Every AI Success in NHS Wales Has One Thing in Common: It Didn't Go Through DHCW

In the year to September 2024, Wales did something no other UK nation had done: it deployed an artificial-intelligence stroke platform — Brainomix 360, which reads a brain scan in real time to fast-track emergency stroke treatment — across the whole country, fifteen stroke centres, on a national basis. Mechanical-thrombectomy rates rose from 1.8 per cent towards 3 per cent nationally, with early-adopter sites reaching 7 per cent and thrombolysis rates above 20 per cent. (Source: NHS Wales Executive, "Revolutionary new technology set to transform stroke care in Wales," September 2024; Life Sciences Hub Wales case study, 23 September 2024; Brainomix, 21 October 2024.)

It was a genuine clinical-AI win — one of the few in the UK delivered nationally rather than in a single trust.

It was procured by the National Stroke Programme Board working with the NHS Wales Executive, with Life Sciences Hub Wales acting as facilitator and a single central project manager coordinating fifteen health-board IT, governance and clinical teams.

Digital Health and Care Wales — the Special Health Authority the Welsh Government created in 2021, the direct successor to the nationally discredited NHS Wales Informatics Service (NWIS) and built on the same leadership and the same delivery culture — is named nowhere in any of it. Not in the procurement. Not in the funding. Not in the delivery. Not even as the provider of the networks the software runs on.

That is not an exception. As this article sets out, it is the pattern. Every AI success in NHS Wales shares a single characteristic: it did not go through DHCW.

Artificial intelligence is the breakthrough politicians across the UK — and in Wales — now name most often as the answer to the forces crushing the NHS: the backlogs, the burnt-out workforce, the diagnostic bottlenecks. On the official telling, it is the single biggest opportunity in a generation of health policy. And the body the Welsh Government created and tasked to deliver it in Wales did almost nothing. Every AI project that has worked in NHS Wales has worked despite never involving DHCW — or perhaps because it never did. Time after time, programme after programme, at every level of cost and every pitch of stated urgency, DHCW has proved unable to deliver. AI is simply the newest entry in that record — and, on the thing politicians call the most important opportunity of the age, the most damning.

The body that was given the monopoly

DHCW was not a bystander to AI in Wales. It was, on paper, the lead.

When the Welsh Government created DHCW in April 2021, it gave a single body a national monopoly over NHS Wales's digital, data and technology infrastructure. And when Wales turned its attention to AI, DHCW was placed at the centre of the machinery built to govern it.

In June 2023 the Welsh Government and the NHS Wales Executive established the AI Commission for Health and Social Care in Wales — a senior advisory body whose terms of reference were approved on 31 October 2023, chaired by the then NHS Wales Chief Digital Officer, Mike Emery, with its secretariat provided by the Welsh Government and Life Sciences Hub Wales. (Source: gov.wales, "AI Commission for Health and Social Care"; Public Health Wales Knowledge, Research and Information Committee paper, 6 June 2024.)

DHCW is a founding-partner member of that Commission — it says so on its own website. DHCW is represented on it by its Executive Director of Strategy, Ifan Evans, and by an Independent Member, Rowan Gardner. And the Commission's terms of reference do not give DHCW a watching brief: they assign it a workstream — "improving AI capability in health and social care: Digital, data and technology roles." (Source: AI Commission terms of reference, in the Public Health Wales paper, 6 June 2024; dhcw.nhs.wales/data/artificial-intelligence-ai.)

So when DHCW's record on AI is examined, the framing is not "AI was never its job." AI was explicitly its job. It was tasked, named, and given a workstream.

The AI that happened anyway — without them

Hold the monopoly's mandate next to the AI that Wales actually delivered. Every documented win was driven by a body other than DHCW.

Stroke AI — Brainomix 360, described above. The detail that matters here is how it was delivered: a single central project manager — hosted outside DHCW — coordinating fifteen health boards' IT, governance and clinical teams into one national rollout. The cross-boundary, many-stakeholder coordination DHCW routinely cites as the very reason it cannot deliver was achieved here, in months, by other people, around it. (Sources, open in a new tab: NHS Wales Executive; Life Sciences Hub Wales; Brainomix, September–October 2024.)

