The Coherence Problem: Why NHS Transformation Always Fails

A laser and a lightbulb both emit photons. They use the same fundamental physics. The difference is that in a laser, every photon is in phase with every other photon. The waves reinforce each other. The result is a beam that can cut steel, measure the distance to the moon, or transmit data across continents without loss.

A lightbulb scatters its photons in every direction, at every frequency, out of phase with each other. The energy is real. It is also, in any engineering sense, wasted. A lightbulb cannot cut anything. It cannot transmit a signal. It can only radiate heat and a diffuse glow.

The difference between these two devices is coherence. Not power. Not energy. Not intent. Coherence.

This is not an abstract distinction. It describes a structural problem facing healthcare systems worldwide, wherever care crosses organisational boundaries. But it is the National Health Service, with its 1.5 million employees, its £180 billion annual spend, and its uniquely fragmented organisational landscape, that illustrates the problem most vividly.

The NHS contains some of the most skilled clinicians in the world, operating in some of the most sophisticated facilities ever built. And yet it routinely fails to transfer a patient’s medication list from one building to another. It loses responsibility for patients at organisational boundaries the way a lightbulb loses photons: in every direction, continuously, as a fundamental property of how the system operates.

This is not a technology problem. It is not a funding problem. It is not even, primarily, a management problem. It is a coherence problem. And coherence, as physics has understood for over a century, is not something you can bolt on afterwards. It is a property of infrastructure, or it is absent.

What coherence actually is

Let’s be precise, because this isn’t metaphor. In physics, coherence describes the degree to which the components of a system maintain stable, predictable phase relationships with each other. When waves are coherent, they can interfere constructively. Energy accumulates. Signals propagate. Work gets done. When waves are incoherent, they interfere destructively as often as constructively. Energy dissipates. Signals degrade. The system produces noise instead of signal.

This matters beyond optics. In 1944, Erwin Schrödinger published What Is Life? ^[1], a slim book that would go on to inspire Watson and Crick’s discovery of DNA. Schrödinger posed a deceptively simple question: how do living systems maintain their extraordinary internal order in a universe governed by the second law of thermodynamics, which dictates that all systems tend toward disorder?

His answer was what he called “negative entropy,” or negentropy. Living systems, Schrödinger argued, survive by continuously extracting order from their environment. A cell doesn’t achieve order and then coast. It actively maintains order through metabolism, through constant energetic work against the tide of entropy. Stop the metabolism and the cell doesn’t slowly wind down. It dies. The order collapses. ^[1]

This is the foundational insight that the NHS has never absorbed: coherence is not a state you achieve. It is a process you must continuously run. It requires infrastructure that actively maintains order against the natural, inevitable, thermodynamically guaranteed tendency of complex systems to fragment.

The entropy of care

In 2024, Herman Aksom published a paper titled “Organizational Disintegration” ^[5] in which he articulated something that anyone who has worked inside a large institution already knows intuitively: organisations have an inner tendency to gradually drift from ordered and patterned design toward fragmentation, disorganisation, and disorder. This disintegration, Aksom argues, is a by-product of increasing organisational complexity. As organisations grow more complex to adapt to their environment, they inevitably generate disorder processes: loss of control, loss of coordination, the emergence of tensions, inconsistencies and contradictions within their own structures and functions. ^[5]

This is not a failing of specific organisations. It is, in the language of thermodynamics, a law. Without active countermeasures, without the organisational equivalent of metabolism, entropy wins. Every time. The only variable is speed.

Now consider the NHS. It is not one organisation but hundreds, each with its own governance, its own IT systems, its own institutional culture, its own understanding of what words like “referral” and “discharge” actually mean in practice. A patient moving from a GP surgery to a hospital to a community mental health team crosses at least three organisational boundaries. At each boundary, information degrades. Context is lost. Responsibility becomes ambiguous. The signal attenuates.

