What fails in hospitals begins with how hospitals are designed. This is a manifesto for designing the operating system first — and the building second.
We believe that what fails in hospitals begins with how hospitals are designed.
Hospitals are among the most complex human systems ever created. Thousands of people, technologies and machines perform tens of thousands of interconnected tasks every day. Decisions create actions. Actions create information. Information moves between people and systems. Every process depends upon another.
And yet we have never truly designed the hospital as a complete system.
We design the building.
We plan the structure, the rooms, the corridors, the utilities, the equipment and the adjacencies. Years of professional expertise and enormous amounts of capital go into determining what the hospital will physically become. But the operating system that determines how the hospital actually works is rarely designed with the same discipline.
Instead, it evolves.
It is assembled over years from a thousand thumb drives: policies inherited from other hospitals, procedures brought by incoming leaders, regulatory requirements, accreditation standards, vendor workflows, legacy systems, consultants, previous employers, departmental conventions and local workarounds. Eventually, this accumulation becomes the way we do things here.
It may be documented. It may be governed. It may be accredited.
But documentation is not design.
And we believe this distinction explains far more about hospital performance than we have been willing to acknowledge. The inefficiencies, unnecessary costs, workforce pressures, process variation, fragmented information, interoperability failures, safety problems and repeated transformation failures we see across healthcare are not simply problems to be managed individually. They are symptoms of something deeper.
The system itself was never designed as a system.
At its most fundamental level, a hospital is humans and machines performing tasks within processes. Jobs are bundles of those tasks. Departments organize them. Processes connect them. Decisions direct them. And data is the digital footprint of a process.
If the work itself was never coherently designed, we should not be surprised when processes vary. If concepts were never consistently defined, we should not be surprised when systems cannot communicate. If workflows emerged rather than being engineered, we should not be surprised when staff create workarounds to make them function.
For decades, humans compensated for this. Experienced nurses knew which procedure did not quite reflect reality. Physicians understood the exceptions. Managers knew whom to call. Staff carried enormous amounts of unwritten organizational knowledge in their heads and bridged the gaps between what the hospital said should happen and what actually had to happen.
That world is ending.
Three forces are ending it.
Artificial intelligence changes the equation. AI cannot safely operate on organizational ambiguity at scale. It cannot reliably automate a process that the organization itself cannot define. It cannot interpret five conflicting meanings of the same operational concept and somehow know which one we intended. And it cannot safely inherit decades of accumulated process variation simply because we digitized it.
We cannot automate what we have not defined.
We cannot scale what we have not standardized.
We cannot intelligently automate a hospital that was never intelligently designed.
Cybersecurity makes the same problem impossible to ignore from another direction. When tasks, identities, systems, permissions, information flows, decision rights and dependencies are poorly understood, we do not merely create inefficiency. We create vulnerability.
And the workforce will not scale. Demand for care is rising faster than the supply of people trained to deliver it. No hospital can hire its way out of this, and no country can train its way out of it quickly enough to matter. For a generation, healthcare's answer to rising demand has been more people. That answer is closing.
Which means tasks must change hands. But tasks cannot be reassigned — to machines, to agents, to different professions, to patients themselves — unless we can first say precisely what those tasks are. Redistribution requires definition. A hospital that cannot describe its own work cannot decide which parts of that work a human must keep.
None of these three forces created the underlying problem.
They exposed it.
They are forcing us to confront something healthcare should perhaps have confronted decades ago: before we design the spaces, we must understand and design the system that will inhabit them.
Systems before spaces.
We are asking the building to absorb a decision we never made.
Much of what we call flexibility is a hedge against a specification we never wrote. When the operating model — how work is done, who does it, with what staffing and what systems — is not designed before the building is, flexibility becomes the insurance policy against that omission. And we pay for it in capital.
This is not an argument against adaptable buildings. It is an argument about sequence. What actually prevents a hospital from being reconfigured later is rarely structure. It is licensing, staffing and system configuration. Buying structural flexibility to solve an operating problem is expensive, and it does not solve it.
The organizational chart tells us who reports to whom. It was built for an era in which work was performed almost entirely by humans grouped into jobs, departments and hierarchies. But jobs are bundles of tasks, and those bundles are coming apart. Some tasks will remain exclusively human. Some will move to machines. Some will be executed by AI under human supervision. Some will remain human-led but become unrecognizably augmented.
A hospital cannot be designed around an organizational chart that is being disassembled. It must be designed around the work itself: what has to happen, why, what information it requires, what decisions it creates, who or what should perform it, how it connects to everything around it, and how we know it was performed correctly. The workforce, the technology, the digital architecture and the physical environment all follow from that.
This is not another digital transformation. It is not another process-improvement initiative. It is not about adding AI to the hospital we already have. And it is certainly not about making yesterday's hospital faster.
We should not automate the past.
We have an opportunity to reconsider the hospital from first principles — to intentionally design its operating system, its language, its work, its intelligence and its physical form as parts of one coherent whole.
Most hospitals, of course, cannot start over. Very few of us are handed an empty site and a clean sheet. But this discipline does not require a new building. It requires knowing how the hospital is supposed to work — and an existing hospital can be examined, defined and redesigned as a system without pouring a single new foundation. The difference is only where you begin. A new hospital can be designed correctly from the start. An existing hospital must first be understood. Both are design problems. Only one of them begins with concrete.
Either way, the work begins by questioning assumptions that have become so familiar we no longer recognize them as assumptions.
Why does this job exist?
Why is this task performed by a human?
Why is this information collected?
Why do two systems describe the same thing differently?
Why does this room exist?
Why are we designing the building before we have designed what happens inside it?
The answer cannot continue to be:
Because that is how hospitals work.
We are entering territory where no organization, profession, technology company or discipline has all the answers.
Neither do we.
We are all learning together. We will learn by questioning. By designing. By simulating. By testing. By measuring. By building. By failing intelligently. By improving what works and discarding what does not.
We will learn by doing.
And what we learn should not disappear into another consultant's report, another policy manual or another thumb drive. It should become part of a living hospital operating system — one that can continuously learn, improve and be adapted across facilities, health systems and countries.
The ambition is not a smarter building. It is not a paperless hospital. It is not a hospital with more technology.
It is a hospital that knows how it is supposed to work. A hospital in which humans and machines operate as parts of one intentionally designed system. A hospital whose physical structure follows its operating requirements rather than defining them. A hospital capable of safely absorbing technologies that have not yet been invented.
A hospital that can learn. A hospital that can adapt. A hospital designed for intelligence from the beginning.
So we are asking for something specific.
Before the next hospital is drawn, define how it is meant to work. Before the next system is purchased, define the work it is meant to carry. Before the next transformation program is approved, ask whether it is redesigning the hospital or merely accelerating it.
And if you sit on a board, a ministry, an investment committee or a design review: refuse the drawings until someone can tell you what happens inside them.
That is the demand. Design the system first.
We believe the next generation of hospitals should not simply be built.
They should be designed.
Intelligently.