AI clinical scribes. The pilot that ministers cite most often — AI scribing to cut the assessment backlog in children's neurodevelopmental services — was run by the Centre for Digital Public Services (CDPS) with Cwm Taf Morgannwg University Health Board, using the Magic Notes tool, tested at Swansea University's simulation suite. The same tool is in production use in adult social care at Neath Port Talbot Council. Powys Teaching Health Board's flagged digital tool for health visitors, eFRAIT, was built by the University of South Wales. (Sources, open in a new tab: Centre for Digital Public Services; gov.wales AI Plan launch release, 18 November 2025; gov.uk AI Knowledge Hub.)

None of these — not one — was delivered by DHCW. Nor is there a counter-example hiding elsewhere: DHCW has not produced a single AI tool of its own in operational use anywhere in NHS Wales.

The throughline is hard to miss. Each AI success came from a clinician-led programme board, a delivery unit, or a council finding a route around the national monopoly — and succeeding across precisely the complex, multi-stakeholder landscape DHCW likes to cite as the alibi for its own permanent non-delivery. The distributed models delivered. The monopoly did not. When the recipe for an AI win in NHS Wales is reverse-engineered, it reduces to a single step: do not route it through DHCW.

Inside the monopoly: the AI that never came

If the AI was happening everywhere except DHCW, what was happening inside DHCW?

Across five years of DHCW's public board and committee meetings — dozens of them between 2021 and 2026, in the full meeting recordings rather than the heavily edited published minutes — artificial intelligence is mentioned only a handful of times, and never as something delivered. It appears as a recurring strategy bullet ("automation and AI"), voiced largely by the Executive Director of Strategy. It appears as a thing still to be planned: at the board in March 2024, more than nine months after the AI Commission was stood up, the position was that DHCW still needed to "navigate" who owned AI strategy and would "need a plan around it." And it appears as a future justification for the National Data Resource — AI as the thing the data platform would one day enable.

The single concrete healthcare-AI deliverable on DHCW's record is not a tool. It is a training module: an "introductory training module on artificial intelligence for healthcare," commissioned in November 2024 from the Wales Institute of Digital Information (WIDI) — a body of which DHCW's Chief Executive is a listed director — an interest she has never separately declared, and which appears on DHCW's own register only as a "Nil Declaration". (Source: DHCW public Board meeting, 28 November 2024, Mission 4 half-year performance item; widi.wales director listing; DHCW Declarations of Interest register. See the WIDI related-party record.)

The only named AI deliverable in DHCW's own corporate plan — the "AI orchestration and deployment engine (AIDE)" discovery in its Integrated Medium Term Plan 2025-28 — is an internal back-office capability, and it is past due. (Source: DHCW Integrated Medium Term Plan 2025-28, Workflow Service deliverables — a published DHCW planning document.)

And the AI revolution in software engineering — the AI-assisted "agentic coding" tools that are reshaping how digital organisations build — is, on DHCW's record, entirely absent. The high-water mark is a single board remark, in 2023, that using AI to help developers was "going at pace" and "something we need to wrestle with professionally as engineers." Then silence. No adoption, no pilot, no policy appears on the public record in the three years since.

There is one place DHCW does operate AI at scale: Microsoft Copilot. It uses it to write its board minutes. And in April 2026, the Executive Director of Strategy, Ifan Evans, used a Copilot prompt — built with Sam Lloyd, the executive who negotiated DHCW's Microsoft contract — to generate the committee's own delivery-confidence assurance commentary, which was then read to the meeting as a "neutral, objective, independent view." (Source: DHCW Performance and Delivery Committee public meeting, 30 April 2026 — from the meeting recording. Exchanges of this kind routinely do not survive into DHCW's published minutes; see How They Falsify the Public Record.) The one operational AI use the national digital body can demonstrate is the automated authorship of its own assurance that delivery is on track.

Tasked, named — and absent from the plan

The external record matches the internal one.

DHCW is a founding partner of the AI Commission — yet the Commission publishes nothing of its own: no standalone reports, minutes or recommendations appear on the public record for 2025 or 2026, only endorsements channelled through the Welsh Government. And against the AI-capability workstream the Commission's terms of reference assigned to DHCW, there is no published delivery record at all.

Then there is the document that ought to be DHCW's charter. In November 2025 the Welsh Government published its national AI strategy, "AI Cymru: Shaping a Smarter, Fairer, More Prosperous Wales" — the AI Plan for Wales. The national body for NHS Wales digital, data and technology is not named in it once. No DHCW deliverables. No DHCW role. The plan delegates health-AI governance to the Commission — now renamed the AI Health and Social Care Advisory Group — and routes the clinical examples through health boards and delivery units. (Source, new tab: gov.wales/ai-plan-for-wales.)

