Leva Health | Safe AI execution for NHS productivity

The problem

The NHS does not lack information. It lacks the capacity to act on all of it in time.

Operational teams work across fragmented systems, incomplete pathway states, competing priorities and manual coordination. Important issues can be visible without the next action happening quickly enough.

01Signals sit in different systems from the people who act on themUnderstanding one stalled pathway can mean three logins and a phone call. The information exists; assembling it is the work.
02The next step depends on somebody rememberingChasing, checking and escalating are held in the heads of experienced staff, because nothing else is holding them.
03Nobody can say what any single change actually didServices improve for many reasons at once. Without a declared comparator, attribution is guesswork.

How it works

From signal to accountable execution

Seven stages. Two of them are the reason an NHS organisation could let software do any of this at all.

Understand operational state
State is projected from the event history on every read rather than stored and mutated.
Inputs
Synthetic pathway events, in the order they occurred.
Authority
None required. Reading state changes nothing.
Recorded
State version and hash recorded with each projection.
Detect a constraint

What it produces

Every decision leaves a record you can check.

Not a log written after the fact. The record is how the system works: each row states what was decided, on whose authority, and what evidence supported it — and is hashed to the row above.

The extract alongside includes a refusal. Asked to do something that would set clinical priority, the policy engine declined and routed it to a clinician. That is the architecture working, and it is written to the same ledger as everything else.

How the authority model works

thin_slice_v1seed 20260901ledger 8cbed31c9e59f482Synthetic data

The first fourteen audit records from a synthetic demonstrator run
Seq	Elapsed	Class	Event	Outcome	Record
1	02:00	—	Constraint detected	Prerequisite never requested

Bounded by design

Capability is useful only when authority is explicit

Technology analyses, detects and prepares

Interpreting state, identifying constraints, assembling a candidate response with its evidence attached.

People authorise material action

Named roles. Two signatures where impact is higher. An expired approval is never treated as consent.

Clinical decisions stay with clinicians

Diagnosis, treatment, priority, suitability and eligibility are routed, never decided by software.

Safety does not depend on asking a model to behave.

Leva Health treats AI reasoning as a component inside deterministic policy, validation, approval and audit controls. Nothing in the current build calls a model at all — the architecture reserves a bounded place for one, and that place is deliberately empty until it can be shown to earn its keep.

Current focus

Starting where operational friction is visible and consequential.

Our current research focus is elective pathway coordination. We are investigating whether bounded AI can help teams identify operational constraints earlier, distinguish administrative action from clinical judgement, and coordinate the next permitted step across complex pathways.

It is a starting point, not the boundary of the company. Whether the architecture genuinely transfers to other operational environments is one of the questions the research is designed to answer honestly, including the possibility that it does not.

Explore the elective pathway research

Evidence

We separate what works, what is modelled and what remains unknown

Working in the synthetic demonstrator
Mechanics you can exercise now and check against the test suite: detection, policy enforcement, approval, execution, monitoring, recovery and a tamper evident record.
Modelled under declared assumptions
Operational arithmetic whose inputs are stated defaults rather than measurements, shown next to the figures they produce.
To be tested through funded R&D and NHS evaluation
Everything about real data, real workflows, real adoption and real effect. None of it established, none of it claimed.

No economic figures appear on this site. The demonstrator exposes the assumptions and the arithmetic together, so a reviewer can change a load bearing assumption and watch the number move. A figure without its assumptions is not evidence.

No NHS deployment, integration, partner, customer, pilot, certification, saving or patient outcome. None of these is claimed and none exists.

Collaborate

We would rather be told this is wrong now than in two years.

We are speaking with NHS operational leaders, clinical safety specialists, information governance professionals, evaluators and health economists to shape a credible route from synthetic demonstration to real world evidence. Review of the demonstrator takes about 25 minutes and is by individual invitation.

Share an operational perspective See what we are asking

Turn operational insight into safe, accountable action.

The NHS does not lack information. It lacks the capacity to act on all of it in time.

From signal to accountable execution

Understand operational state

Detect a constraint

Every decision leaves a record you can check.

Capability is useful only when authority is explicit

Technology analyses, detects and prepares

People authorise material action

Clinical decisions stay with clinicians

Starting where operational friction is visible and consequential.

We separate what works, what is modelled and what remains unknown

Working in the synthetic demonstrator

Modelled under declared assumptions

To be tested through funded R&D and NHS evaluation

We would rather be told this is wrong now than in two years.

Read further

Propose the permitted next action

Check authority and evidence

Obtain human approval where required

Execute through an approved integration

Record, monitor and evaluate