Healthcare has published more AI strategy than any sector outside of technology. Every major payer. Every biopharma. Every health system with a budget large enough to hire a consultant. The documentation is meticulous.
In 2025, .
One category was different. , tools that transcribe clinical conversations in real time, generated $600 million in revenue. Not through incremental growth. Through sudden scaled adoption across systems that had been piloting AI for years without committing to anything.
The technology was not new. The capability had existed long enough to be piloted multiple times by the same organisations. What changed was not the tool. Something changed in the organisations that chose it.
The gap is not narrowing
The pattern is not specific to healthcare.
: 88% of organisations report regular AI use. Seven percent have it fully scaled. : 74% of companies struggle to achieve and scale AI value. Four percent are creating substantial value. BCG and MIT Sloan, surveying the same period: 47% of organisations have no strategy at all for what they are going to do with AI. They adopted the tool before the thinking arrived.
The gap between adoption and scaled deployment has not narrowed as the technology matured. It has widened. that the share of companies scrapping the majority of their AI initiatives jumped from 17% in 2024 to 42% in 2025. The average organisation abandoned 46% of its proofs of concept before they reached production.
More investment. More pilots. More abandoned projects.
The gap became a hiring problem
The most visible organisational response was a new job title.
Chief AI Officers appeared across financial services, retail, and healthcare. Major organisations appointed them, announced them, gave them C-suite mandates and direct CEO reporting lines. The role was created to close the gap between AI strategy and AI execution.
spent their first year documenting use cases. Not setting direction. Not committing to anything. Compiling inventories of what other parts of the organisation were already building.
HSBC's response was instructive. When the bank appointed its first Chief AI Officer in April 2026, it simultaneously expanded the remit of its Chief Technology Officer, creating an explicit separation between the strategy role and the execution role. The gap between knowing what to do and doing it had become wide enough to require two separate C-suite positions to cover it.
What the conventional diagnosis misses
The explanations for this gap are consistent across industries.
The technology is still developing. The ROI is unclear. The regulatory environment is uncertain. Legacy systems make integration expensive. The right use case has not yet been identified.
Each of these locates the failure in the environment: in the information available, in the clarity that has not yet arrived, in the external conditions that have not yet resolved. Each one implies the same solution. More information, more time, more analysis before committing.
The ambient scribes case makes this diagnosis hard to sustain.
The technology that transcribes clinical conversations existed before healthcare committed to it at scale. The ROI became clear after the commitment, not before. The regulatory environment did not change between 2023 and 2025. What changed was whether the sector had crystallised a position.
The ROI became clear after the commitment, not before. What changed was not the tool. What changed was whether the sector had crystallised.
Not a decision problem. A crystallisation problem.
The describes how strategic thought converts uncertainty into movement: Reason, Crystallise, Decide, Communicate. The cycle stalls. The question is always where.
The AI strategy problem is a . Not a failure to decide, but a failure to compress what is known into a working position clear enough to commit to. Not simplified. Not reduced to false certainty. Crystallised: the essential structure has become visible enough to act on.
The organisations still in pilot purgatory have not reached the decision phase. They are stuck before it. And the failure is invisible from the inside, because more reasoning feels like the answer.
Another pilot feels like progress. Another use-case audit feels like due diligence. Another cross-functional working group feels like appropriate care. None of them are wrong. They just do not address the crystallisation problem. They deepen it.
This is why the CAIO pattern is diagnostic. An organisation hired an executive to solve a crystallisation problem. The executive produced reasoning outputs, because that is what the organisation's intelligence rewarded. HSBC then split the role in two, because the organisation had learned that one person cannot bridge the gap between having AI intelligence and acting on it. The gap had become structural.
The stall generates its own deepening. Every reasoning activity feels like progress. The crystallisation problem remains invisible.
Strategy is not a roadmap. It is a position.
The question most boards are asking is a reasoning question: what AI strategy should we adopt?
It will not produce a crystallised answer.
The organisations that are moving have a different question. Not what should we do with AI, but what are we committed to that would be visibly wrong if it turned out to be wrong. Ambient scribes gave healthcare a crystallised position. Not a framework. Not a roadmap. A falsifiable claim: this capability, at this scale, for this workflow.
The sector did not get more information before it committed. It compressed what it had. The $600 million that followed was not the cause of the decision. It was the evidence that crystallisation had happened.
The incapacity to state that sentence is a diagnostic. It does not point to an information problem. It points at a failure to crystallise.
Crystallisation does not respond to more analysis. It responds to the willingness to have a position before the position is safe.