How AI Is Used in Healthcare: Real Use Cases, Rules and Limits
Where AI actually helps in healthcare today, from imaging and clinical notes to scheduling, plus the FDA, HIPAA and fairness rules that decide how fast it spreads.
AI in healthcare is used mainly in 3 areas: medical imaging, where FDA-authorized tools help flag findings on scans; clinical documentation, where ambient AI drafts visit notes for clinicians to review; and operations like scheduling, billing and patient messages. It assists clinicians rather than replacing them, and HIPAA, FDA rules and anti-discrimination law shape how it is adopted.
Photo: Unsplash
On this page
- Where AI is used in healthcare today
- Medical imaging: the most regulated use
- Clinical documentation: the fastest-growing use
- Scheduling, messages and the back office
- The rules that shape AI adoption in healthcare
- What AI in healthcare still cannot do
- Where to start: a checklist for practices and clinics
- The bottom line
- Sources
Ask a physician what they would change about their job, and many will not mention diagnosis at all. They will talk about the hours spent typing notes after clinic, the inbox full of patient messages and the scheduling gaps that leave rooms empty one hour and overbooked the next.
That is where much of AI’s real impact in healthcare is landing. The headline-grabbing uses, like algorithms that read scans, are real and regulated. But the quieter uses in documentation and operations are spreading fastest because they save time without making clinical decisions. This guide covers the main use cases, what the FDA and privacy rules require and where AI still falls short.
This is general information, not medical advice.
Where AI is used in healthcare today
| Area | What AI does | Who uses it | Regulatory touchpoint |
|---|---|---|---|
| Medical imaging | Flags possible findings, prioritizes urgent scans, measures structures | Radiology, cardiology, ophthalmology | Often an FDA-regulated medical device |
| Clinical documentation | Drafts visit notes from recorded, consented conversations | Primary care, specialists, hospitals | HIPAA; usually not an FDA device when clinicians review the draft |
| Patient messages and triage | Sorts and drafts replies to portal messages, routes requests | Clinics and health systems | HIPAA; clinical review of medical advice |
| Scheduling and access | Predicts no-shows, fills cancellations, balances provider calendars | Front desk and operations teams | HIPAA |
| Revenue cycle | Suggests codes, checks claims before submission, drafts appeals | Billing and coding teams | HIPAA; payer and billing rules |
| Clinical decision support | Alerts and risk scores inside the EHR | Hospitals and clinics | FDA CDS guidance; HHS anti-discrimination rule |
For a deeper look at how clinics roll out these tools, see our companion guide to AI in healthcare in 2026.
Medical imaging: the most regulated use
Imaging was the first area where AI became an everyday clinical tool, and it remains the largest. The FDA maintains a public list of AI-enabled medical devices authorized for marketing in the US, and radiology accounts for the large majority of entries. The agency notes the list is not comprehensive and is updated periodically.
Typical imaging tools:
- Triage and prioritization. Software flags scans that may show a time-sensitive finding, so a radiologist reads them sooner.
- Detection aids. Tools mark areas that may need a closer look on mammograms, chest images or other scans.
- Measurement. Software automates measurements that used to be done by hand, such as heart chamber volumes.
- Screening. In April 2018 the FDA authorized the first AI-based device to detect more-than-mild diabetic retinopathy in adults with diabetes without a clinician first interpreting the images, allowing screening in primary care offices.
Most imaging AI works as a second set of eyes. A licensed clinician still reads the study and makes the diagnosis.
How the FDA regulates changing AI
Traditional medical devices change rarely. AI models can be retrained and improved. To handle that, the FDA finalized guidance in December 2024 on predetermined change control plans (PCCPs) for AI-enabled device software. A PCCP lets a manufacturer describe planned model updates and how it will validate them, so authorized changes can be made without a new marketing submission each time.
Clinical documentation: the fastest-growing use
“Ambient” documentation tools listen to a consented patient visit, transcribe it and draft a structured clinical note. The clinician reviews, edits and signs the note in the electronic health record (EHR).
