For AI healthcare investors
AI is scaling faster than healthcare can understand or govern.
A market this large has never been this unobservable. Over a thousand companies are selling AI into clinical workflows, and nothing independent records what is actually happening inside them.
Act one · The historical ledger
This market has run this experiment before.
01
1998
Roughly 2.4 million websites existed. Small enough for humans to catalogue by hand, and Yahoo’s directory did exactly that.
02
2000
Over 17 million. Directories broke under their own growth, and the company that indexed reality instead took the market.
03
Today
The FDA has authorised more than 1,000 AI enabled medical devices, from a handful a decade ago, and over 1,000 companies are selling AI into clinical workflows.
04
Still
Healthcare AI is catalogued by hand: vendor claims, committee decks, isolated studies. Nothing indexes what is actually happening.
Act two · The problem
AI adoption is becoming impossible to see.
Adoption did not solve the problem, it created a new one. Deployment happens in thousands of separate places at once and no one holds a complete view of it. These are the questions the market cannot answer about any of the companies selling AI into healthcare.
01
Which tools are actually used
Not announced, not purchased. In genuine daily use, in a named specialty, in a real organisation.
02
Where they succeed
Which workflows they hold, and which they were quietly withdrawn from after the pilot.
03
Where users struggle
The friction that never reaches a press release and never reaches the vendor either.
04
Which vendors improve
Performance after deployment, over time, rather than at the moment of the study.
05
Which are gaining trust
Clinical trust is directional and observable, and today nobody measures its movement.
Act three · One organisation
Take a single large academic health system.
It may run dozens of AI systems at once: imaging triage in radiology, risk scoring in emergency medicine, ambient documentation in outpatient clinics, coding and scheduling in operations, research tooling across departments. Each was bought separately, justified separately, and each is now live.
Nothing independent shows which of those systems are genuinely in use, where adoption quietly failed, which were abandoned after the pilot, or what the clinicians using them actually experience. The vendors each hold their own version. The organisation holds fragments.
The market holds a press release.
Act four · The precedent
Every technology wave produced a winner that indexed reality.
The internet
Too much information
Google indexed reality
Markets
Too much complexity
Bloomberg created market intelligence
Professional life
Too many disconnected people
LinkedIn created professional identity
Healthcare AI
Too many systems, workflows and experiences
The layer does not exist yet
In each case the durable company was not the one that built the technology. It was the one that became the place the market went to understand itself.
Act five · Why current answers fail
Four answers exist today. None of them close the gap.
01
Vendor claims
"The product works." The company measuring its own performance has different incentives from the people using it.
02
Academic evidence
"The study showed benefit." A trial captures controlled conditions, rarely the messy reality of everyday deployment.
03
Hospital committees
"We evaluated this tool." The knowledge is real, it stays inside one organisation, and nothing compounds.
04
Surveys
"We asked users." A snapshot of one moment, not continuous intelligence that moves as reality moves.
The structural failure is the same in all four: vendors know what they build, organisations know what they buy, users know what they experience, and nothing connects the three.
Act six · The layer
A continuously learning intelligence layer, built from what people actually observe.
01
People
The clinical and operational users who experience an AI system in their actual work.
02
Organisations
The hospitals, physician groups, ACOs and virtual care platforms where deployment really happens.
03
Technologies
The specific systems, held as canonical identities rather than vendor brand names.
04
Evidence
What is independently observed, corroborated and contradicted over time.
05
Outcomes
What actually changed as a result, in workflow and in practice.
Not a directory and not a review site. A graph in which every observation is attached to a real organisation, a real specialty and a real technology, so intelligence accumulates instead of expiring.
Act seven · Why now
Three forces converged.
01
Deployment exploded
Over a thousand companies are already selling into healthcare workflows. Volume alone has broken evaluation.
02
Governance became mandatory
Organisations are now accountable for AI they cannot currently see, and accountability without visibility is untenable.
03
AI became infrastructure
When AI runs daily operations, understanding adoption matters as much as understanding the technology.
Act eight · Why it compounds
Every participant increases the intelligence available to everyone.
01
People record what they observe
One sentence, attached to a real organisation, specialty and technology.
02
Observations spread
Across more organisations and more specialties, which is breadth rather than volume.
03
Intelligence improves
Independent corroboration turns scattered accounts into an answerable record.
04
Records become valuable
For every technology, company and category in the market.
05
More participants arrive
And the loop begins again, one turn stronger than before.
A database depreciates. A network of independent observation appreciates, and the record of what happened cannot be reconstructed later by anyone who was not there.
The trillion dollar question is no longer whether healthcare will use AI.
It is who becomes the intelligence layer that helps healthcare understand and govern it.