Leva Health exists to help NHS teams turn fragmented operational signals into coordinated action. Our ambition is not to replace clinical judgement or automate responsibility. It is to give operational teams better execution capability: understanding what is happening, identifying what is getting in the way, preparing the appropriate next step and ensuring every action remains within explicit human authority.
Our mission
Help NHS teams release capacity and improve patient flow by turning operational signals into safe, accountable action.
Productivity is not simply asking people to work faster. It is reducing avoidable coordination, preventing important tasks from being lost between systems and helping skilled NHS staff spend more time on the work that needs their judgement.
The execution gap
Healthcare organisations generate vast amounts of operational information, but insight alone does not change a pathway. Someone still has to understand the issue, establish the appropriate action, check who has authority, coordinate the people and systems involved, and verify the result.
Leva Health is being built around that gap. We believe the next important generation of healthcare AI will not be defined only by what a model can recommend. It will be defined by whether technology can participate safely in real operational work while remaining explainable, constrained and accountable.
An execution layer for NHS operations
Leva Health is developing an operational execution architecture that can interpret pathway state, identify constraints, propose administrative responses, apply deterministic authority policies, route material actions for approval and preserve a traceable evidence record.
The AI component is deliberately hemmed in. It may support interpretation and planning, but it does not own authority. Policy, validation, approval and audit controls determine what can happen next.
Operational data and pathway state
Synthetic events projected into canonical state, versioned and replayable.
Detection and bounded reasoning
Deterministic detectors today. One validated place is reserved for AI, inside the same controls; it is currently empty.
Authority, policy and approval controls
Deny by default. Class resolved from the catalogue, not chosen by the planner. Named human approval where required.
Execution, monitoring and evidence
Idempotent execution through simulated connectors, outcome monitoring, and a tamper evident record.
Beginning with follow-up integrity
A review that a clinician intended and that nobody booked exposes the challenge clearly: the intention and the booking sit in different systems, owned by different teams, and nothing compares them. The first Leva Health demonstrator uses synthetic pathway data to test the underlying execution mechanics on exactly that case. Ophthalmology is where the published evidence is clearest, which is a reason to start there rather than evidence about anywhere else.
It demonstrates how an operational issue can be detected, how a permitted response can be prepared, how prohibited clinical action is refused, how human approval controls material action and how the resulting record can be checked for tampering. It does not claim real NHS savings, outcomes, integration or validation. Those are questions for prospective research in appropriate NHS environments.
How we build
- Human authority is explicit
- Responsibility is never hidden inside an AI response.
- Clinical judgement stays clinical
- The system must not cross into diagnosis, treatment or clinical prioritisation without the appropriate authorised professional.
- Evidence travels with the action
- Important outputs should be traceable to their inputs, rules, assumptions and approvals.
- Unknown means unknown
- Modelled opportunity is not presented as an achieved outcome.
- Safety is architectural
- Controls must be testable outside the prompt and model.
- Evaluation starts before deployment
- Comparators, success metrics and failure criteria should be specified in advance.
- Productivity should benefit staff and services
- Released capacity matters only if it can be used meaningfully.
The company behind it
Leva Health is developed by Orbixa Technologies Ltd, a UK software company registered in England and Wales under company number 16519488. Orbixa is the legal entity behind this work and the data controller for personal data collected through this site. The privacy notice says what that means in practice and what is still outstanding.
We are building technical, clinical, operational and evaluation capability in the United Kingdom, and assembling a multidisciplinary delivery network across NHS operations, clinical safety, information governance, technical architecture, evaluation and health economics.
Named advisers, delivery partners and collaborating organisations will appear here only once they have agreed to be publicly associated with Leva Health. We will not list anyone who has not.
Help us test the right questions.
We want to work with people who understand the operational reality: NHS teams, clinical safety and governance specialists, researchers, evaluators and technology leaders. If the problem is relevant to your work, we would value your perspective.