The top of the pathway is not about using AI better. It is about teaching the rest of a medical practice to use it well, in a lunchtime in-service, an onboarding week, a classroom or a client engagement. An instructor who teaches a rule wrongly makes the error once for every learner who heard it, and in a practice that error travels from a slide into a note, a claim and a patient's record. This level is written for practice managers and trainers, compliance and privacy officials, physician champions, revenue-cycle and patient-access trainers, programme faculty, group educators and consultants.
The first three modules set the foundations. Module one describes the room you actually get — a physician between patients, a medical assistant rooming again at one, a front desk lead who touches the tool most, a biller nobody planned for, half an hour over sandwiches — keeps every adoption figure inside the base and date it came from, and names the two failures you teach against: refusing to use anything, and accepting everything unread. Module two designs a session backwards from one duty somebody performs differently afterwards, separates the four duties that differ by role, keeps the supervision line visible, and cuts the rest. Module three classifies a document before teaching it — final rule, proposal, guidance or frequently asked question — recognises vacated text still online, keeps a currency log, and runs the correction owed to a room taught wrongly.
Modules four to six are delivery. Module four faces the signature asymmetry: the work of a dozen people funnels into one clinician's signature, and each seat can put different things into a record or a claim. It teaches hearing the objection underneath the objection about workload, blame and replacement, opening in the profession's own vocabulary while labelling ethics guidance and association advocacy precisely, and taking a public challenge from the most senior clinician without needing to win it. Module five scripts demonstrations that fail on purpose on fictional material — an invented rule, superseded text, a vacated page, a wrong plan year, a planted detail — and rehearses recovery when the live run surprises you. Module six builds fictional patients, charts and claims that carry the teaching load, explains why a de-identified real record is not a classroom substitute, and runs the debrief where the learning lands.
Modules seven and eight are the hardest content to teach accurately. Module seven draws the boundary between an AI output and a licensed judgement in the three places a team touches daily — the signature on a note, the final code on a claim, a message to a patient — holds a room steady when the honest answer is that no state board rule is published, and replaces the cautionary anecdote with rehearsed verification. Module eight teaches the documentation, scribe and messaging evidence by design rather than by headline: what each trial measured, in what setting, against what comparator, and what it left unmeasured; why a productivity change proves neither upcoding nor compliance; and the evaluation gap behind all of it.
Modules nine and ten teach the material that varies. Module nine teaches permitted uses against authorization, minimum necessary as a decision made many times a day, the business associate relationship in words a new receptionist can act on, the parts that remain unsettled, and notification timing with two state statutes as labelled examples. Module ten puts the state question first, teaching recorded state AI statutes side by side on a scope, trigger, exemption and enforcer grid rather than merging them, including one whose duties never took effect; establishes payer type before any coverage discussion; explains why a Medicare supervision condition is about supervising people; and sends a room home with a lookup method.
Modules eleven to thirteen cover assessment, materials and the limits of the role. Module eleven writes items whose wrong answers are the mistakes practice staff actually make, runs performance tasks watched at a workstation, writes rubrics that describe observable work, and keeps the training record — while claiming nothing about validity, because none of it is validated and passing it qualifies nobody. Module twelve writes the one-page card a front desk lead can still use a month later, uses the accessibility guidelines as a technical reference while stating precisely whom that standard and the federal web rule bind, and teaches language access as a duty a covered practice owes patients. Module thirteen sets the limits of the front of the room: saying "I do not know" and showing the route to the primary source, redirecting the question that asks for clinical or payer advice, substantiating every claim, disclosing a vendor relationship, and stating accurately what this certificate is.
Module fourteen prepares the Teach-Back, the capstone: one behaviour worth changing, the objective written as that behaviour, every rule evidenced with its source, status and date before you stand up, the demonstration and debrief rehearsed against a clock, and a structured review taken without defending the session. The level ships with a printable workbook and nine templates, and a forty-question scenario examination.
Everything here is professional education. It is not clinical training, it does not teach diagnosis, triage, prescribing, treatment decisions or code selection, and it is not legal, privacy, security or coverage advice. Completing the level earns an independent educational certificate issued by AI Coalition Network with a public verification page. It is not a medical, nursing or other healthcare licence, a coding credential, a compliance certification or a teaching qualification, it carries no professional education hours, and it satisfies no licensing, credentialing or payer training requirement. A session taught by a holder earns no professional education hours either.