Sanctuary Health
Sanctuary Health builds and runs AI applications for hospitals, NHS trusts, insurers, primary care practices and government health organizations, putting engineers alongside receptionists, medical secretaries, nurses and doctors to find where referrals, correspondence, coding, rotas, inventory and appointments get stuck. The Sanctuary Platform joins the applications it builds with governance dashboards for staff oversight and an integration layer into the organization's existing systems. In its GP telephony example, a phone agent takes a repeat blood pressure prescription request, matches the caller to the patient record and queues it for a clinician to sign off. A mental health companion answers questions from a proprietary knowledge base and escalates signs such as insomnia and low mood to a counselor, and an AI search ranks clinical literature by relevance for a healthcare publisher.
Sanctuary's clients average £1.3 billion in turnover, and it names HealthHero, NHS England and Wellabe among them. Professor Manoj Ramachandran of Barts Health Trust credits the team with removing complexity rather than adding it. Across its work Sanctuary reports 50% average cost savings within 12 months and more than 1,000 hours of unnecessary work removed.
Cofounders James Naylor, Jamie Shepherd and Rob Wells started the London company in 2022 as a healthcare content business, which still licenses clinically reviewed media and content APIs at sanctuaryhealth.io, and Livitay Ltd trades as Sanctuary Health. Work starts with a scoping conversation and an engagement shaped to each organization.
Capability Axes
An AI Health Index grade measures what a buyer can verify from public sources on the date shown. It is not a rating of how good the product is. A vendor can build an excellent system and grade low on an axis because it publishes nothing an outsider can check. How grades read
AI is the engine of the Sanctuary Platform. A hospital or insurer commissions applications for specific workflows, and each depends on a model to interpret a caller's request and match a record, answer from a knowledge base and spot concerning symptoms, or rank retrieved papers. The dashboards and integration layer exist to run and connect those applications.
Without the models the platform has nothing to deploy, since the embedded engineering is how it is delivered rather than what it sells. The content licensing business is a separate line and not part of the platform.
The GP phone agent takes a repeat prescription request, matches the patient and queues it for clinician sign off rather than issuing anything, and the mental health companion routes a patient to a counselor when it detects concerning signs. Staff watch the applications through dashboards, and the processor notice preserves human review for automated decisions with legal or similarly significant effects, except where law or consent allows otherwise. The handoffs are named. The thresholds that trigger them, such as what the companion treats as worth escalating, and how staff override an action the system has already taken vary by deployment and are not described.
The current site describes applications built to fit each customer, a governance layer and integrations, with examples that interpret a caller's request, retrieve answers from a proprietary knowledge base and rank literature by relevance. That explains what the AI does in general terms. No model, provider, version or update practice is named for any current application.
An earlier assistant design in older copy described retrieval, summarization, voice and generation built on OpenAI, Claude, ElevenLabs, Pinecone and LangChain, but those names no longer appear on the live platform or enterprise pages, and nothing ties them to what is sold today.
Google Cloud Platform is the principal subprocessor named in Sanctuary's June 2026 processor notice, which also commits to written data processing agreements, customer authorization before any subprocessor is added and a full transfer register on request. No model provider appears on the current site.
Earlier marketing copy listed OpenAI, Claude, ElevenLabs, Pinecone and LangChain for a previous assistant design, but the live platform and enterprise pages name none of them, and which models serve today's phone agent, mental health companion or literature search is not stated. The host is named, while the model layer and the terms governing patient data once it reaches a model are not.
Professor Manoj Ramachandran of Barts Health Trust praises the team for removing complexity rather than adding it, and NHS England, HealthHero and Wellabe appear among Sanctuary's clients. Across its work the company reports 50% average cost savings within 12 months and more than 1,000 hours of unnecessary work removed, with no definition of the cost base, the number of projects or results for any named organization. The phone, literature search and mental health examples show what applications can do. No dated deployment study compares a named site before and after.
Acting as processor, Sanctuary uses patient information only on the healthcare customer's instructions and never for its own purposes. Personal and health data it holds as processor is kept no longer than 60 days after the contract ends unless the customer sets a different period, access to health information is restricted, and the customer is told without undue delay after a qualifying breach. Those are specific commitments on the data Sanctuary holds.
What the terms do not cover is the model path: whether identifiers are stripped before text reaches a model, how long prompts and outputs are kept, and how an unsafe clinical output from the companion would be handled.
Livitay Ltd, trading as Sanctuary Health, acts as a UK GDPR processor when it handles patient or service user data on the instructions of an NHS trust or healthcare provider, which remains controller under a written data processing agreement. Subprocessors are engaged only with the customer's authorization, and patient rights requests go to the controller. The June 2026 processor notice documents that regime in the depth a business associate arrangement would have in the United States. The content business describes itself as HIPAA compliant, but no US business associate agreement is offered for the platform, which fits its UK focus.
The site displays SOC 2, Cyber Essentials, NHS DTAC and GDPR marks, and the processor notice adds the NHS Data Security and Protection Toolkit, encryption in transit and at rest, multifactor authentication, restricted personnel access and annual penetration testing, with prompt notice to the customer after a breach. SOC 2 is presented as Sanctuary's own, separate from Google's hosting certifications. No report, audit period, scope or renewal date is given, and no trust portal offers the artifacts, so coverage of the specific applications is confirmed only through the customer relationship.