The Senedd has noticed. On 1 October 2025, the Welsh Conservative health spokesperson, Gareth Davies MS, told the chamber that the delivery of AI-enabled tools "has so far been abysmal," noting that DHCW — "responsible for major digital programmes, including AI-enabled tools" — had been escalated to enhanced monitoring in March 2025. (Source, new tab: Senedd Plenary, 1 October 2025, Record of Proceedings.)

A founding partner of the national AI Commission. The body a Senedd member names as responsible for AI tools. Absent from the national AI plan, absent from every delivered AI win, and named — by its own legislature — as "abysmal."

Why the monopoly couldn't: the empty foundation

There is a simple reason a body can be tasked with AI and deliver none of it: AI runs on structured data, and DHCW's data platform does not work.

The National Data Resource — the platform sold as the foundation on which Welsh health AI would run — sits effectively empty at year seven. Around £35.7 million spent of a £73.6 million planned lifetime cost; no Health Board using it in service; an information-governance regime so tangled it generated close to a thousand parallel privacy notices; and a modelled benefits case its own leaders quietly restated downward from £151.5 million to roughly £41 million. The one time an AI model was demonstrated on it — a Google Gemini exercise coding clinical documents in 2024, run on the Google Cloud platform DHCW had made a multi-year, dollar-denominated spending commitment to — it impressed the room and was never productionised.

The fuel tank for Welsh health AI was never filled. The full account is set out in National Data Resource: 10 Years Late, £100 Million Short.

The cost of the monopoly

The asymmetry is the scandal. The AI that worked was delivered cheaply, by other people, around the monopoly. The monopoly itself consumed hundreds of millions and produced a training module.

Since April 2021, DHCW has received approximately £600 million of public money. Its Director of Finance has stated the organisation's total demonstrated return at £0.5 million. The full arithmetic — and what it bought — is set out in £600 Million In. £0.5 Million Out and £757K to £49 Million: The Costs They Hide Keep Growing.

This is what an ineffective money furnace looks like in practice: maximum input, negligible output, and the genuine wins arriving from everywhere the money did not go.

Same leaders, same result

A monopoly that cannot deliver AI is not, in the end, a technology problem. It is a leadership and skills problem — and in DHCW's case, the same leadership throughout.

DHCW's Chief Executive, Helen Thomas, has led NHS Wales informatics throughout — Assistant Director of Information at ABM University Health Board, then Director of Information at NWIS from 2017, its interim Director, and DHCW's founding Chief Executive when NWIS was dissolved into it in April 2021. WIDI credits her with having "spearheaded the creation of the NDR" — the empty flagship at the centre of this story. Her Executive Director of Strategy, Ifan Evans — an Oxford history graduate by training, per his own LinkedIn profile — authored the Welsh Government's digital health strategy and then moved into DHCW to deliver it, a path traced in He Wrote the Strategy. Now He Marks His Own Homework.

Beneath them, the gaps are concrete. No staffed AI programme is visible anywhere in DHCW's public record — no AI leadership role, no delivery team, nothing answering the capability workstream the AI Commission assigned it. Health Education and Improvement Wales's own AI skills review found an AI-literacy gap across NHS Wales and had to recommend clinical fellowships to train doctors to check automated outputs. And DHCW's single answer to its own "AI capability" workstream was a training module commissioned from WIDI — a body its Chief Executive directs. The capability was outsourced; the conflict was not declared; the delivery did not come. The leadership question is examined further in An Honorary Professorship, a Fellowship, and an Award and Wales Deserves World-Class Digital Health Leadership.

In March 2025 the Welsh Government escalated DHCW to enhanced monitoring. By 2026 it had moved to Level 4 — Targeted Intervention, with concerns named as "delivery, accountability and leadership." The same leaders remained in post. The case made for the monopoly, when it was built, was that central control delivers faster than a fragmented system. The record of AI in NHS Wales shows the precise opposite: every part of the fragmented system delivered, and the centre did not.

The recipe for digital success in NHS Wales turned out to be simple. Find a way to skip DHCW.

CareNHS invites responses from any individual or body named in this article. If a response is received, we will publish it in full.

Every load-bearing fact in this article is drawn from a public document, an official press release, a published strategy, or a transcribed public meeting. Correspondence: Rose Davies, CareNHS.