SAFE
Internists spend most of their day on things AI cannot do alone: physical examination, in-person assessment of an unreliable narrator, weighing comorbidities and patient preferences, and signing prescriptions and admission decisions under personal license. AI is already eating the documentation layer — notes, coding, prior-auth letters, literature summaries, differential generation, guideline lookup — which changes the workday more than it changes the headcount. Ambulatory internists with heavy panel management face more AI-driven triage pressure than hospitalists doing bedside acute care.
Headcount grew steadily across the period.
Median pay $201,590 → $256,560 +1.8% in real terms
This line is counted by the Bureau of Labor Statistics — the one figure on this page that isn't a judgement of ours. Headcount moves on demand, offshoring, demographics and the business cycle, and automation is one term among several, often not the loudest.
So a falling line is not evidence that AI did it, and a rising one is not evidence that it won't. Both happen in this register: some occupations resist automation and shrink anyway, others are highly automatable and keep growing. The marked year is 2020.
BLS projection, 2024–2034
+3.3% 73,200 → 75,600 on the projections basis
Hard to automate, and growing
The work resists current AI and the BLS projects +3.3% more of these jobs by 2034. Note that safe does not mean well paid — several of the fastest-growing resistant occupations are among the lowest paid on the register.
Different clocks. The score is what current AI could do to this work today. The projection is how many of these jobs will exist in 2034. Everything between the two — how fast employers actually adopt, whether demand grows in the meantime — is why they can point opposite ways without either being wrong.
~2,100 openings a year on average, including replacing people who leave.
DoctorInternistPhysicianOncologistHematologistPulmonologistTrauma DoctorRheumatologistEndocrinologistGeneral InternistGastroenterologistMedical OncologistOncology PhysicianMedical Doctor (MD)Hematology PhysicianMedical HematologistHospitalist PhysicianPrimary Care PhysicianRheumatology PhysicianEndocrinology PhysicianInternal Medicine DoctorGastroenterology PhysicianGeneral Internal Medicine DoctorGeneral Internal Medicine Physician
Holding it up: liability shield . Weakest point: task resistance .
Mixed — a routine tier and a judgment tier Differential generation, guideline lookup, med-reconciliation checks and the entire note-and-coding layer are already being drafted by machine, but the parts that fill the actual visit — palpating an abdomen, deciding whether the dyspneic 78-year-old on five drugs goes home or gets admitted, re-interviewing a patient whose history changes on the third telling — hold the score at 13 rather than 16, because the documentation half of the job is genuinely going.
Hands-on in uncontrolled environments An internist is physically at the bedside auscultating, doing rectal and breast exams, draining an effusion or placing a line on the floor, and rounding through rooms with contagious patients — real hands-on work, but in a hospital or clinic with equipment, nursing support and lighting, which is why this sits at 14 rather than in the linesman-and-roofer high teens.
Licensed human required and personally liable State medical licensure plus DEA registration means the internist's own signature is on every controlled-substance script, admission order, DNR discussion and discharge summary, and a malpractice claim names that physician personally — near the ceiling, short of 20 only because hospital employment and institutional coverage absorb some exposure that an independent practitioner carries alone.
Exists to be accountable for ambiguous calls Deciding how aggressively to treat a frail patient with CKD, heart failure and dementia — where guidelines conflict, the family disagrees, and code status is unresolved — is a call with no protocol and mortality on the other side of it, which is what an 18 looks like; not 20 because much of the panel is protocolised hypertension, diabetes and screening.
The verdict above describes this occupation as a whole. Almost nobody does the typical version of a job — tick what's actually in your week and see how your own mix sits.
Your task mix speaks to task resistance (13/20 here) — how much of the day's work current AI already does. That is the dimension the boxes above are about.
It cannot move the other three. Liability shield (19/20) is whether the law requires a licensed human to sign. Trust premium (18/20) is whether buyers specifically pay for a person. Judgment and accountability (18/20) is whether the role exists to own consequential calls. Those are facts about the occupation's standing, not about which tasks are in your week — a paralegal who does only trial exhibits still holds no licence. Together they are 55 of this occupation's 82 points (67%).
