The single most important architectural fact about aged-care monitoring is that a camera in a resident's room is not allowed to look at the resident. The duty of care doesn't grant the duty of surveillance. The hard problem is producing a useful safety signal without producing a watching signal, and corridor occupancy, counted anonymously, is one of the most useful safety signals in the building.
We don't have an aged-care pilot running yet. This post is about how we'd configure one.
01 · WHY CORRIDORS ARE THE RIGHT UNITWhy corridors are the right unit
Falls happen in the corridor between the bedroom and the bathroom, almost always at night, almost always to a resident moving alone in the dark. The clinical literature is consistent: the four hours between midnight and 4am are the highest-risk window in the facility. They are also the window when a resident is least likely to call for help: the cord is back at the bed, the device is on the nightstand, the fall is in the corridor.
An anonymous occupancy count in the corridor, a single dot moving down a defined zone, is a clinically meaningful signal:
- A dot that stops moving for longer than expected is a candidate wellness check.
- A dot at 3am from a wing that hasn't had one in three months is a candidate review.
- A dot count higher than one at a designated single-occupancy point is a candidate procedural change: two residents converging in a corridor at night is a tripping geometry.
None of those signals require knowing who the dot is. The platform doesn't, because the architecture forbids it.
02 · WHAT A DEPLOYMENT WOULD LOOK LIKEWhat a deployment would look like
A single edge device per corridor segment. The model runs on-chip, emits a dot: anonymous count, anonymous position within the zone, no identity, no demographic, no frame stored. Thresholds and stationary-dwell rules generate alerts; alerts route to the care team's device.
What the deployment would not include, and is incapable of:
- No cameras in resident rooms. We won't deploy any.
- No face matching anywhere on the pipeline. The model can't.
- No identity inference from posture, gait, or schedule patterns. We don't ship those models.
The clinical team would get a corridor density and a stationary-dwell alert. They never get a resident's face or name from the platform, because the platform genuinely doesn't see one to give.
03 · WHAT FAMILIES WOULD ASK, AND WHAT WE'D ANSWERWhat families would ask, and what we'd answer
Three questions come up in every family-engagement session:
- "Is it watching my mother?" No. It counts movement in the corridor. No model of her, no record of her, no identifier of any kind.
- "What if it gets hacked?" There is nothing identifying to leak. A breach would expose corridor counts, the data already on the care team's dashboard.
- "What if it misses something?" It is a probabilistic early-warning, not a replacement for the call-bell, the nurse round or the buddy check. The honest framing is the only one worth giving.
04 · WE'RE LOOKING FOR AN AGED-CARE PILOT PARTNERWe're looking for an aged-care pilot partner
We don't have one yet. We're looking for a single pilot partner in 2026: a facility willing to scope corridor occupancy as a clinical-safety supplement, not a surveillance layer. The architecture solves the objection that every other vendor in the category has to argue around. Talk to us about a pilot or see how privacy works.