The evidence for this is not anecdotal. It is industrial in scale. England experiences an estimated 1.8 million undetected medication errors at care transitions every year, resulting in harm across approximately 31,600 patient episodes ^[29]. More than half of these harmful errors occur at the point of hospital admission, which is to say, at the exact boundary where responsibility transfers from one organisation to another ^[29]. The British Medical Association estimates that 13.5 million working hours are lost by doctors annually due to inadequate IT systems ^[30]. These are not glitches in an otherwise functional system. They are the predictable, measurable output of organisational incoherence.

This is what it looks like at the boundary. A GP refers an elderly patient on six medications into an acute trust. The referral arrives — sometimes by fax, sometimes by e-RS, sometimes by letter scanned to PDF — without a reconciled medication list, without the context of why the GP chose those specific drugs in that specific combination, without a clear statement of what the GP is asking the hospital to do. The admitting junior doctor starts from scratch, re-clerking the patient, phoning the GP surgery that is now closed, making best guesses from the patient’s own uncertain recall. A beta-blocker is missed. The patient deteriorates on day two. Nobody failed. Nobody was negligent. The information existed. The responsibility existed. But at the boundary between two organisations, both degraded to noise.

Researchers at Radboud University Medical Centre in the Netherlands applied Shannon’s information entropy theory directly to hospital organisational structures ^[6]. They identified two critical forms of entropy in healthcare systems: positional entropy, which measures the uncertainty about where a patient is in their journey through the system, and task allocation entropy, which measures the uncertainty about who is responsible for what. Both forms of entropy increase at organisational boundaries. Both are directly correlated with delays in clinical decision-making. ^{[6] [7]}

These researchers put a formal name on something clinicians have always known: the system doesn’t know where the patient is, and it doesn’t know who is looking after them. Not because nobody cares. Because the governance infrastructure for maintaining that knowledge across organisational boundaries does not exist.

The interoperability illusion

At this point, the standard response from the health technology sector is: interoperability. Connect the systems. Adopt FHIR. Build shared care records. And there is truth in this. Interoperability is necessary. But it is not sufficient, and the confusion between the two is actively dangerous because it creates the illusion that the coherence problem is being addressed when it is not.

To understand why, go back to the physics. Signal integrity and signal coherence are different things. Signal integrity asks: can the data travel from point A to point B without corruption? This is what interoperability standards like FHIR address. They define a common language, a shared data format, a way for System A to send a message that System B can parse.

Signal coherence asks a harder question: when the signal arrives, does it mean the same thing at both ends? Does it carry the same implications for action? Does responsibility travel with the data, or does the data arrive orphaned, technically accessible but operationally inert?

You can connect two systems that are working at cross purposes. All you’ve created is faster incoherence. A GP can send a perfectly FHIR-compliant referral to a hospital trust, and if there is no shared understanding of what that referral implies about the transfer of clinical responsibility, about who holds duty of care during the transition, about what constitutes adequate acknowledgement, then the data has arrived but the coherence has not.

This is why, despite two decades of interoperability initiatives, the NHS’s principal challenge remains what a 2020 paper in the Journal of the Royal Society of Medicine identified as the fragmentation of responsibility for IT infrastructure ^[8]. Not the fragmentation of technology. The fragmentation of responsibility. Multiple vendors create proprietary electronic medical records tailored for local trusts with little thought to external data sharing ^[8]. The technology connects. The organisations do not cohere.

The transformation fallacy

This matters right now because the NHS is, once again, about to attempt transformation from the top down. The Ten Year Health Plan, published in July 2025, positions technology as the central enabler of three “fundamental shifts”: analogue to digital, hospital to community, and sickness to prevention ^[12]. It promises to make the NHS “the most AI-enabled healthcare system in the world.” It bets heavily on a Single Patient Record, a massively expanded NHS App, genomic sequencing, wearables, and robotics. ^[12]

The ambition is laudable. The sequencing is backwards. And this is not a uniquely NHS problem. It is a universal principle of transformation that healthcare systems worldwide keep learning and forgetting: you cannot transform from the top of a pyramid that has no base.

