In April 2026, the non-profit amfAR launched a $2 million, three-year study called the HIV Immune Atlas Study. The project uses AI and machine learning to map the HIV reservoir and inform cure research. Separately, the US Centers for Medicare and Medicaid Services (CMS) launched the Health Tech Ecosystem, a voluntary initiative to improve medical record portability, integrate consumer apps with AI tools, and reduce administrative burdens.
These are not speculative announcements. They are funded, scoped, and operational.
Healthcare AI is entering a phase where the conversation has shifted from “can AI work in medicine?” to “which specific applications deliver reliable results?” The distinction matters because the hype phase produced unrealistic expectations. The pragmatic phase is producing measurable outcomes.
Applications That Are Working
- Medical imaging analysis has the strongest evidence base. AI systems that analyse radiology images, pathology slides, and retinal scans have achieved diagnostic accuracy comparable to specialist physicians in controlled studies. The operational challenge is integration with clinical workflows. A model that delivers accurate results but requires physicians to switch between systems adds friction that limits adoption.
- Administrative automation is delivering the fastest ROI. Prior authorisation processing, medical coding, claims processing, and appointment scheduling are high-volume, rules-based tasks where AI reduces processing time and error rates. These applications do not require FDA approval and carry lower regulatory risk than clinical decision support tools.
- Drug discovery pipeline acceleration uses AI to screen molecular compounds, predict drug interactions, and identify candidate molecules for further testing. The amfAR HIV Immune Atlas Study falls into this category: using ML to map biological processes at a scale and speed that manual research methods cannot match.
- Clinical documentation tools that generate structured clinical notes from physician-patient conversations are gaining adoption. These tools reduce the administrative burden that contributes to physician burnout while producing more consistent documentation.
Applications That Are Struggling
- Diagnostic decision support tools face regulatory, liability, and trust barriers. Physicians are reluctant to rely on AI recommendations for treatment decisions because the liability framework has not caught up with the technology. If an AI-recommended treatment causes harm, the question of responsibility remains legally ambiguous in most jurisdictions.
- Population health management tools that use AI to predict patient risk scores and allocate resources have produced mixed results. The models are accurate in aggregate but generate false positives at rates that overwhelm clinical teams with low-value alerts.
CMS Health Tech Ecosystem: What It Signals
The CMS initiative is significant because it addresses the infrastructure layer that health AI depends on. Medical record portability, consumer app integration, and administrative standardisation are prerequisites for effective health AI deployment. Without interoperable data, AI models operate on incomplete information.
CMS attracting participation from hundreds of health tech firms indicates that the industry recognises this infrastructure gap. The initiative is voluntary, which limits its enforcement power. But it creates a coordination mechanism that the fragmented US health tech market has lacked.
What This Means for Your Business
Healthcare AI is a market where infrastructure matters more than algorithms. The organisations capturing value are those solving data interoperability, workflow integration, and regulatory compliance, not those building marginally better models.
FortySeven’s AI in Healthcare and Biotech practice builds the infrastructure layer that healthcare AI depends on. We design interoperable data architectures, clinical workflow integrations, and regulatory-compliant AI deployment pipelines.
