Clinical AI, built by every care team
Describe the tool you need in one sentence. Saana assembles it from validated building blocks, on a compliant foundation, and a physician signs it off before it goes live.
A private practice, a community care network, a hospital ward: each has its own way of following up, reaching out and coordinating. Almost none has a tool that fits.
Most still track patients in a spreadsheet, a personal calendar and a messaging app. It isn't secure, shared or compliant, but it is theirs. The alternative, buying a vendor's software, means adopting someone else's protocol. The gap has two causes.
Compliance now costs more than code. Certified health-data hosting, GDPR and HIPAA, the EU AI Act, medical-device rules: every new tool walks the same road alone, for years.
Software is designed for populations, not for teams. Custom builds exist, but they leave out independent practitioners and small structures.
Saana separates the two. What everyone shares, compliance, is done once in the foundation. What is unique to each team, the protocol, is simply described.
What the engine changes, in four moves
Describe
In plain language, with no spec and no developer.
Assemble
Tiles, queues, rules and campaigns, checked against your data.
Sign off
An agent checks compliance; a named physician approves.
Watch over
Rules monitor between visits and alert the team.
The engine
One sentence becomes a tool.
IIThe foundation
Compliance, done once for everyone.
IIISafety
Three gates before the first patient.
IVPrograms
What teams are already building.
VWho it's for
Six kinds of teams, one foundation.
VIIntegrations
It plugs into what you have.
VIIGetting started
First program live in six weeks.
The engine
One sentence becomes a toolAn independent dietitian's needs look nothing like a care-network coordinator's. The first wants to see this month's revenue, the patients to follow up with, and a reminder campaign before summer. The second wants to know which reimbursement claims are pending with payers, and to produce her regulatory report without losing a day.
The engine handles both requests the same way. It reads the sentence, extracts each intent, picks a validated block for it and checks it against the team's data. A block that can't be computed is never invented: it is flagged "to be defined", and the engine asks a question.
Every intent in the sentence becomes a block
The foundation
Compliance, done onceHosting health data in a certified, in-region data center is not, technically, the hard part. What costs is redoing, for every tool, the impact assessment, the EU AI Act file, consent management and audit trails. The foundation carries all of it once.
Saana hosts the data in the team's region, or connects to the certified host it already uses through an agent installed on its side. Data is normalized to FHIR, and the AI runs on our own GPUs with open models. It never sees a patient's name.
A cross-section of the foundation: six layers, as many tools as there are teams
Safety
Three gates before the patientKeeping an AI-built tool medically coherent, making sure it never gives advice nobody has validated, and knowing when to hand over to a human: that is the hard problem. We turned it into three rules that apply to everything built on Saana.
Sign-in with a verified professional identity, such as a national health-professional card. Every published tool carries the name of the person who approved it.
Saana decides what data can be imported and how far a tool can go.
Until internal testing is conclusive, the assistant stays technical and administrative.
Failing scenarios are stopped at the gate, fixed, then replayed
Programs
One vocabulary, many toolsCOORDIAB, run by the DREAM association and a university hospital in Montpellier, France, operates its entire type 2 diabetes pathway on Saana: QR-code enrollment, shared education assessment, workshops and follow-up across five community centers. More than 200 patients and 30 professionals use it, with no IT team.
On the same foundation, a global pharmaceutical company is running a pilot to screen people with diabetes for cardiovascular risk and alert their care teams.
Each program below is a different arrangement of the same blocks.
Type 2 diabetes support
Enrollment, shared assessment, workshops, follow-up.
Cardiovascular risk
FHIR risk score, validated protocol, care-team alerts.
Practice dashboard
Revenue, appointments, follow-ups, campaigns.
Claims & reporting
Pending reimbursements, regulatory reports, delay alerts.
Intake & triage
Questionnaire before the first session, summary ready on arrival.
Post-discharge follow-up
Day 7 and day 30 check-ins, escalation to the ward.
Who it's for
Six kinds of teams, one foundationWe started with care networks and hospitals because they needed coordination most. The teams asking for Saana now are more varied: independent practitioners who want their own dashboard, clinics that want intake and follow-up tailored to their specialty, digital-health startups that need a compliant product without rebuilding the infrastructure.
Dietitians, psychologists, GPs, nurses
My revenue, my follow-ups, and a reminder for patients I haven't seen.
Coordination across a territory
Shared queues, pending claims, and the report our funder asks for.
Ward tools and care transitions
Call every patient at day 7 and day 30 after discharge.
Patient programs and real-world data
Find patients at risk and support their care teams, under protocol.
Specialty and outpatient clinics
Intake before the visit, follow-up after it, and our activity at a glance.
A compliant product, faster
Launch our care pathway without building hosting, identity and audit.
Integrations
It plugs into what you haveA team's data already lives somewhere: an electronic health record, a lab portal, a calendar, a billing tool, often a spreadsheet. Saana connects to each source, normalizes it to FHIR, and keeps the original system as the source of truth.
When a team already has a certified host, a lightweight agent installed on its side does the connecting, so data never has to leave it.
Every source flows into one normalized record, then out to the tools that need it
Getting started
First program live in six weeksEvery deployment follows the same four steps. The first program is a joint effort: we sit with the team, write the protocol down, connect the data and run the tests. After that, the engine is theirs.
From first conversation to a tool in production
Map the protocol, the users and the data sources. Agree on what success looks like.
Describe the tool, connect the data, assemble the blocks in a beta environment.
Virtual patients, simulated conversations, then sign-off by a named physician.
Go live for the team, with monitoring, versioning and a rollback path.
Where we are
Two applications run on the foundation today, including a pilot with a global pharmaceutical company, within established protocols where their impact can be measured. A connector to France's national patient record is under contract, we are backed by the Mistral AI and OVHcloud programs, and the same foundation is being prepared for the UK, the Gulf and the US.
Next: make the engine usable solo, in an afternoon, by a dietitian or a psychologist; track payer reimbursements and regulatory reporting, the first need of every care network we meet; and extend simulation until a tool can, under supervision, go beyond administrative work.
What tool has your team been waiting years for?
Describe it. We'll show you what Saana makes of it in 30 minutes.