McKinsey’s research consistently shows that 70% of large-scale digital transformations fail to meet their objectives ^[14]. In healthcare specifically, 75% of health system executives surveyed globally report that their organisations cannot deliver on their digital transformation ambitions because they have not built the necessary foundations ^[13]. This is not a statistic about technology. It is a statistic about sequencing. Organisations try to build the visible, impressive upper storeys of transformation on foundations that do not exist.

The NHS has its own proof case: the National Programme for IT, launched in 2002 with a budget of £6.2 billion and the backing of the Prime Minister ^[10]. NPfIT was the largest civilian IT programme ever attempted in the United Kingdom. It aimed to deliver integrated electronic patient records and secure data exchange throughout the NHS by 2010. It was dismantled in 2011, having cost the taxpayer over £12.7 billion while delivering approximately £2.6 billion in actual benefits. ^[10]

Every post-mortem reaches the same conclusion. The programme rushed to award contracts “in almost indecent haste with insufficient planning” ^[10]. It discussed policy changes, procurement strategies, and implementation steps “without stopping to make sure the foundation was set” ^[10]. It was, in the words of the academic literature, “marred by resistance due to the inappropriateness of a centralized authority making top-down decisions on behalf of local organizations” ^[9]. It attempted to impose coherence from above without building the infrastructure that would generate coherence from within.

The parallels with the current moment are uncomfortable. The Ten Year Plan talks about AI scribes, but the King’s Fund points out that “AI scribes can only transform the productivity of the NHS if staff don’t need to spend 30 minutes every morning logging into multiple out-of-date IT systems” ^[16]. It promises a Single Patient Record, while nine NHS trusts still lack any electronic patient record at all ^[11]. It envisions genomic screening at scale, while 45% of NHS services lack even a basic digital pathway ^[11]. The Darzi Review, which set the stage for the plan, found that the NHS “remains in the foothills of digital transformation” and that technology “always seems to add to the workload of clinicians rather than releasing more time to care.” ^[11]

The plan, as Digital Health’s analysis observed, is “extremely long on aspiration but worryingly light on delivery plans” and “overlooks the basics of infrastructure and reliability” ^[15]. The King’s Fund concluded that “the key foundations for good implementation of technology, such as infrastructure, shifting behaviours, and changing workflows and processes, remain largely ignored.” ^[16]

This is the transformation fallacy: the belief that you can deliver the future without first fixing the present. That you can build AI-enabled care on infrastructure that cannot reliably transfer a medication list. That you can create a Single Patient Record in a system where responsibility for that patient fragments every time they cross an organisational boundary.

The countries that have actually achieved what the NHS aspires to did not make this mistake. Estonia, a country of 1.3 million people, now has 99% of its health data digitised, with physicians accessing complete patient histories at the point of care ^[27]. But Estonia did not start with applications. It started with X-Road, a secure, decentralised data-exchange platform that provides the governance layer through which all systems communicate ^[27]. It started with digital identity infrastructure. It built the connective tissue first, then the organs.

Denmark mandated that all healthcare providers use standardised messaging formats, backed by law and financial penalties for non-compliance ^[26]. It built a national eHealth competence centre, MedCom, to govern the standards and infrastructure components before any integrated services were deployed on top ^[26]. Within a decade, it had achieved comprehensive patient data sharing across hospitals, GPs, and pharmacies.

Both countries built foundations before buildings. Both enforced governance infrastructure through regulation rather than relying on voluntary adoption. Both understood, intuitively if not in the language of physics, that coherence must be infrastructural or it will not exist at all. ^[18]

The NHS faces greater complexity due to its size, its multi-vendor landscape, and its deeply embedded organisational fragmentation. But the principle does not scale differently. If anything, it applies more forcefully: the larger and more complex the system, the more critical the governance infrastructure becomes, because there are more boundaries at which coherence can degrade.

