Task-level reachability by current AI systems across the published healthcare sample.
AI and Healthcare jobs
Healthcare sits at both ends of the AI story. Clinical documentation, coding, and transcription are among the most automatable workflows in the economy, while licensed, hands-on care is among the least. The industry's labor shortage means AI mostly arrives as workload relief — but the administrative layer faces genuine substitution.
Estimated potential for task transfer to software.
Estimated potential for AI to expand worker output while keeping human accountability.
Displacement pressure 84 — the most exposed published role in this industry.
Displacement pressure 14 — the strongest anchor role in this industry.
Distribution
How healthcare roles spread across the pressure scale
Each bar counts published healthcare roles in a 5-point displacement-pressure band. Red bars mark scores of 70 or higher. The dashed line marks the industry median of 28.
Documentation is the exposed layer, care is not
The sharpest pressure in healthcare falls on work that produces records rather than care: medical transcription, records abstraction, and coding-adjacent data entry. Ambient clinical documentation and AI coding assistants compress these workflows directly. By contrast, bedside nursing, dental chairside assistance, and physical therapy are physical, licensed tasks where AI's realistic role is drafting notes and flagging changes — reducing burnout more than headcount.
Licensing and liability slow substitution
Healthcare work is wrapped in scope-of-practice rules, privacy law, and malpractice accountability. Automated coding suggestions still require certified review, AI image reads still require a radiologist's sign-off, and someone licensed must still position the patient and operate the scanner. These constraints do not make roles immune, but they shift AI from replacement toward supervised augmentation in most clinical settings.
Imaging anxiety is real but mis-aimed
AI imaging tools read scans remarkably well, which creates anxiety across radiology-adjacent roles. The anxiety mostly misidentifies the target: radiologic technologists and sonographers acquire images through physical, operator-dependent technique that AI cannot perform, and the tools augment the reading step instead. The roles more exposed are the ones feeding the imaging pipeline with paperwork, not the ones producing the scans.
Demographics outrun automation
An aging population keeps demand for nurses, aides, therapists, and technicians structurally strong even as tools improve. For healthcare workers, the practical strategy is rarely fleeing the industry — it is moving from the documentation layer toward licensed, hands-on, or coordination work, or into the informatics and quality-review roles that oversee the AI systems themselves.
Occupation pages
Compare AI risk across healthcare roles
Each page below includes task-level exposure, automation and augmentation scores, wage context, transition pathways, upskilling priorities, and a 90-day planning outline.
SOC 29-1141
LowRegistered Nurses
Documentation and administrative follow-up can change quickly, but hands-on care, clinical judgment, licensing, and patient trust constrain direct replacement.
- Exposure
- 28
- Automation
- 13%
- Augment
- 46%
SOC 29-2061
LowLicensed Practical and Licensed Vocational Nurses
Charting and vitals documentation are being streamlined by AI documentation tools, but administering medication, wound care, patient observation, and hands-on basic nursing are licensed physical tasks with durable demand in aging-care settings.
- Exposure
- 30
- Automation
- 12%
- Augment
- 44%
SOC 31-9092
LowMedical Assistants
Scheduling, chart preparation, and patient messaging can be augmented. Hands-on care, rooming patients, vital signs, specimen handling, and local clinical protocols keep the role comparatively resilient.
- Exposure
- 38
- Automation
- 18%
- Augment
- 49%
SOC 29-2072
ModerateMedical Records Specialists
Chart abstraction, data entry, and record retrieval are exposed to AI coding and documentation tools. Coding accuracy disputes, privacy rules, release-of-information judgment, and clinician clarification keep humans accountable.
- Exposure
- 68
- Automation
- 48%
- Augment
- 44%
SOC 31-9094
Very HighMedical Transcriptionists
Speech recognition and ambient clinical documentation have automated the core dictation-to-text workflow that defined this occupation. Employment is declining sharply, and remaining work centers on editing AI output for terminology accuracy.
- Exposure
- 90
- Automation
- 76%
- Augment
- 24%
SOC 29-2052
ModeratePharmacy Technicians
Counting, labeling, and inventory tasks are increasingly handled by dispensing robots and pharmacy automation. Patient interaction, insurance problem-solving, sterile compounding, and pharmacist support keep certified technicians needed.
- Exposure
- 56
- Automation
- 38%
- Augment
- 42%
SOC 31-9091
LowDental Assistants
Chairside assistance, instrument sterilization, and patient preparation are physical, in-person tasks AI cannot perform. Digital records, AI imaging review, and scheduling tools change the administrative portion of the job, not the clinical core.
- Exposure
- 30
- Automation
- 12%
- Augment
- 36%
SOC 31-2021
LowPhysical Therapist Assistants
Guiding patients through exercises, manual therapy, and mobility training is embodied work that requires presence, touch, and motivation. Documentation tools reduce note-writing burden; the therapeutic relationship and hands-on treatment remain the job.
- Exposure
- 22
- Automation
- 8%
- Augment
- 34%
SOC 29-2034
LowRadiologic Technologists and Technicians
AI image analysis anxiety targets radiologists more than technologists. Patient positioning, radiation safety, equipment operation, and image quality judgment are licensed, hands-on tasks where AI serves as a second reader rather than a replacement.
- Exposure
- 42
- Automation
- 22%
- Augment
- 52%
SOC 29-2032
LowDiagnostic Medical Sonographers
Ultrasound is uniquely operator-dependent: image quality comes from real-time probe technique on a moving patient. AI helps with measurements and documentation, but acquiring diagnostic images, judging anatomy live, and deciding when to extend an exam remain human skills.
- Exposure
- 36
- Automation
- 16%
- Augment
- 50%
Questions
AI and healthcare jobs: common questions
Will AI replace nurses?
No credible scenario replaces nurses near term. AI is absorbing documentation, triage support, and message drafting, which changes the workday but not the need for licensed hands-on care. Persistent staffing shortages mean the tools are deployed as workload relief, and clinical judgment, patient trust, and physical care remain human-accountable tasks.
Which healthcare jobs are most at risk from AI?
Medical transcriptionists face the clearest substitution, since speech recognition and ambient documentation automate their core task directly. Medical records specialists and billing-adjacent roles are next, because abstraction, coding, and claims workflows are structured and rules-based. Even there, compliance review keeps humans in the loop, but the task mix is shrinking fastest at the documentation layer.
Is radiology a safe career with AI reading scans?
Safer than the headlines suggest for technologists. AI reading tools target interpretation, which is the radiologist's step. Radiologic technologists and sonographers physically position patients, operate equipment, manage radiation dose, and judge image quality in real time — licensed, hands-on work. The technology increases throughput, which tends to support demand for skilled operators rather than eliminate them.
What healthcare roles should admin workers move into?
The strongest bridges convert documentation knowledge into oversight work: clinical documentation integrity, health information analysis, coding quality audit, and pharmacy or imaging automation technician roles. These paths reuse medical terminology and workflow knowledge while adding AI-supervision skills, and they typically preserve or increase wages compared with pure transcription or records-entry positions.