Scale. Add the next agent in the build order. Reuse the backbone.
How to read this table
This table reframes 118 AMC AI agents as a periodic system — not a checklist. The position of each agent tells you what it does, how mature it needs to be, and how it relates to the agents around it. It is purpose-built for the academic medical center tripartite mission: clinical care, education, and research.
Two axes — domain and autonomy
Columns (groups) = the WHAT. Each column is an AMC function — Patient Access, Clinical Care, Quality & Safety, Medical Education, Faculty & Learner Affairs, Research Operations, Clinical Trials, Pharmacy, Patient Experience, Compliance, and so on. Read down a column to see how a function progresses from informational chatbots to autonomous, governed systems.
Rows (periods) = the autonomy & risk profile. Period 1–2 agents answer questions. Period 3–4 agents complete transactions and run workflows. Period 5 agents make recommendations with evidence. Period 6 agents coordinate across the tripartite mission. Period 7 agents act with significant autonomy and require the strongest governance.
The two special rows
Lanthanides (yellow band, 57–71) — Operational Backbone. Identity, eligibility, document OCR, payments rail, master data, EHR/LMS connectors, queue management, messaging. These are the shared engines every other agent depends on. They're not user-facing, but they're load-bearing. Underinvest here and everything above gets brittle.
Actinides (rose band, 89–103) — Governance, Risk & Public Trust. HIPAA Guard, Stark/AKS, IRB Compliance, ACGME Compliance, equity impact, AI oversight, bioethics review, cyber risk, adverse-event learning. These cut across every domain — they are the institutional license to operate in academic medicine.
Where simple tools live in the table
The Period 1–2 row is reserved for cross-cutting informational helpers (login, scheduling, HR, billing, IT, patient FAQ, compliance Q&A). Domain-specific simple tools — for example, IRB policy lookup or SOAP-note guidance — are absorbed into related agents rather than given their own cells:
IRB policy access is covered by Compliance Q&A Agent (Aj) at Period 2 (HIPAA, IRB, ACGME, and institutional policy lookup), backstopped by IRB Compliance Agent (Ds) in the actinide row for continuous protocol oversight.
SOAP and clinical note assistance shows up as Clinical Documentation Agent (Bv) at Period 5 — ambient capture and note drafting — because in modern AMC workflows, real value is at the documentation tier, not at template-only Q&A.
What AMC leaders should be considering
Build the floor before the ceiling. Period 6–7 agents only work if the lanthanides underneath them are solid. Many AMCs overinvest in flagship agents (ambient documentation, sepsis alerts) and underfund the shared identity, master-data, and EHR/LMS connectors that make those flagships work at scale.
Governance runs in parallel, not sequentially. The actinides aren't a phase 3 deliverable — HIPAA, Stark/AKS, IRB, ACGME, equity, and bioethics need to stand up alongside the first transactional agent that touches PHI, learner data, or research subjects. In academic medicine, these are bright-line risks that don't tolerate "we'll get to it."
Sequence by risk, not by enthusiasm. A Period 2 informational agent (Compliance Q&A, Patient Helpline) carries low risk and ships fast — it builds organizational confidence. A Period 5–7 decision-support or autonomous agent (Sepsis Detection, Adaptive Curriculum, Research Autonomy) needs evidence, governance scaffolding, and human-in-the-loop review before launch.
Pick a portfolio, not a project. Use the Start with Your Priority section above to assemble 5–8 agents that together advance one strategic outcome — for example, clinician burnout combines Inbox Triage, Clinical Documentation, Discharge Planning, Care Network, and Care Journey, supported by Queue Management (backbone) and Experience Monitor (governance). One agent in isolation rarely moves the needle.
Align to the AMC pressure stack. Clinician burnout, documentation burden, value-based contracts, the NIH and trial economics, ACGME/LCME requirements, HIPAA exposure, ransomware targeting, equity accountability, the 340B and Stark/AKS environment, and the AI use-policy question — this table is designed to map your roadmap to those forcing functions, not against them.
Talk in symbols, not slideware. The two-letter symbols (Aa, Bm, Cz) make this a shared vocabulary across CEO, CMO, CMIO, CRO, dean, and the board. It lowers the activation energy for governance conversations — "Bm needs a Dm before it ships" beats a 30-page deck.