Skip to content
Barion AI

Products / Clinical

In development

ASSISTIVE INTELLIGENCEFOR CLINICALDECISIONS.

Designed to support qualified clinicians, never to replace them. Treatment options compared against real patient context, with the reasoning and the risks shown, and the decision left where it belongs.
Domain
Clinical decision support
Status
In development
Available for care
No

The problem

Hard cases are hard because the context is scattered

Not because clinicians lack knowledge. Because the information needed to compare options sensibly is spread across systems, notes and memory, and the comparison itself is never written down.
Complete context
The relevant history exists, but assembling it for one decision is manual work done under time pressure.
Option comparison
Several treatment paths may be defensible. The comparison between them usually happens in someone’s head and leaves no record.
Risk identification
Interactions and contraindications are knowable, but only if every relevant item of context was actually in view.
Transparent reasoning
A recommendation without visible reasoning cannot be challenged, taught from, or reviewed later.
Human judgement
The parts that need clinical judgement should stay with the clinician. Support means removing the assembly work, not the decision.
Traceability
Months later, someone may need to see what was known and what was considered at the time.

Intended workflow

How it is designed to run

This is the intended shape of the system, not a description of software you can use today.
  1. 01

    Patient context

    History, current presentation and relevant prior treatment, assembled rather than searched for across systems.

  2. 02

    Relevant medical information

    Retrieval narrowed to what bears on this case, with the source of each item kept attached to it.

  3. 03

    Option generation

    Treatment paths that are actually defensible for this patient, not a generic list.

  4. 04Gate

    Contraindication and risk checks

    Interactions, comorbidities and known risks tested against the specific context.

  5. 05

    Ranked considerations

    Options ordered with the trade-offs made explicit, including what would change the order.

  6. 06

    Explanation

    Why each option ranked where it did, in language a clinician can interrogate and challenge.

  7. 07Human

    Clinician decision

    The clinician chooses. The system records what was shown, what was chosen and when.

Conceptual workflow. Not a released product.

Human control

Control is the design, not a setting

In a clinical context, the control model is the product. These are the commitments the system is being built around.
Layered context, held in one frame rather than searched across systems.
01The decision stays with the clinician
The system ranks and explains. It does not select a treatment, and it has no route to act on a patient record on its own.
02Every recommendation is explainable
A ranking that cannot be explained is a defect, not an output. If the reasoning cannot be shown, the option should not be presented.
03Uncertainty is surfaced, not smoothed
Where the context is incomplete or the evidence is thin, that is stated on the option rather than hidden behind a confident number.
04Actions require defined authority
Anything that writes to a clinical record or reaches a patient sits behind an approval by a person with the authority to give it.
05History is available for review
What the system received, what it proposed and who decided, retained so a case can be reviewed long after the appointment.
06Escalation is a designed path
When context is missing or a contraindication is detected, the intended behaviour is to stop and escalate, not to proceed on a best guess.

The same control mechanisms are used across every Barion product. How guardrails and observability are engineered.

Intended users

Who this is being designed with

Clinicians and institutions who own difficult decisions and are accountable for them.
  • Primary care physicians
  • Specialists working complex cases
  • Clinical decision boards
  • Healthcare institutions

Intended audiences. No deployment has taken place.

Status

Where development has actually reached

Stated plainly, because overstating clinical readiness would be both wrong and dangerous.
Designed
The decision model, the intended workflow and the control requirements above.
Being built
The reasoning, ranking and explanation layers on Barion Core, which are shared with products already running.
Not started
Clinical validation, regulatory assessment, and any deployment into a care setting.
Open now
Design partnership with clinicians and institutions willing to shape the comparison and the guardrails against real practice.

Continue

Where to go next

Discuss a clinical AI partnership

We are looking for clinicians and institutions who want to shape this against real practice. Design partnership, not deployment.

Discuss a clinical AI partnership All products