Claude-style AI, for facility staffing
Call-offs become covered shifts — without losing control.
VanaHR is an AI staffing partner for senior care. Your coordinator asks in plain words; it reads live state, ranks who can actually cover, and proposes the next move. It proposes — a human confirms. Every consequential action is state-bound and auditable.
A scoped pilot on one building. Measured against your baseline.
Ana just called off for tonight’s 7p–7a on 3 North.
That opens a CNA gap at 19:00. Six staff are eligible and free — ranked by reliability, overtime risk and credential status.
Proposed
Send offers to the top 3 eligible CNAs
Nothing is sent until you confirm · read live at 18:42
You already know how this goes.
The call-off arrives at the worst time
Evenings, weekends, holidays — when the bench is thinnest and the coordinator has already gone home.
The scramble is manual
A phone tree down a printed list. Who’s eligible? Who’s already at 40 hours? Whose license lapsed? Nobody holds that in their head at 6pm.
Agency is the default fallback
When the scramble fails, the shift goes to agency — or goes uncovered. One is expensive; the other is a survey risk.
How it works
The recovery loop, governed end to end: propose → confirm → dispatch → trace.
- 1
Say what happened
“Ana called off for tonight’s 7p–7a on 3 North.” Plain words — no forms, no codes.
- 2
It reads live state
It re-reads the schedule, eligibility, credentials and overtime before it answers — and tells you when something changed underneath.
- 3
It proposes; you confirm
A ranked, reasoned recommendation arrives as a confirm card. Nothing leaves the building until a human says yes.
- 4
It dispatches and traces
On confirm, offers go out through the same governed handler your console uses — recorded once, exactly once, in an immutable ledger.
Built to keep you in control
The question isn’t whether AI can send a text. It’s whether you can explain, months later, exactly what it did and why. That’s the part we engineered first.
It proposes; a human confirms
The assistant can never take a consequential action on its own. Every write stops at a confirm card shown to an authorized person.
Bound to the state it read
A proposal is tied to the facts it was based on. If the world changes before you confirm, it invalidates rather than acting on a stale picture.
Exactly once, or not at all
Confirmed actions dispatch through the same handler your console uses, recorded once in an append-only ledger. No double-sends.
Auditable by construction
Who proposed, who confirmed, what state it was bound to, what happened — an immutable trail, not a log you have to take on faith.
PHI stays on the safe path
Health details are redacted before they could reach any external model, and a thread carrying them is blocked from that path entirely.
Role-scoped, tenant-isolated
Access is enforced in the database itself, per tenant and per role — not merely hidden in the interface.
What it’s worth
Three numbers decide whether this is real. We don’t publish invented averages — we measure these against your baseline, in your building, during the pilot.
Fill rate
How many open shifts end up covered by your own staff instead of agency — or not at all.
Measured against your baseline
Time to fill
How long a gap stays open, from call-off to a confirmed, accepted assignment.
Measured against your baseline
Agency spend
What you stop sending to agency once your own bench is reached first, and faster.
Measured against your baseline
Why there are no numbers on this page. We’re early, and we’d rather show you nothing than show you someone else’s marketing math. The pilot captures your current fill rate, time-to-fill and agency spend before anything changes — so the comparison is yours, not ours.
Who it’s for
Skilled nursing, LTACH and assisted living — from a single building to a multi-state group.
Staffing coordinators
Primary userThe person living the scramble. VanaHR is their surface: ask in plain words, get one ranked answer, confirm, move on.
DON / ADON
Approves what matters. Sees why a person was recommended — reliability, overtime, credentials — before an assignment is confirmed.
Administrators
Fill rate, time-to-fill and agency spend across the building, with an audit trail that survives a survey.
Staff (CNA / LPN / RN)
Offered only shifts they’re actually eligible for, call off in one sentence, and never spammed with shifts they can’t take.
Straight answers
Is our data safe?
Access is enforced in the database per tenant and per role, not just hidden in the interface. Health details are redacted before they could reach any external model, and a thread carrying them is blocked from that path entirely. We are not currently operating under a signed BAA — if that’s a requirement for you, say so in the pilot request and we’ll address it directly rather than quietly.
Does it replace my coordinators?
No — it removes the phone tree, not the person. The assistant cannot take a consequential action alone; every write waits for a human to confirm. Coordinators stop being a lookup engine and go back to judgment.
What about staff who don’t want texts?
Messaging is consent-gated per person and every attempt is recorded. Staff who haven’t consented are never contacted, and a “stop” is honored as an instruction, not a suggestion.
How fast can we pilot?
One building, with your existing roster and schedule imported from a spreadsheet. We capture your baseline first, so the result is measurable rather than anecdotal.
What if it recommends the wrong person?
You’ll see why it ranked them — reliability, overtime risk, credential status, eligibility — and you can decline. It never assigns anyone on its own, and a proposal bound to stale facts invalidates instead of acting.
Start with one building.
A scoped pilot, your baseline measured first, and a person on the other end of this form — not a sequence of automated follow-ups.
- One facility, your existing roster and schedule
- We capture fill rate, time-to-fill and agency spend before anything changes
- Nothing reaches your staff until your coordinator confirms it