Evidence landscapes
What the published record shows — and does not show — for each major category of healthcare AI, before you look at a single vendor.
A vendor review scores a specific product against a published method. An evidence landscape does something different: it describes the state of independent validation, the regulatory picture, the transparency norms and the security expectations for an entire product category, drawing only on peer-reviewed literature, regulatory databases and published frameworks.
The purpose is to give a buyer the base rate before they read a marketing deck. If the independent evidence for an entire category is sparse, a vendor claiming robust validation should be asked to show it. If the regulatory path for a category is genuinely ambiguous, a vendor asserting clarity should be asked to explain how they arrived at it. The landscape is the context that makes those questions possible.
These pages name no vendors and review no products. They describe what the evidence record contains for the category, where the gaps are, and what a buyer should ask before signing anything.
The Institute makes no clinical, safety, regulatory or outcome claim about any named product.
Published landscapes
Ambient clinical documentation AI
The state of independent evidence for AI-powered ambient scribes and clinical documentation tools. Covers the validation gap, the regulatory gray area, transparency norms, data training rights, and the questions a buyer should ask before procurement.
Revenue cycle AI
What the published record shows for AI in prior authorisation, coding, denial management and charge capture. Covers the difficulty of isolating AI-driven ROI, the integration variable most pilots ignore, and the regulatory picture for tools that sit outside clinical decision-making.
How these are built
Each landscape is assembled from peer-reviewed literature, FDA regulatory databases, published frameworks from organisations such as NIST and the Coalition for Health AI, and publicly available policy documents. No vendor-supplied materials are used. Every quantitative claim carries a dated, named source in the same sentence. Where the evidence is absent or contested, the page says so rather than filling the gap with a plausible number.
Landscapes are reviewed at least every six months and carry a last-reviewed date. If you find an error or a missing source, email [email protected] with the page, the sentence, and what you say the correct position is. We correct established errors within one working day.