Most engagements concern administration: appointments, referrals, coding, equipment tracking and correspondence. The mental health companion is different. It answers patient questions from a knowledge base and triages them to counseling, with the site's example flagging insomnia and low mood for counselor escalation, and the GP phone agent handles prescription requests that a clinician then signs.
Triage of mental health symptoms is clinical work, yet no regulatory position, device registration or explanation of why the companion sits outside device rules is given. The NHS assurance marks cover deployment and security, not regulatory status.
Governance is built into the platform as its own layer, with dashboards that show staff what each deployed application is doing and workflows designed alongside the receptionists, secretaries and clinicians who run them. Sanctuary also displays NHS DTAC, the NHS assessment covering clinical safety, data protection, security, interoperability and usability for digital health tools. Those give customers a named process and an oversight view.
No subgroup results, bias testing or evaluation by language and caller type is offered for the phone agent or the mental health companion, so how the models perform across patients is left to each customer to establish.
The healthcare customer is the controller and handles patient rights requests with Sanctuary's help, and in use the dashboards and steps such as clinician sign off on a repeat prescription keep a person in a position to review what the system did or queued. That is a mechanism for catching errors. Sanctuary's public app terms attach foreseeable loss provisions to Livitay, its consumer mobile app, and govern personal use only.
Remedies for a misrouted referral, faulty triage or another wrong output in an enterprise deployment, including any warranty or correction deadline, sit in each customer agreement, and none is public.
An integration layer connects each custom application to the customer's existing systems, and earlier Careflow examples included EMR integration, referral management and tracking and automated patient communication. In the GP telephony example, matching a caller to a patient record precedes a task for a clinician, which implies read access to the practice system.
No EHR vendor, interface standard or record operation is named for any deployment, and no production connector is shown with a defined read or write scope, so integrations are described per engagement rather than demonstrated. The content API at sanctuaryhealth.io serves media, not patient records.
Engineers work on site inside the customer's workflows, but the applications run in the cloud. The processor notice names Google Cloud Platform as principal subprocessor, with data hosted in the United States under the UK extension of the EU and US Data Privacy Framework, and a full transfer register available on request, while contractual safeguards cover other parties processing data outside the UK. Hosting location and transfer basis are therefore stated. Isolation between customers, whether an NHS trust can keep processing in the UK, and where model requests are served beyond the host are not described.
A deployment starts with a conversation about an administrative problem, then Sanctuary scopes it, places engineers inside the organization and delivers applications with dashboards and integrations, so cost follows the workflows chosen and the systems to connect. The structure of the engagement is visible before contact. No rate card, per user price, usage rate or recurring platform fee is published for the operational software, and support terms are set per engagement. The licensed educational content library at sanctuaryhealth.io is priced separately under its own licensing offer.
Hospitals, NHS trusts, insurers, primary care and public health organizations are the buyers, with clients averaging £1.3 billion in turnover, and the application range runs from GP reception and mental health support to clinical correspondence, referral coordination, staffing rotas, inventory and equipment tracking and scheduling. Every deployment is built to the organization's own processes, so the application one site gets may differ from another's. The NHS, Barts and HealthHero relationships ground the UK focus. No validated scope maps any application to a specialty, patient population or point in the care pathway.
Alternatives to Sanctuary Health
Other products that do the same kind of work as Sanctuary Health, and what each one does.
Flexbone, based in Atlanta, builds AI agents for administrative operations, primarily in healthcare; its privacy policy describes it as a healthcare automation company working with providers and payors.
GenHealth.ai supplies AI agents that perform healthcare administrative work inside existing applications. The current offer spans intake, eligibility, prior authorization, resupply, billing and denial follow up through web interfaces, fax and telephone.
Paratus Health builds generative voice agents for outpatient facilities, medical practices, and dental groups. The service handles inbound and outbound patient conversations, including appointment scheduling, intake, insurance eligibility, reminders, and routine billing requests.
Talkie.ai sells AI front desk agents to independent medical practices and physician groups.
Arini provides an AI receptionist for dental offices and dental service organizations. It answers inbound calls, checks schedules and books against the appointment types, providers and slots a practice has set up.
Counsel Health is an AI native primary care service. Adults chat with Counsel AI for free about symptoms, lab results, medications and next steps, and add a board certified physician to the same chat when they need a diagnosis, a prescription, labs or a referral.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
| Entry Price | Pricing Basis | BAA Tier | Implementation | Source |
|---|---|---|---|---|
|
Custom engagement
|
Sanctuary scopes each platform engagement around the workflows, engineering work and system connections a hospital, insurer or health organization needs, starting with a scoping conversation. | UK deployments involving patient data run under a written UK GDPR data processing agreement. No US business associate agreement is offered for the platform. | Embedded engineering, integrations and governance dashboards are part of each deployment's scope, with implementation and support priced per engagement. | Vendor Published |
No rate card, per user price or usage rate appears for the operational platform. The educational content library is licensed separately.