Diagnostic Medical Sonographers: EXPOSED

EXPOSED

The hard part of sonography is acquisition, not interpretation: angling a transducer through body habitus, holding an obese patient's liver in view during a breath-hold, chasing a fetal heart in a moving 28-week gestation. AI is genuinely good at reading a saved clip and increasingly good at auto-measuring EF, IMT, and biometry, but it cannot find the window or coax a scared patient into position — and Caption-style AI acquisition guidance so far assists the operator rather than replaces them. Registry credentialing (ARDMS/ARRT) plus radiologist sign-off keeps a human in the loop; the exposed slice is the protocol-driven measurement and preliminary-report writing.

10-year outlook: Demand keeps growing with imaging volume; expect AI to absorb measurement and preliminary reporting while sonographers' value concentrates in difficult scans, procedures, and patient handling.

66/100 risk resistance · ±3 on re-score

Score — 66/100 resistance

Five dimensions, 0–20 each, summed. Higher means more protected. The arithmetic is shown so you can check it: 14 + 18 + 9 + 12 + 13 = 66.

Task resistance 14/20

Tasks largely resist digitisation. How much of the day-to-day work current AI already does at usable quality.

Embodiment 18/20

Hands-on in uncontrolled environments. Physical work in unpredictable environments. Robotics lags language models badly.

Liability shield 9/20

Certification preferred, not legally required. Is a licensed human legally required to sign, and personally liable if it goes wrong?

Trust premium 12/20

Some relationship component. Do buyers specifically pay for a human — presence, care, accountability?

Judgment & accountability 13/20

Meaningful discretion. Does the role own consequential calls made with incomplete information?

This verdict was revised after a second scoring. Two independent runs of the same rubric returned 67/100 and 62/100. The score above is their average, and the change of label held whichever way the rounding went — so it reflects the two runs genuinely disagreeing with the first published number, not a rounding artefact.

This score sits on a verdict boundary. At 66/100 it is one point from SAFE. Re-scoring moves results by a point or two, so here the score is more informative than the label.

Confidence: high · reviewed 2026-08-11 · how scoring works

Tasks already automatable

auto-measuring standard biometry and cardiac calculations
drafting preliminary findings/tech worksheets
flagging obvious pathology on saved clips
protocol selection and image labeling
scheduling and exam documentation

What survives

transducer manipulation on difficult body habitus
real-time window hunting and image optimization
patient positioning, breath-hold coaching, calming anxious patients
recognizing an urgent finding mid-scan and extending the exam
guiding needles in ultrasound-assisted procedures

Active moats: embodiment, trust, judgment

How to future-proof this job

Stack specialty registries where scanning skill is scarcest — RVT (vascular), pediatric echo, MSK, or fetal echo — so your value sits in acquisition difficulty, not measurement volume.
Move into interventional/procedural ultrasound: US-guided biopsies, vascular access, and pain injections, where you're at the table with the physician.
Own the AI QA loop: become the tech who validates auto-measurement output, rejects bad auto-traces, and documents why — the person the radiologist trusts on borderline studies.
Take on point-of-care ultrasound training for ED/ICU clinicians and residents; teaching probe skills is a task no model performs.

Field report — do you do this job?

Has AI actually changed your work?

Self-reported and unverified — a sentiment signal, not a survey. One response per person per occupation; you can change your answer.

From people who do this job

Nobody has filed one yet. If you do this work, you know things the rubric can't see.