Why it is spreading:
- It targets burnout directly. Documentation time is one of the most common complaints in clinical work.
- It keeps humans in charge. The clinician approves every note, which lowers the clinical risk compared with autonomous tools.
- It fits existing workflows. Many tools integrate with major EHR systems.
The risks are real, though. Drafted notes can contain errors, omissions or details the patient never said, which is the same AI hallucination problem seen in any generative tool. Clinicians must read every note before signing, and practices need clear patient consent processes for recording.
Key takeaway: The safest high-value AI uses in healthcare keep a clinician in the loop: AI drafts, flags or prioritizes, and a licensed professional decides. Tools that act on their own face much tighter FDA scrutiny.
Scheduling, messages and the back office
Some of the quickest wins are administrative:
- No-show prediction. Models estimate which appointments are likely to be missed, so staff can send extra reminders or overbook carefully.
- Waitlist and cancellation filling. Automated outreach offers open slots to patients who want an earlier visit.
- Portal message sorting. AI categorizes incoming messages (refill, billing, clinical question) and drafts replies for staff or clinicians to approve.
- Coding and claims. Tools suggest billing codes from documentation and check claims for errors before submission.
- Prior authorization. Software gathers required documentation and drafts requests.
These uses rarely involve FDA oversight, but they almost always involve protected health information, so privacy rules apply in full. Our guide to telehealth and remote patient monitoring covers related digital front-door tools.
The rules that shape AI adoption in healthcare
This is not legal advice. Confirm current requirements with qualified counsel, as several rules have changed recently.
HIPAA. Any AI vendor that creates, receives, stores or transmits protected health information on behalf of a covered entity is generally a business associate and must sign a business associate agreement (BAA). Never put patient information into a consumer AI tool that has not signed a BAA. HHS also expects administrative, physical and technical safeguards for electronic health information under the HIPAA Security Rule.
FDA device rules. Software that diagnoses, treats or analyzes medical images is generally regulated as a medical device. The FDA’s Clinical Decision Support Software guidance, revised in January 2026, explains which decision support tools fall outside device rules. A key factor is whether a clinician can independently review the basis for the software’s recommendation rather than relying on it blindly.
Anti-discrimination rules. In May 2024 HHS finalized a rule under Section 1557 of the Affordable Care Act that requires covered entities to make reasonable efforts to identify and reduce discrimination risks from patient care decision support tools, including AI, with a compliance date of May 1, 2025. Check the HHS Office for Civil Rights for the current status.
EHR transparency rules. ONC’s HTI-1 rule, finalized in early 2024, required certified EHR developers to provide “source attribute” information about predictive decision support tools, similar to a nutrition label. In December 2025 ONC (now ASTP/ONC) proposed the HTI-5 rule, which would remove those requirements. As of this writing, check healthit.gov for whether it has been finalized.
State laws. Several states have passed rules on AI in healthcare, including disclosure requirements for AI-generated patient communications and limits on AI in insurance coverage decisions. Check the rules in each state where you practice.
What AI in healthcare still cannot do
- Replace clinical judgment. AI tools support specific tasks. They do not examine patients, weigh values or take responsibility.
- Work equally well for everyone. A model trained mostly on one population may perform worse on others. Ask vendors how tools were validated across age, sex, race and care settings.
- Transfer perfectly between sites. A model that performs well at one hospital can perform worse at another with different equipment, patients or workflows. Local testing matters.
- Guarantee accuracy. Generative tools can invent details, and predictive tools produce false alarms and misses.
- Explain every decision. Many models cannot show exactly why they produced a result, which complicates clinician trust and review.
Where to start: a checklist for practices and clinics
- Pick a pain point, not a technology. Documentation time, no-shows or message backlog are common starting points.
- Confirm the regulatory status. Ask whether the tool is FDA-authorized, and for what intended use, or why it is not a device.