Embodiment (14/20) is also a property of the work rather than the worker, but we don't tag individual tasks as physical or not, so the picker can't tell you anything about it. That's a limit of this tool, not a claim.
Did we get the list right? Tell us what's missing — the tasks are written from the outside, and you're reading this from the inside.
The moves above are yours to make. This is the other half: what would have to change in the world for the occupation itself to score higher. None of it is in any one person's gift, but it is where the floor actually comes from. Scores here are not a one-way ratchet. Only two of the five dimensions — task resistance and embodiment — track what machines can do. The other three track law, what buyers will pay for, and who is answerable, and those move in both directions, often in response to the same pressure AI creates. If every lever below landed, this occupation would score around 90/100, still SAFE.
Task-mix shift: documentation, coding, prior-auth and guideline lookup are the routine tier and are already being absorbed (Abridge, Nuance DAX). If they go entirely, the residual day is undifferentiated-complaint workup, multi-morbidity trade-offs, goals-of-care conversations and admit/discharge calls — the tier current models cannot close. Watch for panel sizes rising while visit length for complex patients also rises.
State medical board or legislative rules explicitly barring autonomous AI from diagnosis/prescribing and requiring a named licensed physician of record to review and sign AI-generated recommendations — as in the pattern of California AB 3030 (2024, AI-generated patient communications must be disclosed and reviewable) and Texas/Illinois bills restricting AI clinical decision-making without physician sign-off. Also DEA/state rules keeping controlled-substance prescribing to a licensed prescriber personally.
Payer and hospital adoption of formal AI-override documentation requirements — e.g. a requirement that the attending record a rationale whenever an AI sepsis/deterioration alert or utilization-review recommendation is not followed, making the physician the explicit owner of the deviation. Already appearing in Epic deterioration-index workflows and CMS conditions-of-participation discussion.
Growth of direct primary care and concierge internal medicine, where the sold product is named-physician continuity; and payer/employer contracts that price a human continuity relationship separately from AI-triaged 'virtual first' tiers. Also malpractice insurers offering premium credits only where a physician conducted the in-person encounter.
The limit. Already at 82; liability_shield, trust_premium and judgment_accountability are near their practical maxima, so realistic headroom is a few points from task-mix shift, not a category change. The real risk here is not displacement but scope reallocation — NPs/PAs plus AI absorbing the ambulatory panel-management tier, shrinking headcount without lowering any dimension score.
| New York-Newark-Jersey City, NY-NJ | 10,380 | $105,960 -59% |
| Baltimore-Columbia-Towson, MD | 2,100 | $322,890 +26% |
| Phoenix-Mesa-Chandler, AZ | 1,960 | $259,320 +1% |
| Minneapolis-St. Paul-Bloomington, MN-WI | 1,440 | $330,200 +29% |
| Washington-Arlington-Alexandria, DC-VA-MD-WV | 1,370 | $223,250 -13% |
| Dallas-Fort Worth-Arlington, TX | 1,350 | $276,520 +8% |
| Detroit-Warren-Dearborn, MI | 1,230 | $149,990 -42% |
| Boston-Cambridge-Newton, MA-NH | 1,220 | $243,230 -5% |
| South Bend-Mishawaka, IN-MI | 140 | $485,900 +89% |
| Lafayette-West Lafayette, IN | 120 | $416,860 +62% |
| Jackson, TN | 160 | $412,460 +61% |
We have no reported case of a named organisation automating this occupation. Not one deployment, not one announcement.
That is worth saying out loud next to a score of 82. The verdict above is about what the work exposes — what current AI could do to these tasks. It is not a claim that anyone has done it. For this occupation those two things have come apart completely: the capability argument is on this page, and the evidence column is empty.
Has AI actually changed your work? One tap, anonymous, and the running tally is public. Nothing else is asked of you.
Rather than check back: get the digest and we'll tell you what changed — or watch a single occupation from its own page.