Every wing mapped.
Every second counts.

Why is a hospital harder to navigate in an emergency than almost any other building?

A hospital is rarely one building. It is an original structure from one era with additions bolted onto it across decades, connected by bridges, tunnels and corridors that were designed to solve a circulation problem rather than to make sense to a stranger. Each expansion arrived with its own numbering logic. Walk far enough down a single hallway and the room numbers change convention at a junction, because that junction is where one building ends and another begins.

Floor levels compound the problem. Two towers built thirty years apart rarely line up. The third floor of one is a half level off the third floor of the other, joined by a ramp or a short flight of stairs that does not appear on any signage. Staff who work the building every day absorb this and stop noticing it. Nobody who is new to the building has that advantage.

A lot of the people in a hospital are new to it: traveling clinicians, agency and per diem staff, rotating residents, vendors, contractors working on a renovation, volunteers. During a serious incident the number of people who have never walked the building grows sharply, because that is when mutual aid arrives. Officers and medics from neighboring jurisdictions come through a door they have never used, into a floor plate they have never seen, and are expected to move immediately.

Then there is the part that makes a hospital different from a school, an office or a hotel. Many of the people inside cannot leave on their own. A patient in an intensive care bed, a patient mid-procedure in an operating room, an infant in a neonatal unit: none of them are evacuating down a stairwell. Lockdown and evacuation in a hospital are decisions about which areas hold, which move horizontally to the next fire compartment, and what equipment and staff move with them. Those decisions need an accurate picture of the building, at speed, held in common by everyone making them.

What does a hospital digital twin actually contain?

Ark builds the twin from measured 3D capture of the real facility. LiDAR captures the interior. Drone capture handles roofs, exteriors and the campus around the buildings. The raw point clouds run through Ark’s proprietary processing into cloud meshes: an accurate as-built model of the hospital as it stands, not as the drawings said it would be built.

Two things Ark deliberately does not claim. There is no real-world latitude and longitude for interior rooms, and no interior geolocation. RTK lat/long is applied to the exterior, for grounds, approach routes and the outside of the buildings. Inside, wayfinding runs on the grid, true north and consistent labeling, which is how people actually talk to each other under stress. Nobody reads coordinates over a radio. They say a grid square, a room label and a stairwell number.

How does a live twin help during a workplace violence incident?

The core problem in the first minutes is not information. It is shared information. Hospital security knows the building. Arriving law enforcement does not. Clinical leadership knows where the patients are who cannot be moved. The incident commander outside knows what units are inbound. Each of them is working from a different mental model of the same place.

A live twin gives all of them one picture. The grid reference is the mechanism: security can say a location, an officer at the door can find that location in the model on a phone or laptop, and both are looking at the same geometry with the same labels. Routing works in both directions, toward a location for the people who need to reach it and away from it for the people who need to move patients and staff out of the path. Because the model is measured, it also carries what a floor plan cannot convey: sight lines down a corridor, where an approach is exposed, which doors open into which spaces, how a stairwell lands.

Cameras a hospital already runs can be surfaced inside the twin through API integration, so a live feed sits in its actual spatial position rather than as a numbered tile on a wall of monitors. Ark integrates with partner systems, including ZeroEyes gun-detection integration and Verkada hardware, so detection and video from those systems appear in the same operating picture as the building itself.

One clarification worth making plainly, because the categories are easily blurred: Ark does not sell panic alarms. Ark builds and maintains the facility intelligence layer. If your organization runs a duress or alerting program, that is a separate system.

Can the twin keep up with a hospital that is always under construction?

A hospital is never finished. A wing is renovated, a unit is converted, a nurse station moves, a door is sealed and a new corridor is opened. That is exactly how a set of drawings in a binder dies. It was accurate once. Nobody can tell you when it stopped being accurate, and nobody discovers the gap until the moment it matters.

Ark treats the twin as a living lifecycle model, not a one-time deliverable. Areas are recaptured and the model is updated as the building changes, over the life of the facility. Renovation work becomes a scheduled update rather than a reason the model quietly goes stale. That also makes the twin useful to the people running the construction, because the as-built geometry of what is actually there is the starting point for planning what comes next.

What about HIPAA and patient privacy?

Ark models the facility, not the people in it. The platform is built around facility intelligence: building layout, asset and infrastructure locations, and operational data about the physical plant. It does not store patient records and does not hold protected health information.

That distinction should be checked rather than taken on trust, and Ark does not claim a certification or third-party audit it has not stated it holds. Bring your compliance and privacy leads to the demo. The right conversation is a specific one about what is captured, where it lives, who has access to it and what your health system requires, before anything is scanned.

How does it connect to the systems we already run?

The twin is not intended to replace the stack a health system has already bought and trained people on. It is intended to give that stack a place to sit. API integration brings data from systems you already run into the model, so cameras, sensors and operational feeds are viewed in the context of the building rather than in isolation from it.

Delivery is through the browser. There is nothing to install, which matters in an environment where endpoint software goes through a review cycle measured in months, and where the people who most need the model during an incident are not on your network at all.

Access is role-based and works across a multi-campus health system. A security operations center, a facilities director, a clinical leader on one campus and a mutual aid partner do not need the same view or the same permissions, and they do not get them. Conventions stay identical across every site: the same grid, the same labeling, the same structure, so someone who can read one campus can read all of them.