- Get a signed BAA before any patient data touches the tool.
- Ask for validation evidence from settings and patient populations similar to yours.
- Run a time-limited pilot with a small group of clinicians or staff, and measure time saved and error rates.
- Set review rules. Decide who checks AI output and how errors are reported.
- Tell patients. Explain when AI is used, especially for recording visits, and get consent where required.
The bottom line
AI in healthcare is already practical in imaging, documentation, patient messages and operations. The fastest-growing tools save clinician and staff time while keeping a person responsible for decisions. Adoption depends as much on rules as on technology: FDA oversight for diagnostic tools, HIPAA for anything touching patient data and growing attention to fairness and transparency. Start with one workflow problem, demand evidence and keep clinicians in the loop. For related terms, see our glossary.
Sources
- FDA: Artificial Intelligence-Enabled Medical Devices list — public list of AI-enabled devices authorized for US marketing
- FDA: Clinical Decision Support Software guidance — January 2026 guidance on which decision support software is regulated as a device
- HHS: Business Associates under HIPAA — when vendors handling protected health information need a BAA
- ASTP/ONC: HTI-5 Proposed Rule — proposal to remove HTI-1 decision support source attribute requirements
- NIST AI 600-1: Generative AI Profile — risks of confabulated content in healthcare summaries
Frequently asked questions
How is AI used in healthcare today?
AI is used mainly for medical imaging analysis, drafting clinical notes, sorting patient messages, predicting appointment no-shows and supporting billing and coding. In imaging, FDA-authorized tools flag possible findings for radiologists. In documentation, ambient AI drafts notes from consented visits for clinicians to review. Nearly all of these tools support professionals rather than making independent decisions.
Is AI in healthcare regulated by the FDA?
Some of it is. Software that diagnoses, treats or analyzes medical images is generally regulated as a medical device and needs FDA authorization. Many administrative tools, such as scheduling or billing software, are not FDA devices. Some decision support software is excluded from device rules if it meets criteria in FDA's Clinical Decision Support guidance. This is not legal advice.
Is AI in healthcare HIPAA compliant?
AI tools can be used in a HIPAA-compliant way, but compliance depends on how they are set up. A vendor that handles protected health information for a covered entity generally must sign a business associate agreement and apply required safeguards. Consumer AI apps without a BAA should never receive patient information. The practice remains responsible for its own compliance.
Will AI replace doctors?
No. Current AI tools handle specific tasks, such as flagging findings on a scan or drafting a note, and a licensed clinician remains responsible for diagnosis and treatment. AI is more likely to change how clinicians spend their time, reducing typing and administrative work, than to replace them. Patients still need examination, judgment and communication that AI cannot provide.
What are the risks of AI in healthcare?
The main risks are inaccurate output, bias against patient groups underrepresented in training data, privacy breaches, over-reliance by busy clinicians and poor performance when a tool is moved to a new setting. Generative tools can also invent details in notes or summaries. Validation, human review, signed BAAs and ongoing monitoring reduce these risks but do not eliminate them.
How much does AI cost for a medical practice?
Costs vary widely by tool and vendor. Ambient documentation and other clinical AI tools are commonly sold as per-clinician subscriptions, while imaging AI may be licensed per site, per study or bundled with equipment. Some EHR vendors include AI features in existing contracts or add-ons. Ask vendors for full pricing, including integration, training and support.
Do patients have to consent to AI being used in their care?
It depends on the use and the state. Practices commonly get patient consent before recording visits for ambient documentation, and state recording laws may require it. Some states also require disclosure when AI generates patient communications. Consent and disclosure rules are evolving, so check your state's requirements with counsel. This is not legal advice.
Was this guide useful
No reviews yet. Share your experience and help other readers.
Write a review
Manishdai
Manishdai writes about the technology that is reshaping every industry — from AI agents and cloud platforms to healthtech, fintech and the IT companies building it all across the United States.
manishdai.com