Technology
The intelligent twin, governed by medicine.
Averra’s twin is a private model of your health. It updates as new readings arrive, spots changes that matter and helps clinicians decide what to do next. Here is how it is designed to work.
Signals in
No single reading tells the whole story.
The twin is useful because it combines signals, and because each one reaches it with the person’s permission.
Wearables
Read from Apple Health and Health Connect by Averra’s own integration layer, with permission and without a third-party aggregator.
At-home blood tests
A simple finger-prick kit used at home and posted to a laboratory. No clinic visit is needed to confirm a change.
Health records
Shared by the person with consent, and normalised to the FHIR R4 international standard.
Symptom checks
Symptom assessment is designed to run on certified medical-device software, not on a general-purpose chatbot.
The model
Seven body systems, one record.
Each module scores named clinical factors. Risk is stratified with an established population methodology used by health systems worldwide, and thresholds are referenced to NICE, ADA, GINA, GOLD and AASM guidance.
- Cardiovascular
- Metabolic
- Respiratory
- Brain and mental wellbeing
- Musculoskeletal and mobility
- Sleep and recovery
- Women’s health
Decisions out
One escalation path, four levels.
Every module drives a single ladder. Every change of level records the factors that fired, their weights and the rule version, so each decision can be explained after the fact.
Illustration
Your normal, and the moment it changes
- Your usual range
- A reading over 30 days
- A change is noticed: a home test checks it
- L0
Monitor
Keep watching the personal baseline.
- L1
Confirm
Something has changed. Check it properly.
- L2
Doctor
The evidence warrants a conversation.
- L3
Urgent
Act now.
Prime Edge AI
Reasoning that starts on your phone.
Averra’s AI runs in two tiers, presented to the person as one conversation.
- On the device. Routine reasoning runs on the member’s own phone: private and fast.
- In the cloud, by exception. Harder questions go to a more powerful AI model in Averra’s secure cloud.
- Hardware attestation. The cloud tier only accepts requests from genuine, unmodified app installs on real devices.
Architecture detail is available to partners under a non-disclosure agreement.
Powered by Prime Assist
Yesterday, 21:10
What moved my metabolic score?
Meal timing, mostly — 4 of your last 7 dinners landed after 22:00.
Today, 07:58
Why is my readiness lower this morning?
Readiness is 74, down 8 from your weekly average. The main driver is HRV — 38 ms last night vs your 46 ms baseline. Sleep length was normal.
Late meals preceded 4 of your last 5 low-HRV nights. Yesterday you logged dinner at 22:40.
- Oura · nightly HRV · 30 nights
- Meals · self-logged
twin-insight v1.4 · not a diagnosisWhy?
The Averra app · demonstration data
Omnichannel AI assistant
One assistant. Every conversation.
Prime Assist is Averra’s AI assistant. It will help members on phone calls, WhatsApp, iMessage and in-app chat, in their own language, from customer service to aftercare, and hands every decision that matters to a person.
- Designed to answer straight away. Day or night, with no queue.
- Any language, even a mix. It replies in the member’s own language, and keeps up when they switch mid-sentence.
- Always says it’s an AI. At the start of every conversation, with a route to a person whenever they want one.
- Escalates on written rules. Urgent, clinical or sensitive conversations go straight to a named person, with the full transcript.
- Never gives clinical advice. It arranges care. Licensed clinicians provide it.
Prime Assist AIHello, I’m Prime Assist, an AI assistant from Averra, the health service that comes with your cover. Is now a good time to talk?
MemberYes, go ahead.
Prime Assist AIYour recent readings have shifted a little from your normal. Nothing urgent, but worth checking. Shall I send you a home test kit?
MemberPlease do.
Prime Assist AIDone. When your result is in, a clinician will explain it in a short video, in your language.
Home test kit ordered · Follow-up arranged
Talk to a person, any time
Prime Assist AIHi, it’s Prime Assist from Averra, an AI assistant. How are you feeling after yesterday’s appointment?
MemberMuch better, thanks. When is my follow-up?
Prime Assist AIThursday at 10:30, by video with the same doctor. Shall I remind you the evening before?
MemberYes please.
