The healthcare AI tool landscape

The healthcare AI landscape is large and moving quickly. Learn the categories and their risk profiles — specific vendors and features change constantly (this is a 2026 snapshot; verify before relying on any specific tool, and naming a tool here is not an endorsement).

The categories, roughly by clinical risk:

Higher clinical risk (they influence diagnosis or treatment):

• Medical imaging / diagnostic AI — analyze radiology, pathology, cardiology images to flag findings. This is the largest FDA-cleared category (roughly three-quarters of authorized AI-enabled devices are radiology). Many are FDA-cleared medical devices (Module 3).
• Clinical decision support (CDS) — risk scores, early-warning systems (e.g., sepsis prediction), drug-interaction checking. Some qualify for an FDA exemption, some are regulated devices (Module 3).
• Patient-facing symptom checkers / triage chatbots — the highest-risk consumer-facing category, because patients may act on output without a clinician.

Lower clinical risk (they support documentation and operations):

• Ambient clinical documentation / AI scribes — listen to the visit and draft the note (Abridge, Microsoft's Dragon Copilot, Suki, Nabla, Heidi). The clinician still reviews and signs, so risk is bounded — but transcription can introduce errors (Module 2).
• Administrative and revenue-cycle AI — prior authorization, coding, billing, scheduling, claims. High value, lower clinical risk — though payer-side automation has drawn major controversy (Module 2).
• Operational copilots — draft, summarize, and retrieve from clinical or administrative systems.

The critical distinction — is it a regulated medical device? Some healthcare AI is an FDA-regulated Software as a Medical Device (SaMD); some is not. That line (Module 3) determines what oversight applies. As a rule of thumb, software that analyzes a medical image or a physiologic signal, or that a clinician relies on without being able to independently check its reasoning, tends to be regulated as a device.

Why think capabilities, not vendors: vendors and product names change every quarter, but the capabilities and their risk profiles are stable. An imaging tool carries device and diagnostic-accuracy risk whether it is called Tool A or Tool B this year. Evaluate healthcare AI by what it does, what clinical risk that creates, and its regulatory status and evidence — not by brand.

The mindset: the healthcare AI stack ranges from lower-risk documentation and administrative tools to higher-risk diagnostic and decision-support tools — some of which are FDA-regulated devices. Learn the categories and their risk profiles, evaluate tools by capability, evidence, and regulatory status rather than brand, and match your caution to the category. Knowing which category creates which risk is the foundation for deploying healthcare AI responsibly.