Alcidion

Miya Precision is integration and workflow software first. It consolidates real time data from a hospital's existing systems into one record and runs journey boards, bed management, observations, emergency department workflow, virtual care and, at some sites, a full electronic patient record. Most of that works without a model, and the independent evaluation at Alfred Health attributes its results to the journey boards and two way integration.

The learned layer is growing on top. Concept Detection, inside the Miya Noting module, reads clinicians' free text notes, identifies medical concepts and suggests SNOMED CT codes. Miya Scribe captures dialogue into note templates and Miya Insight condenses the integrated record into a short summary, both developed with Google Cloud. A proprietary Miya AI Service module hosts these features. The company also describes long standing rules based intelligence for deterioration detection, predictive analytics and discharge planning, which supports the AI rather than replacing the workflow software underneath.

Alcidion's AI features suggest and clinicians decide. Concept Detection proposes concepts and SNOMED CT codes, and each one needs clinician validation before it is added to the patient's record. Miya Scribe fills Miya Noting templates that wait for clinical review, and the coded problems it extracts are presented to the clinician for confirmation. On the flow side, staff act on the journey boards themselves, and at Alfred Health a change to an estimated discharge date requires a reason.

What a clinician sees when the system is unsure, how often suggestions are rejected, and whether drafted text is marked as machine generated once it is in the record are not described. For a documentation assistant those details decide whether review stays meaningful at volume.

Miya Precision works in layers. It converts data from a hospital's existing systems into FHIR resources in real time, maps it to clinical terminologies, and runs a clinical decision support engine that suggests actions from local protocols and the patient's own data. The company says this older layer is rules based, built on terminology management, workflow orchestration and natural language processing over clinical notes, and that newer features run through the Miya AI Service, a proprietary module for deploying AI inside customer environments. Third party algorithms can be plugged into the platform and managed there too.

Concept Detection has a narrow, defined task: free text notes in, medical concepts and suggested SNOMED CT codes out. Miya Scribe and Miya Insight were built with Gemini on Google Cloud's Vertex AI, according to Alcidion's March 2025 announcement.

No model version, training data, update cadence or accuracy figure is given for Concept Detection, the decision support or either generative product.

The generative features run on Google Cloud. Alcidion's March 2025 announcement says Miya Scribe and Miya Insight were built with Gemini on Vertex AI, and its 2026 annual report says both are being brought forward with Google. Concept Detection and later features run through Alcidion's own Miya AI Service, and the platform can host third party algorithms beside its own.

Beyond the model layer the chain is thin. Alcidion lists no subprocessors and does not name the provider behind Alcidion Cloud, and its standard terms allow subcontractors to process customer data without naming them. Whether customer data trains any model, and whether notes and transcripts sent to Gemini stay in the customer's country, are not stated. NHS trusts and Australian state health services treat where processing happens as part of procurement, so those answers come up early in any deal.

Two independent studies sit behind the platform, both about its flow and virtual care workflows rather than its AI. A Monash University evaluation of Miya Precision journey boards at Alfred Health, funded through the Digital Health Cooperative Research Centre, reported length of stay down 12.1 percent over 18 months, outliers at admission down 17.7 percent and bed management call time down 67 percent, comparing one site before and after.

In June 2026 the ANZ Journal of Surgery published the iCAD trial, a Phase I/II randomized feasibility study at Sydney Local Health District in which Miya Precision ran the virtual care pathway for acute uncomplicated diverticulitis. Across both phases 93 percent of virtual care patients completed treatment without crossing to inpatient care, there were no major complications, and 88 percent rated their care excellent against 40 percent of inpatients. The model is now standard practice in the district.

The company also reports results of its own: doctors at South Tees saving up to an hour a day on administration, and stays about five days shorter during the first phase at Herefordshire and Worcestershire's community hospitals. Concept Detection and the generative features have no reported accuracy figures.

Access control is where the platform is specific. Patient records use claims based authorization, and release 7.11.0 in September 2026 added a three tier access model in which break glass access needs a justification and an individual PIN and is audited. Alcidion's standard terms keep patient data in the production instance unless agreed otherwise, encrypt backups at rest, and rule out patient data in the monitoring it runs on customer systems. The company describes the Miya AI Service as a secure, controlled architecture for running AI inside healthcare environments, and its UK privacy policy sets out its role as a data processor for NHS trusts.

How long patient data is kept, whether it is used to train or tune any model, and whether it is de identified before reaching the generative features built with Google are not stated.

HIPAA does not reach Alcidion's markets; data protection law does. The privacy policy of Alcidion UK Limited says the company processes patient identifiable information as a data processor for NHS customers under UK data protection law, with the trust responsible for the patient's privacy notice. The company also has Australian and New Zealand policies, and the UK policy itself covers people who work for customers, prospects and suppliers.

The data processing agreement itself, its scope and any terms binding subcontractors are not public, so the contractual handling of patient data sits in each customer's contract.

Alcidion holds ISO/IEC 27001 certification and Cyber Essentials Plus, both renewed during its 2026 financial year, and its annual report describes regular penetration testing, cyber insurance and incident response processes alongside them. Its UK listings on the G-Cloud 15 framework name the British Assessment Bureau as the certifying body for ISO 27001, give a Cyber Essentials Plus certificate number, and cite the NHS Data Security and Protection Toolkit under organization code 8J464.

Inside the product, records use claims based authorization, and the September 2026 release added restricted access tiers in which break glass access needs a justification and a PIN and is audited.

