For physicians, healthcare executives, and researchers

Clinical AI needs physicians in the decision.

Sourced guides, a vendor question list, and close reads of new research, open to everyone. Join the network to work through it with colleagues deploying the same tools.

Join the network Open the evaluation kit

Overview

AI in medicine needs physicians. Enthusiasm for these tools runs well ahead of the evidence for them, and someone has to judge each one before a hospital depends on it. That judgment belongs with the people who know the patient.

AIMOCS helps physicians judge clinical AI on the evidence, and work it out with the healthcare executives and researchers they deploy it with.

Figure 1 — Enthusiasm and evidence, over time since the demo. The gap is where judgment is needed.

1,000+

AI-enabled medical devices authorized by the U.S. Food and Drug Administration.

Source: a peer-reviewed taxonomy across 1,016 authorizations (Singh et al., npj Digital Medicine, 2025), read against the FDA’s own device list in our tracker.

The library

113 pieces on evaluating and running clinical AI, open to everyone. Every piece lists its sources and the date it was last updated.

The kit is our clinical AI evaluation kit: 16 questions to ask a vendor, an evaluation checklist, and a governance charter, built to take into the meeting where a tool is bought.

Guides 73 Glossary 29 Statistics 8 Briefing 3

Figure 2 — Reported performance, and the point where a ward actually operates.

AUROC explained for clinicians

Evaluation · 10 min

AI scribes: what the evidence actually shows

Ambient scribes reached millions of encounters before the first randomized trial reported a result. What the peer-reviewed record actually shows.

Ambient AI · 11 min

A physician's hands annotate a printed chart with a pen beside a stethoscope in warm window light.

How to read an AI validation study

Evaluation · 10 min

Figure 3 — Four cells. Only one of them wakes anyone up.

Sensitivity, specificity, and AUROC

Glossary · 4 min

Figure 4 — The demo checklist.

The clinical AI evaluation kit

16 questions for the vendor demo, a reconciled evaluation checklist, and a governance charter.

Working document

Figure 5 — Performance by day. The schedule changes; the model does not.

Dataset shift and model decay

Evaluation · 8 min

The whole library

Teamwork

No one profession can put a model into clinical care alone. The physician knows the patient and the ward. The researcher knows how the model was built and where it fails. The executive owns the budget and the risk. Leave one of them out and the deployment fails where that person would have caught it.

The AIMOCS network is where the three work it out together. Members post a case or a question and hear from all three sides. Membership includes:

The community: post a case or a question, reply, and react.
The member directory, with each member's role, institution, and country.
The full briefing, including the part below the line.
Live events: register, and watch the replays.
Ten self-paced courses, 129 lessons, and a certificate of completion for each course you finish.

The model is not the practice of medicine. The physician is.

Figure 6 — Where the clinician goes. Not a phase of the project; a part of the system.

Figure 7 — Three professions, one deployment. Signals converge at the network and continue as one line.

Courses and the cohort

Membership includes ten self-paced courses in four tracks, from programming foundations to AI strategy and governance. Every lesson ends in a hands-on exercise, and each course you finish earns a certificate of completion.

The cohort is the live version: one small group at a time, working through the material together, by application. Members hear first when the next one opens. About the cohort

Compared with certificate programs

University certificate programs in AI for healthcare are rigorous and often excellent. Harvard, Stanford, and Johns Hopkins all run programs worth completing. A network answers a different need.

AIMOCS and three university programs, row by row. Sources follow the table.

Feature	AIMOCS	Johns Hopkins	Harvard	Stanford
Last content review	Syllabus v2026.Q3; all 129 lessons written Sep 2026	Re-taught live with each 10-week online cohort [1]	Runs as a live 3-day program [2]	Core content released 10 Aug 2023; CME renewed through 9 Aug 2029 [3a][3b]
Generative + agentic AI	Core to two of four tracks: LLMs end to end, agentic AI engineering, scribes, golden-set evals, MCP, EHR agents, and governance. A separate course covers working with Claude	Curriculum runs from LLMs to goal-directed agentic AI, plus an n8n masterclass [1]	Curriculum covers generative AI; hands-on sessions are demonstrations [2]	No agentic-AI module in the published syllabus; core CME content released Aug 2023 [3][3a]
What you build	A hands-on exercise in every one of 129 lessons, across ten courses	Guided projects + masterclass build [1]	Demonstrations only [2]	Recorded exercises + capstone [3]
What continues after	The member community, live events and replays, and the research briefing	Program ends at certificate [1]	Program ends at certificate; alumni events vary [2]	Program ends at certificate [3]

The two work together. A course teaches the foundations. The network is where your questions go once the systems are live. The full comparison covers each program cell by cell.

The library is open to everyone. The network is where you use it.

Members post a case or a question, and physicians, healthcare executives, and researchers answer. Membership also includes the member directory, the full briefing, event registration and replays, and ten self-paced courses with certificates of completion.

Join the network Start reading

One email address and a six-digit code.