Level 1 taught how AI systems work and fail with practice and patient information, and which duties do not move when a tool arrives. Level 2 is where that becomes practice. It is written for patient access, referral and prior authorization staff, medical assistants, billing and coding staff, practice managers, privacy and security officials, and the clinicians who review AI drafts and sign AI-captured notes. Every module takes one task that fills a practice day and builds a supervised, repeatable, defensible way to do it, with the reviewer's step shown and the reviewer's role named.
Modules one to three build the instrument. Module one sets the operating model: triaging a task on consequence, data class, verifiability and who is qualified to own it; a workflow from source material to an output a named person has reviewed; reading a vendor's accuracy claims for who measured what, on which population; and keeping the AI use record while the work happens. Module two builds the reusable practice and payer prompt: a context header naming payer type, state, date of service and plan year, supplied policy text instead of recall, constraints that keep the output administrative, and a design that treats instruction-like text inside faxes and referrals as data rather than commands. Module three turns shared prompts into a library with named owners, data classes and versions, retesting tied to real change events such as the fiscal-year coding guidelines and CMS manual revisions, and retirement.
Modules four to six are production work. Module four assembles the verified patient fact file a draft may rely on, writes closed-file drafting instructions, works through invented facts and omissions, and settles who must read a draft before it goes out, with anything pertaining to clinical information belonging to the licensed clinician. Module five extracts referral packets, faxes and outside records into a fixed field list the practice owns, verifies every field against its source, flags what the checklist misses, and teaches the recognition of substance use disorder records in an inbound packet. Module six is the signed note: what a clinician checks before signature, that it documents the care given and records nothing that did not happen, and who may prepare and who must finalise it.
Modules seven and eight are research and verification. Module seven explains what a grounded citation over payer policy proves and what it does not, the five ways a linked answer still misstates a rule, and how a practice protects the documents its research tools read. Module eight is the five-step Rule and Status Verification Protocol: does the instrument exist, does the quoted text appear where it is cited, is it final or proposed or vacated, whom does it bind and from what date, and how the check is recorded.
Modules nine and ten cover the money and the message. Module nine works coding support without ever choosing a code: documentation support and the coder's judgement when a tool suggests a code or a level of service, provider queries and the uncertain outpatient diagnosis, and what the law requires once a wrong claim has gone out. Module ten turns marketing copy into a list of claims the practice can substantiate, sets the rules for describing the practice's own AI honestly, applies the Consumer Reviews Rule to AI-written testimonials, and separates advertising law from the HIPAA marketing authorization.
Modules eleven to thirteen widen to the front door, the payer and the data. Module eleven draws the administrative boundary of an intake assistant, builds category-based escalation that never asks staff or software to judge urgency, and covers disclosure and website tracking. Module twelve assembles prior authorization requests and appeals from signed documentation, states what a Medicare Advantage plan must base a decision on, and keeps straight which timeframes and denial-reason duties bind the plan rather than the practice. Module thirteen classifies practice material before a tool sees it, separates treatment, payment and operations from uses needing an authorization, and works business associate terms, subcontractors and tracking vendors.
Module fourteen closes the loop where the work is proved: the use record tested against the questions a practice manager, a privacy official or an investigator would ask months later, the first hour after an AI incident, the breach presumption and its four-factor assessment through to notice, and what follows: an identified overpayment, voluntary reporting and a dated change to how the practice works.
Statements of authority are labelled throughout as law or rule, professional standard or guidance, best practice, emerging practice, or an AI Coalition Network recommendation, with whom each binds and the date it was checked. What binds the practice directly is kept apart from what binds a manufacturer, a certified health IT developer or a health plan, and one state's statute is never presented as a national rule. The level ships with a printable workbook and twelve templates, a final examination that draws forty scenario questions from a reviewed bank, and a capstone Practice Task Workflow Portfolio built on fictional practices and patients.
Everything here is professional education. It is not clinical training, it does not teach diagnosis, triage, prescribing, treatment decisions or how to select a code, and it does not replace the judgment of any licensed clinician or coding professional, legal counsel, privacy or compliance review, payer requirements, or the law of a practice's own state. 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 or a compliance certification, it carries no professional education hours, and it satisfies no licensing, credentialing or payer training requirement.