter commercially

The strategic interpretation is straightforward: reaching a clinician through software already present in the workflow may remove some obstacles that a separate product would face. It may reduce the need to discover another application, establish a new habit or move output manually between unrelated tools.

These are plausible mechanisms, not a measured causal finding about the Accurx partnership. Existing access does not eliminate configuration, training, governance, support or the need for the output to be useful. It may make the first step easier without settling the rest of the journey.

For the earlier product context, iatroX's Tandem care-partner analysis discusses its documentation and aftercare direction. The question here is how an NHS team reaches and implements the relevant workflow.

The commercial value also depends on who makes the decision. A clinician can be interested in a tool while a practice needs agreement on its use. A central organisation can approve a product while individual staff still decide that it does not fit their consultations. Distribution reaches the opportunity; implementation determines what happens to it.

For the funding story, this helps distinguish model capability from delivery capability. A company seeking to expand clinical AI needs more than an impressive demonstration. It needs a route through which the intended users can obtain, configure and continue using the product.

Integration is a set of behaviours, not a badge

The EMIS and SystmOne write-back described in the May 2025 Accurx announcement is a useful concrete starting point. It does not establish that every Tandem function is present in every Accurx deployment or that every action occurs automatically.

An organisation should ask what the particular implementation actually does. Does it identify the correct encounter? What is transferred? Which elements require review? Where do coding suggestions appear? What remains a separate task? These are demonstration questions rather than assumptions about undocumented behaviour.

A useful walkthrough includes a correction. Ask the supplier to change a clinically meaningful detail in an original fictional encounter and show where the amendment is made, how the final version is recognised and what appears in the record. Then ask what happens if transfer is interrupted. A normal-path demonstration cannot answer an exception-handling question.

The same approach applies to documents beyond the note. Accurx's original announcement described referrals, Advice and Guidance requests and patient communications. Whether a particular output is drafted, sent or merely prepared for review should be established in the version and setting being considered, not inferred from a broad product description.

Follow the implementation through to a completed task

The following proposed monitoring framework separates stages that can otherwise disappear into a single adoption statistic. It is an original framework for local evaluation, not a report of observed Accurx performance.

Stage	What to record	What the measure does not prove
Available	Staff and settings for which access is enabled	That anyone has attempted a session
Started	People who initiate a first eligible use	That the output is completed or useful
Completed	Encounters reaching a reviewed, usable output	That the workflow saved time overall
Repeated	Continued use over a defined observation period	That every user or patient benefits
Improved	A relevant outcome compared with an appropriate baseline	That the result generalises beyond the evaluated setting

Keep the period and denominator visible at every stage. Counting individual users, encounters and organisations together obscures the point at which a rollout is struggling. A practice with a high completion rate among a few enthusiasts may need a different intervention from one where many people start but abandon the process.

Reasons for non-use are informative. A clinician may have an unsuitable consultation type, a technical problem, a workflow preference or a concern about the output. The monitoring process should be able to distinguish these explanations without assuming that every non-user needs more training.

Assurance and support are part of delivery

Accurx's 13 October 2025 implementation article discussed support for governance, a trust centre, clinical safety officer training and sessions with its clinical team. These are dated examples of the implementation work the company described, not a claim that every announced session remains available today.

The practical lesson is that the delivery layer includes helping an organisation understand how a product is used and supported. A practice should request the current documentation for its proposed deployment rather than rely on an old announcement, a neighbouring practice's configuration or the identity of the underlying technology supplier.

Responsibility for support also needs to be clear. When a problem occurs, staff should know where to report it, how urgent issues are handled and who communicates the resolution. These are sensible questions for a service-level discussion; they are not claims about a particular contractual division of responsibility between Accurx and Tandem.

The evidence should match the implementation

Tandem's study published on 10 July 2026 provides evidence about recorded editing behaviour and perceived changes among sustained adopters in a Swedish healthcare setting. It does not automatically establish the size of any benefit for an NHS team using Accurx Scribe.

An NHS service can regard those findings as relevant background while measuring its own complete workflow. The companion article, Tandem's 29% Documentation-Time Reduction: What the Published Study Actually Shows, explains why self-reported note-finalisation time is different from an objective measure of all administration.

The outcome should match the implementation goal. For a note workflow, examine the work required to reach a reviewed, usable record. For communication, examine whether the required message reaches its intended recipient and whether additional correction is needed. For clinician experience, ask staff directly while keeping their responses distinct from note-quality measures.

Neither a large audience nor a favourable satisfaction response is a substitute for that task-level evaluation. Equally, the absence of a randomised trial for every local configuration does not mean a team must ignore its own carefully collected experience. The important point is to describe what was actually measured.

Where separate reference and learning tools fit

iatroX publishes this article and is included here as an example of a different access route, not as a replacement for Accurx Scribe. According to its platform overview and Tutor guidance, checked on 14 September 2026, it offers clinical reference and structured learning that a clinician can explore separately from their organisation's documentation system.

A fictional GP might use an approved scribe for the record and later formulate a non-identifying learning question about an uncertainty encountered during the day. Inspecting a source or working through a related question is a different activity from recording the consultation. This example does not assume a Tandem-to-iatroX connection, transfer of patient data or NHS-wide endorsement.

A team seeking documentation integration should therefore evaluate the Accurx Scribe deployment on its own merits. A clinician seeking additional reference or learning support should assess those tools against that separate purpose. Complementary use may be sensible, but access to one product does not constitute approval of another.

The post-funding question is whether the delivery route produces more completed, useful work with an acceptable review burden. Distribution can create the opportunity. Sustained, appropriately evaluated use is what gives the opportunity clinical meaning.

Frequently asked questions

Is Accurx Scribe powered by Tandem Health?

Yes: Accurx's 27 May 2025 announcement describes Accurx Scribe as built on Tandem's AI scribe technology. The partnership predates Tandem's September 2026 Series B.

Does access through Accurx mean everyone uses the scribe?

No: availability, first use, repeated use and completed workflows are different measures. The original announcement's more-than-200,000-staff figure described an access route, not an independently verified count of active scribe users.

Can clinicians use separate reference and learning tools alongside a scribe?

They can serve different purposes, subject to the relevant organisational policies and appropriate handling of information. Separate use does not imply an integration or permission to transfer identifiable patient material between products.

Explore a separate clinical learning question with Ask-iatroX →