Every hospital runs on an operating system. Most are never designed. They accumulate — from vendor defaults, inherited job descriptions, workarounds, and decisions made under construction deadlines. By opening day the operating system exists, but no one wrote it, no one owns it, and no one can change it without breaking something. Intelligent by Design writes it first.
This is the point most people get wrong on first contact. The operating system is the specification of how the hospital works — the decisions normally made implicitly, late, by default, or by a vendor. Four things are being designed.
What every entity, event, and measurement is called, and how those definitions relate to each other. This is the substrate everything else compiles against. Get it wrong and every system procured afterward inherits the error permanently.
Every process decomposed to task level, with each task assigned to a human, an agent, an automation, or a machine — the reasoning recorded, the assignment revisable as capability changes.
Who decides what, under which conditions, with what escalation, and what remains auditable afterward.
Headcount, floor area, adjacency, infrastructure load, and lifetime cost are outputs of the three above, not independent choices.
Software is then selected to satisfy the specification, rather than the specification being discovered after the software arrives. The same is true of the building.
The operating system is built through eight sequential planning blocks, each producing a defined, version-controlled artifact.
Services, volumes, case mix, acuity, payer and regulatory assumptions. The anchor.
The governed language and data foundation: nomenclature, master data, ontologies, knowledge graphs, and data definitions.
How work actually moves, before rooms and adjacencies exist.
Work decomposed into tasks and allocated. Humans lite, agents heavy.
The enterprise digital, data, interoperability and AI architecture — including the model, agent and evaluation stack, and the digital border-control and trust layer.
The operating model in written, trainable, auditable form.
FTEs derived from the work that remains after allocation, not from legacy ratios.
The operating model translated into space, adjacencies, infrastructure, and cost.
AI appears three times, doing three different jobs. Block 3 decides whether a task is done by an agent at all. Block 4 specifies what that agent runs on — models, orchestration, data access, evaluation, monitoring, containment. The governance layer sets the controls both must satisfy. Allocation, architecture, governance. None of the three substitutes for the others.
Iteration between blocks is expected. It is also governed — documented, approved, version-controlled.
By the time it begins, the building has largely been determined — because the work has been determined.
Design the building first and the organization spends the next forty years compensating for decisions it never made. Work changes before space changes. The operating system is the middle term between intelligence and the built environment, and it is the term almost no one starts with.
Because the operating system is a specification rather than an installation, it is versionable, licensable, and reusable across facilities of the same type. It is refined per deployment through controlled governance — never rebuilt from zero.
That property is a direct result of it not being software.