This is not an argument against the Ten Year Plan’s goals. Shifting care from hospitals to communities is essential. Harnessing AI and data for better outcomes is inevitable. Prevention over treatment is obviously right. But none of these shifts can be delivered on top of a system that cannot reliably coordinate care across organisational boundaries today. You cannot shift care to the community if the governance infrastructure for tracking responsibility, consent, and clinical intent across community providers does not exist. You cannot deploy AI at system scale if the data it needs is fragmented across incompatible systems with no coherent provenance trail. You cannot prevent disease through population health analytics if the population’s health data is trapped in organisational silos that were never designed to talk to each other.

The transformation the NHS needs is real. But transformation, in any domain, follows a universal pattern. You build the foundations. You establish the governance. You create the infrastructure that will maintain coherence under load. Then, and only then, you build the services on top. Reverse the sequence and you get NPfIT: £12.7 billion spent, £2.6 billion delivered, and a generation of institutional scar tissue that made the next attempt harder. ^[10]

What transformation practitioners already know

I have spent thirty years delivering digital transformation across government, defence, financial services, and healthcare: at GCHQ, British Airways, RBS, Lloyds Bank, the Home Office, and others. Not as a policy adviser. As the engineer in the room when things either worked or didn’t. And the pattern described above is not theoretical to anyone who has actually done this work. It is the lived experience of every serious transformation practitioner I have ever worked with.

There is a sequence to transformation. It is not arbitrary. It reflects a structural dependency chain in which each layer requires the one beneath it to exist before it can function. Violate the sequence and the programme fails, regardless of budget, ambition, or political will. The sequence looks like this:

First, you establish a safe operating environment. In cloud engineering, this is called a landing zone: a governed, secured, baseline infrastructure in which work can happen without immediately generating risk ^[21]. AWS’s Well-Architected Framework makes this explicit, stating that “before you architect any workload, you need to put in place practices that influence security” ^[22]. The landing zone includes identity management, access controls, logging, encryption, and network isolation, all established before a single application is deployed. The principle is not controversial in engineering. You do not build the house before you lay the foundations. You do not deploy workloads into an environment that cannot safely contain them.

Second, you establish a shared language. Eric Evans, in his seminal work on Domain-Driven Design, calls this the “ubiquitous language”: a common, rigorous vocabulary shared between technical teams and domain experts ^[19]. Evans argues that without this shared language, the friction introduced by constant translation between different vocabularies undermines every subsequent effort. As Martin Fowler summarises it, the language must be “based on the domain model used in the software, hence the need for it to be rigorous, since software doesn’t cope well with ambiguity” ^[20]. In healthcare, the absence of this shared language is acute. The word “referral” means different things to different organisations. “Discharge” carries different implications depending on context. “Responsibility” is used loosely precisely where it needs to be used precisely. Without a grammar for the system, every message crossing an organisational boundary is subject to interpretation, and interpretation is where coherence degrades.

Third, you establish coherence at the infrastructure layer, not the application layer. This is the insight that platform engineering has formalised. Spotify calls it the “golden path.” Netflix calls it the “paved road.” The principle is identical: you embed governance, security, observability, and compliance into the infrastructure itself, so that every team building on top of it inherits those properties automatically. Coherence becomes a property of the platform, not a responsibility delegated to individual teams who may or may not implement it correctly.

Fourth, you create order through small, verifiable steps. John Kotter’s research on organisational change consistently demonstrates that breaking transformation into incremental stages, building confidence and competence at each level before advancing to the next, is not a nice-to-have. It is a prerequisite ^[25]. The MIT CISR Enterprise AI Maturity Model, based on a survey of 721 companies, found that financial performance improved at each stage of maturity, and that the capabilities required at each stage must be built sequentially ^[24]. You cannot skip stages. The Infosys Enterprise AI Maturity Model states this even more bluntly: “Maturity is not about speed. It is about readiness.” ^[24]

Fifth, you implement first principles and validate them before scaling. You prove the pattern works in a constrained environment before extending it. Then, and only then, you scale those principles across the system. And finally, once the foundations are proven and scaled, you build out the capabilities and features that everyone wanted to start with.

