The NHS is moving much faster than charities on AI

Here is something that should bother charity leaders. The NHS, an institution whose procurement processes are legendary for their slowness, where IT systems still run on decades-old infrastructure, and where 45% of services lack a digital pathway, is moving faster on AI than the charity sector.

Not experimenting with AI. Deploying it. In clinical settings, at scale, with patients.

AI-assisted chest X-ray analysis is now live in 64 NHS trusts through the AI Diagnostic Fund, checking for over 120 conditions including cancers and pneumonia. The world's largest AI mammography trial launched in early 2025 across thirty centres, testing whether AI can safely replace one human reader in breast screening across 462,000 studies. AI tools are predicting kidney failure at Sheffield Teaching Hospitals six times faster than manual methods. Newcastle Hospitals became the first in the world to trial AI that predicts how burns will heal. And that's just radiology. In pathology, the National Pathology Imaging Co-operative has gone live across three major London trusts, with digital pathology networks expanding in Devon, Cornwall, and across Wales.

The charity sector, by contrast, has 76% of organisations "using AI." Mostly for drafting emails and grant applications.

The infrastructure gap

The difference is the infrastructure behind them.

The NHS invested approximately £250 million in the AI Lab from 2019 onwards. It built the Federated Data Platform, a £330 million data infrastructure programme now live across 73 trusts and 41 integrated care boards. It created specific procurement frameworks for AI, including a £150 million Healthcare AI Solutions framework led by NHS Shared Business Services. It published deployment guidance and minimum standards. It established a national registry of vetted AI suppliers in January 2026. The 10-Year Health Plan, published in July 2025, includes a dedicated AI strategic roadmap with nationwide deployment of validated algorithms planned from 2027.

This is what treating AI as a strategic investment looks like. Not someone in the digital team trying ChatGPT on their lunch break. Dedicated funding. Purpose-built procurement routes. Data infrastructure laid before AI tools are deployed on top of it. A national strategy with a timeline and accountability.

Now compare that with the charity sector's position. The 2025 Charity Digital Skills Report tells the story clearly. Yes, 76% of charities are using AI tools. But only 2% at a strategic level. Just 44% have a digital strategy, down from 50% the previous year. Board digital skills are poor at 28% of charities, and 36% say their CEO has poor AI skills. Only 48% have an AI policy, and 37% have taken no actions at all to move forward with AI.

The NHS faced the same objections charities raise. Governance is complex. Data quality is inconsistent. Staff are overstretched. Budgets are tight. Cultural resistance is real. And yet the NHS has over 80 AI-driven innovations in deployment, with more coming. What did it do differently?

What the NHS got right (and what it got wrong)

It would be dishonest to pretend the comparison is straightforward. The NHS has advantages charities lack: central coordination through NHS England (though NHS England is now being absorbed into the Department of Health and Social Care in a two-year process begun in 2025), government funding commitments, and the political pressure of waiting lists and workforce shortages driving urgency. The 470 consultant radiologist vacancies and £216 million annual spend on outsourced scan reporting created a burning platform that most charities don't have.

And the NHS hasn't got everything right. Contracting for AI tools took four to ten months longer than anticipated. Adoption of the Federated Data Platform has been uneven, with several major trusts pushing back. The BMA raised concerns about the Palantir contract. Shifting objectives and limited capacity slowed the AI Lab's work. An independent evaluation has been commissioned precisely because the scale and complexity of the programme demands scrutiny.

But three things the NHS did are directly relevant to charities planning for 2026 and 2027.

It built data infrastructure before deploying AI at scale. The Federated Data Platform isn't glamorous, but without connected, accessible data, AI tools have nothing useful to work with. Most charities haven't done this foundational work. Their data sits in disconnected spreadsheets, inconsistent CRMs, and email inboxes. No AI strategy will succeed on that foundation.

It created dedicated procurement pathways for AI rather than forcing AI purchases through frameworks designed for traditional IT. Charity procurement processes, where they exist at all, aren't designed for technology that changes every few months. Some charities are still running CRM procurement processes that take eighteen months. AI needs a different approach: smaller commitments, faster evaluation, willingness to test before committing.

And it treated AI as a strategic priority with dedicated capacity. The NHS created the AI Lab, appointed AI leads, and ring-fenced funding. In the charity sector, AI responsibility typically falls to whoever in the digital team has bandwidth, which often means nobody. A strategic priority without dedicated capacity is just a wish.

The planning question

None of this is a criticism. Charities operate with a fraction of the NHS's resources, and nobody expects a housing charity or a hospice to build a national AI deployment platform. The scale is different.

But the principles aren't. If a charity's plan for 2026 or 2027 includes anything about AI, three questions are worth asking. Is there a line in the budget for data quality work, even a modest one? Who, specifically, is responsible for AI progress, and do they have time allocated to it? And does the board understand enough about AI to provide meaningful oversight, or are they rubber-stamping decisions they don't fully grasp?

The NHS didn't wait until AI was mature, safe, and simple before committing. It started with the messy, expensive groundwork of getting data infrastructure and governance in place. Most of that work isn't visible and none of it is exciting. But it's the reason the NHS can deploy AI in clinical settings while the charity sector is still mostly using it to rewrite thank-you emails.

The planning window is open. The question for charity leaders is whether they're willing to invest in the boring parts that make the interesting parts possible.