There is no trust center or security page on the website, and the scope of the ISO 27001 certificate, including whether it reaches the Australian business and the AI services, is not shown.

Alcidion has taken an AI feature through device registration in both of its main markets. Concept Detection is included on the Australian Register of Therapeutic Goods as a Class I medical device (ARTG 522634) and registered with the MHRA in the United Kingdom as Class I, announced in February 2026. The company says providers in both countries increasingly require that registration before adopting AI tools, and its annual report says it registers products as software as a medical device where applicable.

The rest of the decision support carries no stated position. Patient flow, bed management and journey boards are operational tools, but deterioration detection and alerts on clinical issues can fall under device rules in both countries, and Miya Scribe and Miya Insight are not described as registered.

An AI Council set up during Alcidion's 2026 financial year oversees the responsible use of AI, weighing opportunities and risks against regulatory, security and ethical requirements, and the annual report describes responsible AI policies with human in the loop controls and data governance. Concept Detection's device registration brings its development under the quality and safety requirements of device rules.

No validation results, monitoring practice or subgroup analysis is published for any AI feature. Concept detection and summarization built on one health system's notes can behave differently in another with different documentation habits and coding, and no results are given by country or by patient population.

Every AI output passes a clinician before it counts. Each Concept Detection suggestion needs validation before it reaches the record, and scribe drafts and the problems extracted from them wait for review. Alcidion's own risk statement lists model reliability, explainability and the unintended consequences of AI assisted decisions as risks, with those human in the loop controls as the mitigation.

The standard terms on its G-Cloud listings put the rest on the customer. The customer must ensure that all clinical rules and algorithms configured in the services are appropriate for their purpose, Alcidion does not promise the software is fault free or fit for a particular purpose, and liability is capped at the fees paid in the first 12 months, or £100,000 for data protection breaches.

No accuracy or error figure is published for Concept Detection, the decision support or the generative features, and there is no correction route a patient or clinician can use against the vendor. Contracts also run long: University Hospitals Sussex signed for seven years with an option to ten, and the AI inside a platform on that term will change several times.

Integration is the platform's core job. Miya Precision integrates in real time with a hospital's record and administration systems, converts their data to FHIR and serves FHIR resources to other applications. Named examples include real time integration with Oracle Cerner at Peninsula Health, the SA Health statewide record at Northern Adelaide Local Health Network, and two way FHIR integration with the record and patient administration systems at Alfred Health, where estimated discharge dates write back to both. Gold Coast Health's remote monitoring connects to its existing record. At University Hospitals Sussex and North Cumbria the platform is the record itself.

Integration documentation, an API reference and marketplace listings are not public.

Customers choose hosting in Alcidion's cloud, on premise or a hybrid of the two, and the company's 2026 annual report notes a growing number moving to Alcidion's cloud. For UK customers the location is committed: Alcidion's G-Cloud listings state that data is stored and processed in the United Kingdom, on Alcidion Cloud infrastructure backed by industry partners in a UK data center, and its standard terms set a 99.95 percent availability target for the production environment. Unless agreed otherwise, patient data is held only in the production instance rather than in test environments, and backups are encrypted at rest.

Many customers of the acquired Kyra products still run older on premise systems, with a path onto the cloud platform offered when they are ready.

The cloud provider behind Alcidion Cloud, the tenancy model, and where Australian and New Zealand customers' data is hosted are not stated. The standard terms also allow third parties to run parts of Alcidion Cloud, without naming them.

In the UK, Alcidion sells through the government's G-Cloud 15 framework, and its Miya Patient Flow listing carries a price list dated January 2026. Fees are set by hospital size and charged per bed unless otherwise agreed: £808,500 a year at 500 beds, £924,000 at 1,000 beds and £1,039,500 at 1,500 beds, stepping up about 5 percent in each of the next two years, or £2,548,797 to £3,277,025 over a three year term. The fee covers the software license, maintenance and round the clock support, delivered as a managed service on Alcidion's cloud.

Implementation and professional services are charged on top at a blended £1,100 a day, data migration and change management are scoped separately, and terms under 36 months are prorated. Alcidion reserves the right to reprice each opportunity by scope, and its standard terms carry an early termination fee. Its other UK listings, Miya Noting and Better Meds electronic prescribing among them, link pricing documents of their own.

Outside the UK the company reports contract values rather than prices. The University Hospitals Sussex record agreement signed in 2026 is worth about A$35 million over seven years, extendable to about A$49 million over ten, and an expansion with Leidos for the Australian Defence Force added A$12.3 million. How the AI features are licensed alongside the platform is not stated.

Public health systems in the United Kingdom, Australia and New Zealand are the customer base, with early work in Canada and the Middle East. Named deployments cover inpatient flow (Alfred Health, Western Health, Peninsula Health, Northern Adelaide Local Health Network), emergency departments (the major trauma center at Southampton General Hospital and three departments at Hampshire Hospitals), trust wide platforms (South Tees, North Cumbria Integrated Care, and University Hospitals Sussex from 2027), community hospitals (Herefordshire and Worcestershire), virtual care (Sydney Virtual Hospital, Gold Coast Health, Goulburn Valley Health) and military health through Leidos for the Commonwealth of Australia.

The Kyra products bought from Telstra Health in 2026 add outpatient queue and referral management and more than 30 Australian hospital and health service customers, most of Queensland's among them.

Independent studies cover inpatient flow and acute virtual care. The emergency department, military health and AI work rests on deployment reports, and clinical specialties are not listed.