This is not an idiosyncratic methodology. It is the consensus of every major cloud provider, every serious enterprise architecture framework, and every practitioner who has survived enough failed programmes to understand why they failed. AWS builds landing zones before workloads ^[21]. Google Cloud’s transformation framework positions foundational infrastructure as a prerequisite to service delivery. The TOGAF enterprise architecture standard sequences capability development from baseline through to optimised. Scott Hanselman, borrowing from Maslow, captured the practitioner’s frustration perfectly: a team worried about code aesthetics when their deployment process is manual and their builds are unverifiable is asking whether they are eating enough leafy greens without first establishing whether they have food for tonight. ^[23]

The relevance to healthcare should now be obvious. The NHS’s repeated transformation failures are not mysterious. They are the predictable result of attempting to start at the top of this sequence rather than the bottom. The Ten Year Plan wants AI, genomics, wearables, robotics, and a Single Patient Record ^[12]. These are all top-of-stack capabilities. They all depend on governance infrastructure that does not exist. They all require a safe operating environment for clinical data that has not been established at the inter-organisational layer. They all assume a shared language for responsibility transfer that no one has defined.

Every transformation practitioner seeing this recognises the pattern immediately. It is the programme that buys the software before establishing the data governance. The migration that deploys applications before building the landing zone. The digital strategy that promises AI before cleaning the data. The NHS is not failing at transformation because it lacks ambition or talent. It is failing because it keeps starting in the wrong place.

Requisite variety, or the lack of it

In 1956, the cybernetician W. Ross Ashby formulated what he called the Law of Requisite Variety ^[2]. The law states, in essence, that any system attempting to regulate another system must possess at least as much variety in its responses as exists in the system it is trying to regulate. A thermostat with only “on” and “off” can regulate room temperature adequately because the problem space is simple. But if the environment were to vary across a hundred dimensions simultaneously, the thermostat would be useless.

Stafford Beer, the British management cybernetician who built on Ashby’s work, restated this more bluntly: only variety can absorb variety. ^[3]

Healthcare delivery operates in a domain of effectively infinite variety. Every patient is different. Every clinical presentation is different. Every combination of comorbidities, social circumstances, patient preferences and institutional constraints creates a unique problem space. The governance infrastructure that is supposed to coordinate care across organisational boundaries has, by comparison, almost no variety at all. It consists of referral letters, discharge summaries, and fax machines. In some places, literally fax machines.

The mismatch is not just large. It is categorical. The variety of the clinical reality overwhelms the variety of the coordination infrastructure. In Ashby’s terms, the regulator cannot regulate ^[2]. In Beer’s terms, the system is not viable. ^[3]

Beer’s Viable System Model offers another critical insight. He defined what he called the meta-system: “a collection of sub-systems which looks after the operational elements so that they cohere in that totality called the Viable System” ^[3]. The meta-system doesn’t replace the operational units. It doesn’t centralise control. Beer was emphatic that operational units must be as autonomous as possible. The meta-system exists to provide the connective tissue, the coordination, the shared context that allows autonomous parts to function as a coherent whole. ^{[3] [4]}

The NHS has no functioning meta-system for inter-organisational care. Individual trusts, GP practices, community services, and mental health providers are autonomous operational units. That part of Beer’s model is well represented. But the meta-systemic layer that should maintain coherence between them is either absent or so weak as to be ineffective. There is no infrastructure that continuously maintains the phase relationships between organisations: who is responsible for this patient right now, what is the clinical intent of this transfer, has consent been established for this data to travel, what outcome is expected, and who will know if it isn’t achieved.

POSIWID

Beer coined another phrase that cuts to the heart of this: POSIWID, or “the purpose of a system is what it does” ^[4]. Not what it says it does. Not what it was designed to do. Not what well-meaning people within it want it to do. What it actually, observably, measurably does.

