■ MODERATE RISK ■ Healthcare
No, though the job is being rebuilt around them. Continuous monitoring is replacing the ritual annual medical, but fitness-for-duty verdicts, legal certifications, and telling an employer something it doesn't want to hear all require a licensed human with liability.
“AI monitors workplace health data continuously. Your annual checkup is a formality.”
Our AI replacement risk score — how we score jobs
Occupational physicians sit at the awkward junction of medicine, law, and employment. The caseload: fitness-for-duty exams for safety-critical roles, return-to-work assessments after injury or illness, health surveillance for hazard-exposed workers (hearing, lung function, blood lead), workers' compensation evaluations that end up contested, workplace risk assessments, and advising employers on accommodations — all while serving two masters, since the patient and the paying client are different parties with different interests.
Data is eating the surveillance half. Wearables and environmental sensors now track noise dose, air quality, fatigue, and biometrics continuously, making the annual audiogram-and-questionnaire ritual look like sampling error. AI flags exposure anomalies and health drift in real time, triages which workers actually need physician review, and drafts the surveillance reports and certificates that consume clinic hours. Routine pre-employment medicals for low-risk roles are already migrating to questionnaire-plus-algorithm pipelines with nurse oversight, thinning a traditional revenue line.
What stays human is everything with a signature and a dispute attached. Certifying a pilot, driver, or crane operator fit for duty carries legal weight that regulators assign to licensed physicians; workers' comp opinions get cross-examined; and the delicate judgment calls — is this worker's back claim consistent with the MRI, can this role be accommodated, is this employer's testing program ethical — involve navigating conflicting interests no algorithm is trusted to referee. Meanwhile mental health, ergonomics of remote work, and aging workforces are expanding the specialty's remit. Our 33: the screening factory automates briskly, but the medico-legal referee — the physician whose signature survives cross-examination — does not.
Automatability: our editorial assessment of current and near-term AI capability
The shift is underway: sensor-based surveillance and algorithmic triage are in deployment at large employers now, and routine screening medicals will be substantially automated by 2030, with physicians reviewing exceptions rather than everyone. Through the 2030s the specialty consolidates around medico-legal certification, complex cases, and program design — fewer routine clinic hours, more judgment work. Regulatory requirements for physician sign-off anchor the profession well beyond that horizon.
Changing, not declining. Routine screening medicals — long the specialty's bread and butter — are automating via sensors and algorithmic triage. But the medico-legal core (fitness-for-duty certification, contested compensation cases) legally requires physicians, and workplace mental health, aging workforces, and monitoring-ethics questions are expanding the advisory role. Physicians who move up from screening to judgment work will find plenty of demand.
For low-risk roles, largely yes — questionnaire-plus-algorithm pipelines with nurse oversight already handle many pre-employment checks, and continuous wearable monitoring beats an annual snapshot for surveillance. For safety-critical certification (pilots, drivers, heavy equipment operators), regulators require a licensed physician's signed judgment, and that requirement is not loosening. The exams that survive are the ones with legal weight.
It flips the workflow. Instead of examining everyone annually and finding little, physicians will review the workers whose sensor data — noise dose, lung function trends, fatigue markers — actually flags concern. That's better medicine and fewer routine hours. The new physician work it creates: designing monitoring programs, auditing their ethics and accuracy, and handling the disputes when data and worker disagree.
Shift the mix toward what algorithms can't sign. Build expertise in contested evaluations and safety-critical certification, learn the monitoring platforms well enough to design and critique employer programs, and develop the mental-health and accommodation advisory work employers increasingly buy. Letting AI draft your reports is the easy win — the hours saved go to the complex cases that pay better anyway.