Put Seven Extra Hands on Your Team, Not Seven New Risks
Named AI personas draft the routine work and flag what needs a human, each at a defined autonomy level, and none can release a result on its own.
Salus is AI-native, but not AI-in-charge. The intelligence is embodied as a team of named personas, each responsible for one part of the lab, each with a least-privilege permission scope, and each supervised by a human.
Personas are real workers, not a chatbot bolted on. They draft, flag, extract, and monitor. They never release a result on their own. A floating assistant orb on every screen lets anyone summon the persona that owns the screen they are on.
What it does.
The full capability set, not a highlight reel.
Drafts the routine work
Accessioning from manifests, operational reports, and verifications are drafted for a person to approve in one click, never auto-committed.
Flags what needs a human
Out-of-range results, presumptive positives, QC drift, and turnaround-time risk are surfaced first, so attention goes where it matters.
Assists ELR and coding
Validates HL7 and FHIR messages and LOINC/SNOMED coding before anything reaches the state feed.
Watches for signals
Monitors test patterns for anomalies and possible outbreak signals, and drafts alerts for epidemiologist review.
The assistant orb
A floating orb on every screen surfaces the persona that owns that screen, with a proactive tip, a live operational pulse, and a one-click action.
Meet the personas.
Seven named workers, each owning one part of the lab. Each has an autonomy level, and a human in control.
Extracts manifests, label sheets, and CSVs into the Suspense Queue and pre-fills accessioning.
Flags anomalous and out-of-range results and drafts verifications for human sign-off.
Monitors Westgard and out-of-spec events and keeps CAP/CLIA documentation current.
Validates HL7/FHIR and LOINC/SNOMED coding before the state feed.
Tracks turnaround time, bottlenecks, and staffing, and drafts reagent reorders.
Detects outbreak signals and flags cases for investigation.
Supervises the worker personas, and can pause, throttle, or down-scope any of them.
The full architecture.
For your technical reviewers: the complete AI agent ecosystem, the human-in-the-loop pipeline, and the governance model behind it.
What it demonstrates
Why this reassures a cautious reviewer.
- Seven named AI personas, each scoped to one part of the lab, none able to release a result alone
- A human-in-the-loop pipeline: every AI action is drafted for a person to approve
- A governance layer with bias, performance, and audit monitoring built in
- Explainability by design, so every recommendation carries its rationale
Who is in control.
- Each persona has an autonomy level: Suggest, Acts with approval, or Autonomous in scope.
- Cher, the supervisor, can neuter a worker, pause it or lower its autonomy and scope, but can never fully turn one off.
- Disabling a persona is a human-only action. The kill switch always stays in a person's hands.
- Every status, autonomy, and scope change is written to an append-only control log: who changed what, and when.
- Personas hold scoped permissions that mirror staff roles, so a persona can never do something its human counterpart could not.
Configuration.
What your lab controls, without a vendor in the loop.
- Set each persona Active, Paused, or Disabled from the AI Personas screen
- Choose an autonomy level per persona (Suggest / Acts with approval / Autonomous in scope)
- Grant or revoke a persona's permission scope, lab by lab
- Review the full control history for any persona
See it in your laboratory's context.
Explore the live demo, no login and no request form.