The NHS’s stated purpose is to provide integrated, safe, patient-centred care. Its actual output, measured at organisational boundaries, is fragmented responsibility, degraded information, ambiguous accountability, and 1.8 million medication errors per year at care transitions ^[29].

By POSIWID, the system behaves as if organisational survival is the dominant regulating force, not care coherence. Each trust optimises for its own continuity, its own targets, its own CQC rating. This is not malice. It is the natural, entropic behaviour of autonomous organisations without a coherence-maintaining meta-system ^[5]. Without infrastructure that aligns the phase relationships between organisations, each organisation’s rational self-interest produces collective incoherence. The lightbulb doesn’t choose to scatter its photons. It has no mechanism not to.

What coherence infrastructure looks like

If the diagnosis is correct, then the prescription follows. The NHS does not need more energy. It does not need another reorganisation. It does not, primarily, need more money. It needs coherence infrastructure: active, continuous mechanisms that maintain the phase relationships between autonomous organisations as care crosses boundaries.

What are those phase relationships? There are, fundamentally, seven flows that must remain coherent when responsibility for a patient transfers between organisations:

Identity. Is this the same patient at both ends of the transaction? Not just demographically matched, but verified, current, and contextually appropriate.

Consent. Has the patient authorised this transfer of their information and care? Not as a one-time checkbox, but as a dynamic, auditable state that travels with the data. This is what patient sovereignty actually requires.

Provenance. Where did this information originate? How has it been transformed? Can the receiving clinician trust it?

Clinical intent. Why is this transfer happening? What is the sending clinician asking the receiving clinician to do? What problem are they trying to solve together?

Alert and responsibility. Who is responsible for this patient right now, at this moment? When does responsibility transfer? How is that transfer acknowledged? What happens if nobody acknowledges it? This is where Minimum Viable Responsibility Transfer becomes essential.

Service routing. Given the clinical intent, the patient’s location, and the available capacity, where should this patient go next?

Outcome. Was the purpose of the transfer achieved? Did the patient get better? Was the question answered? Can the system learn from what happened?

These seven flows are the phase relationships of healthcare. When they are maintained across organisational boundaries, care is coherent. When they degrade, you get entropy: lost referrals, orphaned responsibility, medication errors, patients falling through gaps.

No amount of FHIR compliance addresses these flows in their entirety. FHIR can carry the data. But the governance of that data, the responsibility that must travel with it, the clinical intent that gives it meaning, the consent that authorises it, and the outcome measurement that closes the loop: these require infrastructure that does not yet exist at the inter-organisational layer.

The cost of incoherence

Entropy has a price. In the NHS, that price is measured in multiple currencies.

There is the human cost: 31,600 patient episodes of harm from undetected medication errors at transitions, every year ^[29]. These are not exotic clinical failures. They are the predictable result of information and responsibility degrading at organisational boundaries.

There is the economic cost: 13.5 million working hours lost annually to doctors fighting inadequate IT systems ^[30]. Clinicians re-entering data that already exists somewhere in the system but cannot be accessed. Administrators chasing referrals that were sent but never acknowledged. The Darzi Review found that NHS staff perceive IT as an additional burden rather than an enabler ^[11]. This is not a technology adoption problem. It is a coherence problem manifesting as technology frustration.

But the deepest cost is one that rarely appears in any report. It is the loss of the capacity for negentropy itself ^[1]. A system that cannot see itself as a whole cannot improve itself as a whole. It cannot learn from cross-organisational patterns. It cannot identify systemic failures that only become visible when you look across boundaries. It cannot adapt, because adaptation requires feedback loops, and feedback loops require coherent information flow, which is precisely what the system lacks.

The NHS is trapped in a thermodynamic spiral. Incoherence prevents the system from seeing itself. Inability to see itself prevents it from building coherence. Each reorganisation adds complexity without adding variety ^[2]. Each IT procurement adds capability without adding coherence. The entropy increases. ^[5]

Coherence as infrastructure, not overhead

Schrödinger’s deepest insight was that life is not a substance or a mechanism but a pattern of organisation maintained against entropy by continuous energetic work ^[1]. The order in a living cell is not static. It is dynamic, active, and dependent on infrastructure that never stops running.

