The three maturities of healthcare AI.
Seventeen categories of AI in healthcare, clinical and administrative, scored on evidence, deployment and economics. A category only scales when all three line up. The method, the scored read of every category, and what moves next. As at August 2026, refreshed quarterly.
Most maps of healthcare AI list companies by specialty or products by technology, and tell a buyer, a board or an investor almost nothing about what is actually ready. This paper asks a harder question: what does it take for a category of healthcare AI to scale? The record gives a consistent answer. Independent evidence, deployment in routine care, and a repeatable way to get paid. All three at once, and the categories that stalled were missing exactly one.
The paper scores seventeen categories, eleven clinical and six administrative, on those three maturities, sets each category's status by its weakest score, and reads the gaps: deployment ahead of evidence is the fragile pattern, evidence ahead of economics is proven work waiting for a budget, and money ahead of proof is where buyers get burned. Every score is backed by dated, cited sources, published openly so the judgement can be challenged.
What is inside
- The three-maturities method, and why the weakest score decides whether a category scales
- The natural experiment: the MASAI randomised trial against the Epic Sepsis Model, and what it teaches about evidence
- The autonomy ladder, the MHRA regulatory line of July 2026, and the 2026 payment inflection: CPT codes, add-on payments and the NHS's £10 billion programme
- The radar: proof against budget across all seventeen categories, with the full scorecard
- Short evidenced reads of every category, from radiology and breast screening to ambient scribes and agentic operations
- Where the opportunity sits for investors, for companies selling into the NHS, and across the clinical decade
Who it is for
Investors and PE operating partners deciding which categories deserve capital and which valuations carry hidden evidence risk. Founders and commercial leaders placing their company against the maturity of its category. Boards and health-system leaders judging which AI is ready to buy and which is still selling ahead of the proof. It is a market read, not investment, clinical or regulatory advice.
Written from an operating perspective
By Andrew Wyatt, founder of Ortent Advisory. Four exits over three decades (Lotus to IBM, Paragon to Phone.com, Apertio to Nokia $240M, Clearswift to Lyceum) plus CGO and COO seats in digital health and life sciences at Sapio Sciences and Lumeon. The paper pairs with the interactive radar, where every score opens to the evidence behind it.