Use cases
Use-case library

What each AI use case actually changes

Most published healthcare AI use cases are a paragraph of benefits and a vendor logo. Every entry here has to fill seven fields before it is allowed to exist: who owns it, what you measure before you start, an effect size with a citation you can open, what changes in the workflow, three ways it fails, its governance risk tier, and the conditions under which the honest answer is don't.

An entry that cannot fill all seven from real evidence is not written. That rule is enforced at build time, not by intention.

Clinical operations · Higher risk

AI-Assisted Mammography Screening

How AI-supported mammography reading increases cancer detection rates, reduces interval cancers, and cuts radiologist workload.

Revenue cycle · Moderate risk

AI-Assisted Medical Coding for Revenue Cycle Management

How NLP-driven AI coding tools improve ICD-10 coding accuracy, reduce claim denials, and accelerate the revenue cycle.

Patient access · Lower risk

AI-Driven Patient Scheduling and No-Show Prediction

How machine learning predicts appointment no-shows and enables targeted interventions to reduce missed visits and idle clinic time.

Clinical operations · Moderate risk

Ambient AI Scribes for Clinical Documentation

How ambient AI scribes reduce documentation time, after-hours charting, and clinician burnout in ambulatory care settings.

How to read one

The two fields most people skip are the ones worth reading first. Three failure modes is where the monitoring plan comes from — a failure you have named before deployment is one you can watch for. When not to do this is the field a vendor will never write for you, and it is usually the reason a pilot that worked somewhere else does not work here.

Every effect size links to its source. Where a number could not be sourced it is not printed, and the mechanism is argued without it.