The same principle applies to healthcare systems. Coherence is not something you can achieve through a programme of work and then move on to the next priority. It is not a project with a completion date. It is a continuous function that must be embedded in the infrastructure itself.

This is why treating coherence as compliance overhead is so dangerous. Clinical safety frameworks, governance requirements, data protection obligations: these are often experienced by healthcare organisations as friction, as cost, as burden. But they are, or should be, the metabolic processes that maintain organisational coherence. The problem is not that they exist. The problem is that they are implemented as bureaucratic checkpoints rather than as living infrastructure.

When compliance is a form you fill in before filing it in a drawer, it generates entropy rather than reducing it. When compliance is embedded in the infrastructure that carries clinical information between organisations, enforced not by auditors but by the architecture of the system itself, it becomes what biology would recognise as metabolism: the continuous work of maintaining order. ^[1]

The distinction is between coherence imposed from above, which decays the moment attention shifts, and coherence that emerges from infrastructure, which persists because the system cannot operate without it. The first is a policy. The second is an architecture. Only the second survives contact with reality.

The lightbulb and the laser, revisited

The NHS will not become a laser. No healthcare system will. Healthcare is too complex, too human, too irreducibly varied for that degree of alignment. Nor should it try. Beer was right that operational autonomy is essential ^[3]. Clinicians must be free to exercise judgement. Organisations must be free to adapt to local conditions. The goal is not regimentation.

But between the lightbulb and the laser, there is a vast spectrum. Right now, the NHS sits so far toward the lightbulb end that most of its energy dissipates at boundaries. Moving even modestly toward coherence would release enormous capacity that currently evaporates as entropy: the hours lost re-entering data, the errors generated by broken handoffs, the opportunities for learning that vanish into organisational gaps.

This requires infrastructure, not exhortation. You cannot make a lightbulb into a laser by asking the photons to cooperate. You have to build the resonant cavity, the gain medium, the mirrors that maintain the phase relationships. In healthcare, those mirrors are the governance flows that maintain identity, consent, provenance, clinical intent, responsibility, routing and outcome across organisational boundaries.

The Ten Year Plan wants to build a digital NHS ^[12]. Estonia and Denmark show that this is achievable ^{[27] [26]}. NPfIT shows what happens when you attempt it without foundations ^[10]. And physics shows why: coherence is not a feature you can add to an incoherent system. It is a property of the infrastructure itself, or it does not exist.

Schrödinger taught us that life maintains order through continuous metabolic work against entropy ^[1]. Ashby taught us that regulators must match the variety of what they regulate ^[2]. Beer taught us that autonomous parts require a meta-system to cohere ^[3]. Aksom has shown us that organisational disintegration is not a failing but a law ^[5].

The question is not whether the NHS needs transformation. It does, urgently. The question is whether we will build the governance infrastructure first, the foundations on which transformation can actually stand, or whether we will once again attempt to build the cathedral before laying the floor.

The test is simple. Take one ICS. Take one boundary — GP to acute trust, or hospital to community mental health. Map the seven flows across that boundary. Measure whether identity, consent, provenance, clinical intent, responsibility, routing, and outcome survive the crossing intact. If they do, you have coherence infrastructure. Build on it. If they don’t — and they won’t — then no amount of AI, no Single Patient Record, no ten-year plan will fix what happens to patients in that gap.

Start there. One boundary. Seven flows. Make coherence measurable before you make it policy. Physics does not care about political timetables. Entropy does not wait for ten-year plans. And the patient whose beta-blocker was missed last night needed coherence yesterday.

Julian Bradder

Founder & CEO, Inference Clinical

Julian builds governance infrastructure for inter-organisational responsibility transfer in healthcare. He has spent 30 years delivering digital transformation across government, defence, financial services, and healthcare. Full profile