People get sick. People get hurt. Bodies fail — slowly, suddenly, predictably, without warning. Medicine exists because that is intolerable in a specific, visceral way: a person in pain is intolerable to another person standing in front of them who might be able to do something about it. Before it is a profession, before it is a system of insurance codes and pharmaceutical supply chains and credentialing exams, medicine is the organized human response to suffering. The structural pull at the field level is Care. Whatever a particular medical job looks like day-to-day, the field exists because the move from suffering to relief is something humans need done well by people who have spent years learning how.
But medicine is not just care; it is care structured by knowledge. The history of the field is the history of figuring out what is actually wrong — why this person is suffering, what is happening inside them that they cannot see, what will help and what will make it worse. Revelation runs underneath the daily work of most clinical roles: a physician looks at a patient and a set of images and a row of lab values and reads what is hidden. A medieval physician and a modern oncologist share the same pull toward the suffering person; what separates them is what they know how to uncover. Resolution sits alongside — when the body is a system and the system breaks, the pull is to make it functional again, whether that means the surgeon reducing a fracture, the physiotherapist rebuilding range of motion, or the psychiatrist restoring someone's capacity to function in their own life. And Discovery runs in the background of every clinical decision via the research and trial work that produced the protocols the clinician now follows; the field is built on a foundation of people whose job is to make the unknown known.
🦊
There's a guide here if you want one
Kitsune can talk through anything on this page — whether it might suit you, what to do next, questions this page doesn't answer. Everything here is yours to read either way.
Clinical documentation burden
Current fact
The barrier
Physicians spending 1-2 hours documentation per hour of patient care, compressing relational care
What changed
Ambient AI scribes generating clinical notes from conversations; 600+ organisations; 70% UCSF adoption; Kaiser 2.5M encounters
Behaviours involved
GP identity restored — time freed from documentation flows back to relational care
AccompanyingCaringListening
Physician-educator with time for patient education rather than charting
CaringExplaining
What this is based on
Kaiser Permanente (2.5M encounters)
UCSF (70% physician adoption)
Emory Healthcare (30.7% well-being improvement)
AHA six-system profile (April 2026)
How this could age
Very low risk on the collapse itself; moderate risk on the optimistic time-reclamation narrative.
What it does not cover
Time saved does not automatically flow to patient care. NEJM AI trial: burnout improved but documentation time reduction was 1.7% (not statistically significant) — mechanism may be cognitive load more than raw time. Institutional choices determine whether time goes to volume or depth.
Assessed June 2026
Diagnostic pattern-recognition monopoly in radiology
Near-term projection
The barrier
Every image required a trained radiologist; persistent shortages create throughput bottleneck
What changed
1,451+ FDA-authorised radiology AI devices; AI triage for urgent cases; market projected to $26.23B by 2034
Behaviours involved
Radiologist as AI-oversight + complex case specialist + clinical integrator
CoordinatingPattern-findingSolving
Radiologist value shifts to interpretation, communication, clinical collaboration
CollaboratingDiagnosingExplaining
What this is based on
Aidoc CARE1 foundation model (cleared Feb 2025)
AI triage for stroke, PE, haemorrhage deployed at major centres
How this could age
Low risk on direction; moderate risk on timeline — the reimbursement gap remains a structural brake on how fast this reaches community practice.
What it does not cover
Reimbursement gap limits adoption. Authorised ≠ routinely used. Community radiologist experience still limited AI partnership.
Assessed June 2026
Data-analysis skill barrier
Near-term projection
The barrier
Dual expertise (clinical + quantitative) required for population-level data analysis
What changed
LLMs and AI data tools lower quantitative analysis floor; bottleneck shifts to clinical question framing
Behaviours involved
Clinician-data analyst finding population patterns while centred on patient outcomes
CaringPattern-finding
Extended diagnostic instinct to population-level data analysis
DiagnosingPattern-finding
What this is based on
Clinician-data scientists at academic centres
Truveta, Flatiron Health platforms
How this could age
Low risk on direction; uncertain on speed.
What it does not cover
[Inference] Exemplars concentrated at academic centres; community practice not yet reached.
Assessed June 2026
Scale constraint on personalised medicine
Near-term projection
The barrier
Too many variables per patient for human synthesis; treatment follows population guidelines by default
What changed
AI integration of genomic, biomarker, history, and outcome data; AI-designed drugs in Phase II/III trials
Behaviours involved
Precision medicine physician using AI synthesis + clinical judgment for individualised care
CaringDiagnosingPattern-finding
What this is based on
Insilico Medicine ISM001-055 Phase IIa (positive)
MSK/MD Anderson precision oncology
Recursion-Exscientia platform
How this could age
Moderate risk — precision medicine has been promised for two decades with uneven delivery, and "every patient gets AI-personalised treatment" remains aspirational for most settings.
What it does not cover
[Inference] Concentrated at academic centres. Community practice earlier in this journey.
Assessed June 2026
Jobs that did not exist five years ago
These are real jobs that exist now and did not exist before the current wave of AI.
Familiar title, new shape
Clinical Informatics Physician / CCIO / CMIO
Physician bridging clinical practice and AI/health technology implementation. ACGME fellowship pathway mandatory from January 2026 (U.S.).
ABPM/AMIA credential change 2025; Faculty of Clinical Informatics (U.K.); Mode 1 archetype expansion · current fact
New title
AI Clinical Reviewer
Physician reviewing AI-generated clinical outputs for accuracy and safety before deployment to practice.
Emerging in health systems deploying LLM clinical decision support; Stanford-Harvard Clinical AI report 2026 · early signal
New title
Precision Medicine Clinician
Physician integrating genomic, biomarker, and AI-synthesised data into individualised treatment decisions.
Formally recognised at MSK, MD Anderson; emerging in cardiology and rare disease · current fact
Familiar title, new shape
AI-Augmented Radiologist
Radiologist whose workflow is structured around AI triage and screening assistance, focusing on complex cases and clinical integration.
Diagnostic Imaging 2026 inflection-point characterisation; deployment at academic centres · early signal
Familiar title, new shape
Clinician-Researcher (AI-Enabled)
Practising physician conducting population-level research using AI data tools without dedicated biostatistician.
Truveta/Flatiron platforms; LLM data analysis capabilities · early signal
New role titles in medicine are slower to formalise than in unregulated fields because of the credential infrastructure. The configurations are emerging faster than the titles.
Bars above the line are the parts of this work that still need a person. Bars below it are what AI can already do. Tap any column to see the actual work behind it.
high ground · holds stronglydeep water · reaches furthest
yours, by strengthAI reach, by depth
The honest read. Medicine's robustly-human profile is the deepest of any field assessed. Four of five dimensions are strong; creative synthesis is moderate and narrowing. The robustly-human work constitutes a large fraction of actual daily practice, which is the structural reason entry-level impact is minimal despite extraordinary AI capability in the information-processing layer.
Medicine has the widest theory-to-deployment gap of any field in this corpus. 1,451+ FDA-authorised AI/ML devices exist; routine clinical workflow integration remains concentrated at large academic systems. Capability is broadly available; deployment is constrained by trust, liability, regulation, and reimbursement rather than by what AI can technically do.
How AI is changing the way in
Getting in is largely unchanged, and that applies fairly evenly across the ways in.
That is everything we currently know about AI in Medicine. It shows where things are moving so you can choose which way in suits you.
People drawn to Medicine are often drawn to these. Most sit in a different part of the terrain.