Nightingale AI
Review workspace
Last year the NHS paid £3.24 billion for clinical negligence on a record 15,236 new claims. Every pound was spent after the harm had already happened. Nightingale AI joins the incident reports, claims and published judgments the NHS already holds, so systemic risk is visible first — and a named person decides what to do, every time.
The National Audit Office and the Public Accounts Committee have put clinical negligence cost on the political agenda and asked the Department of Health and Social Care for a plan to reduce patient harm and manage the cost. Nightingale AI is a concrete answer to that question.
Sources: NHS Resolution, Annual Report and Accounts 2025/26 (HC 421), Table 1, July 2026; National Audit Office, Costs of clinical negligence, October 2025; Public Accounts Committee report, January 2026.
The national Learn from Patient Safety Events service records what went wrong, where, and how often.
What the harm cost, how long it took to surface, and how it was resolved.
How courts have decided comparable facts: standard of care, causation, consent.
Central Alerting System notices and MHRA data that should change what a Trust watches for.
Four datasets. Four owners. Four formats. Nobody connects the incident to the claim to the judgment — so risk is only ever visible in hindsight. Nightingale is the layer that joins them.
Deterministic matching on identifiers, dates, organisation and event codes — exact, explainable rules, no model guesswork. Every flag names the fields it matched and the score it earned. Nothing is merged without a records owner confirming.
Incident counts are compared with the organisation's own baseline. A pattern above threshold raises a banded risk note at the level of a pathway, ward or service — a signal for governance review, never a finding about a person.
Retrieved from The National Archives' Find Case Law service: exact passages, paragraph references and links, under the Open Justice Licence. It cannot cite a case that does not exist — and when there are no close comparables, it says so.
The only place a decision happens. It cannot be skipped and cannot be automated. Every decision, and the evidence pack that follows it, lands on an audit trail that nobody — including us — can edit.
Never determines negligence, settles a claim, starts an investigation, names a clinician or alters a care pathway.
Draft → assessed → under review → accepted, overridden or returned. Nothing auto-progresses; overrides always carry a written reason.
Every load, agent run, decision and export is recorded and cannot be edited or deleted — not by reviewers, administrators or the system's makers.
Every output carries its sources, generator and timestamps. Dossiers export as self-contained evidence packs, readable by a governance panel.
Every record is bound to an organisation code and classified; queries without one are blocked before they run. Built for DSPT and PSIRF.
No real patient-level data enters the system or any training pipeline. Case law is used as metadata and short attributed quotes under the Open Justice Licence.
Most AI-for-NHS proposals bolt safety on afterwards. Here it is the product — which is what makes Nightingale deployable inside the NHS, not just demonstrable.
Systemic risk is seen before more people are hurt by it. When harm does happen, earlier evidence means earlier explanations and earlier admissions.
Risk is framed at the level of a pathway or service — Just Culture, as PSIRF already asks. Duplicate paperwork disappears.
Fewer hours reconciling duplicate filings. Evidence packs ready for PSIRF reviews and legal meetings instead of assembled by hand.
Lower defence costs on claims that should settle early; stronger evidence on claims that should be defended.
A direct route into an annual cost and a £60bn+ liability that the NAO and the Public Accounts Committee have both flagged.
UK-built, auditable AI in the public sector, with a named human accountable for every decision — the standard the 10 Year Health Plan asks for.
Analogue to digital. Treatment to prevention. Preventing harm from systemic risk is prevention. Read the plan
The Government's route to fund UK-built AI that fixes public services. Nightingale is positioned for the NHS productivity challenge. About the scheme
Costs have tripled; liability has quadrupled. Parliament has asked the Department for a plan. A room that already knows the problem is ready for an answer.
Connect one Trust's incident and claims feed inside its own tenant. Outputs are reviewed, nothing is actioned. We measure precision, override rate and lead time — the numbers a board can act on.
To NHS Resolution, the Department of Health and Social Care, or a patient-safety lead who wants to see systemic risk before it becomes a claim.
The live workspace runs on synthetic data. Ask for reviewer credentials and walk the loop yourself.