SAFE
AI already matches or beats cardiologists on narrow read tasks — ECG rhythm classification, echo ejection-fraction measurement, coronary calcium scoring on CT — and those reads are a real slice of the workday, especially for non-invasive practices. But the core of the job is physically inseparable from the patient: auscultation and volume-status exams, catheterizations and stent placement, pacemaker and ICD implants, stress test supervision, and titrating heart failure regimens in patients with kidney disease and five other prescribers. Every diagnosis, prescription, and procedure requires a licensed physician's signature and carries personal malpractice exposure, and patients with a life-threatening chronic condition pay specifically for a named human who will take the call at 2am.
Most of this decline happened after 2021 — it is not the pandemic dip.
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.
BLS projection, 2024–2034
+4.1% 19,400 → 20,200 on the projections basis
Hard to automate, and growing
The work resists current AI and the BLS projects +4.1% 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.
~600 openings a year on average, including replacing people who leave.
PhysicianCardiologistCardiac SpecialistMedical Doctor (MD)Cardiology PhysicianGeneral CardiologistInvasive CardiologistPediatric CardiologistNoninvasive CardiologistNon-Invasive CardiologistHeart Failure CardiologistInterventional CardiologistElectrophysiology CardiologistAPP (Advanced Practice Provider)Cardiology Non-Invasive PhysicianInterventional Cardiology PhysicianDO Physician (Doctor of Osteopathic Medicine Physician)
Holding it up: liability shield . Weakest point: task resistance .
Tasks largely resist digitisation At 14 rather than 18, the diagnostic reading layer — rhythm strips, EF quantification, calcium scores, carotid Dopplers — is genuinely being taken over by algorithms and is a measurable share of a non-invasive cardiologist's billable day, but femoral and radial access, wire manipulation through a tortuous RCA, device interrogation with a patient who has both AF and CKD stage 4, and the actual conversation about whether a 84-year-old gets TAVR are not tasks a model can complete.
Hands-on in uncontrolled environments 16 reflects that the interventional and EP work happens in a cath lab with fluoroscopy, sterile drapes, and a patient whose blood pressure can drop in seconds — sheath insertion, lead placement in the coronary sinus, pericardiocentesis — while acknowledging a meaningful fraction of cardiology is clinic-based echo review and med titration that happens at a desk, which is what keeps this off 19.
Licensed human required and personally liable 20 is the ceiling and correctly so: every stent, every anticoagulation decision, every stress test read carries a state medical licence plus ABIM cardiovascular disease board certification, hospital credentialing and privileging specific to each procedure, and personal named exposure in malpractice suits where missed dissection or delayed cath are among the highest-payout claim categories in medicine.
Exists to be accountable for ambiguous calls 18 fits calls that guidelines explicitly leave open: whether ambiguous chest pain with a troponin of 0.06 goes to cath, whether to escalate to LVAD or shift to palliative, whether to stop the DOAC before surgery in a patient with a mechanical valve — decisions made with incomplete data, under time pressure, where either choice can kill the patient and the cardiologist owns the outcome.
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 (14/20 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 (20/20) is whether the law requires a licensed human to sign. Trust premium (17/20) 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 85 points (65%).
Embodiment (16/20) 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 94/100, still SAFE.
As FDA-cleared autonomous ECG/echo/CT-calcium interpretation absorbs the routine read tier (e.g. Cleerly, Us2.ai, Anumana already reimbursed under CMS NTAP/CPT codes), the residual workday concentrates on structural intervention, EP ablation, advanced heart failure and transplant listing, and multi-comorbidity titration — the tier no model closes. Task-mix shift, no new law needed.
If CMS or the Joint Commission require documented physician override rationale for AI-generated risk scores and treatment recommendations (as several state AI-in-utilization-review laws already require for payer-side denials, e.g. California SB 1120), the cardiologist formally owns the ambiguity call on the record.
Continued shift of volume toward structural/interventional work — TAVR, mitral TEER, LAA occlusion, CTO PCI, leadless pacing — as non-invasive reads are automated. Robotic PCI (CorPath GRX) has not displaced the operator and its vendor exited the market, which is itself evidence the manual ceiling is high.
If ACC/AHA appropriate-use criteria or hospital credentialing require a named attending cardiologist of record be disclosed to the patient for any AI-assisted diagnostic pathway — analogous to radiology's push against unattributed autonomous reads — the human name becomes a purchased feature rather than an implicit one.
The limit. Liability shield is already 20 and cannot rise; state medical practice acts plus malpractice exposure are the binding constraint and there is no headroom above them. Overall score is near the register ceiling — the realistic gains are a few points of task-mix concentration, not a structural change in exposure.
| New York-Newark-Jersey City, NY-NJ | 2,050 | $492,150 -1% |
| Dallas-Fort Worth-Arlington, TX | 940 | $403,340 -19% |
| Atlanta-Sandy Springs-Roswell, GA | 910 | $239,200 -52% |
| Houston-Pasadena-The Woodlands, TX | 460 | $187,550 -62% |
| Boston-Cambridge-Newton, MA-NH | 440 | — |
| Louisville/Jefferson County, KY-IN | 310 | $239,200 -52% |
| Philadelphia-Camden-Wilmington, PA-NJ-DE-MD | 290 | — |
| Salt Lake City-Murray, UT | 230 | $239,200 -52% |
| Seattle-Tacoma-Bellevue, WA | 130 | $662,760 +34% |
| Chattanooga, TN-GA | 40 | $610,080 +23% |
| Omaha, NE-IA | 110 | $598,480 +21% |
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 85. 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.