Follow-up confirmed · Reminder set
Talk to a person, any time
MemberHola, ¿mi resultado ya está listo? I did the test on Monday.
Prime Assist AI¡Hola! Yes, it’s ready. Hay un video corto de un médico que lo explica. ¿Te lo envío aquí?
MemberSí, please. ¡Gracias!
Prime Assist AISent. Si tienes preguntas, I can book you a call with a doctor.
Video sent · Spanish and English, mixed
Talk to a person, any time
AskPowered by Prime Assist AI
Is my next physiotherapy session covered?
Yes, it’s pre-approved and the clinic has been told. Would you like to book it?
Tuesday evening, if possible.
Booked for Tuesday at 18:00, and added to your calendar.
Cover confirmed · Session booked
Talk to a person, any time
Any language. Even a mix of languages in one conversation.
Customer service
Questions about cover, claims, documents and appointments, answered at any hour, in any language.
Outreach when something changes
When the twin notices a change, Prime Assist calls or messages the member on the channel they prefer, explains the next step and arranges a home test.
Care, arranged
A doctor set up when a symptom is mentioned, and the right place found and booked for any test, scan or therapy the doctor advises.
Aftercare and follow-up
Check-ins after a test or a consultation, reminders, and follow-ups booked before they are forgotten.
A named person, when it matters
Urgent, clinical, distressed or out-of-scope conversations go straight to a named person, with the full transcript.
Every word on record
Each conversation leaves a transcript, timestamped actions and a note written back to the record.
Agentic AI
Agents that act. Clinicians who decide.
Averra uses AI agents for defined jobs that would otherwise need a contact centre, a booking desk and a pre-approval team. Each works inside limits set by clinicians, and hands every decision that matters to a person.
AI
Detect
The twin notices a member drifting from their own baseline, or claims intelligence finds a line that doesn’t fit the case-mix.
AI agents
Act
Prime Assist calls or messages the member and books what is needed. Other agents assemble the evidence.
People
Decide
A licensed clinician, or your own team, makes the call. Never the AI alone.
System
Record
Every step is logged with its factors, weights, rule version and evidence, ready for audit.
Clear limits
What our AI never does.
Trust in AI comes from knowing where it stops. These limits are part of the design.
- Never diagnose or prescribe.Licensed clinicians diagnose and treat. Averra flags, routes and explains.
- Never price, underwrite or decide cover.Member health data is never used to set a price, decide cover or assess a claim.
- Never make a care decision on its own.Clinicians decide. Averra arranges the care they advise.
- Never pretend to be human.Prime Assist tells members it is an AI at the start of every conversation.
- Never raise an alarm from a wearable alone.Wearable data can prompt a check, never an urgent state, without confirmed evidence.
- Never pass members’ raw data to insurers.Insurers see verified outcomes and aggregated insight, not raw health data.
The explanation layer
A flag is only useful if you understand it.
Clinicians record the explanation of a condition or an out-of-range result once. When a person’s data changes, the twin serves the right recording, in their own language. Adding a language is a recording session.
In your language
Explanations recorded by licensed clinicians and served in the member’s own language.
General, then personal
Recordings are general clinical education. Anything specific to a person goes to a licensed clinician.
A path to care
Every explanation sits inside a route to a test or a doctor, so understanding leads somewhere.
Rules the engine obeys
Safety is designed in.
Every deployment is designed to have a documented clinical safety case, owned by a practising doctor and signed off before go-live, following recognised clinical-safety standards. Medical-device classification is confirmed before each deployment.
Wearables raise amber, never red.
Consumer wearable data is treated as wellness-grade. On its own it can prompt a confirmation step, but it can never trigger an urgent state.
Every red rests on evidence.
A doctor-level or urgent state needs a confirmed diagnostic result or a hard rule signed by a clinician.
Medicine owns the thresholds.
Every weight, cut-point and rule is versioned configuration owned by the medical function, with an audit trail. Not by engineering.
Every decision is explainable.
Each change of level records the factors that fired, their values and weights, the time and the rule version.
See the full architecture.
Partners and investors can see the technical architecture, the data flow and a product demonstration under a non-disclosure agreement.