Level 3 taught a team to run AI-assisted accounting work as an operation. Level 4 asks you to take responsibility for it across a firm or a finance function, and to defend what you decided. The work of this level is decision-grade: every module ends with something a partner group, a peer reviewer, an insurer, a client or a regulator could read.
The first three modules place accountability and record it. Module one explains why a partnership spreads responsibility for firm-wide technology until nobody holds it, separates responsibility, accountability and authority, names the output owner for each kind of work, sets out what Circular 230 section 10.36 asks of those with principal authority for a tax practice, and places AI inside the firm's system of quality management without pretending either quality standard names it. Module two writes the governance programme: sized by exposure rather than headcount, with an inventory of uses rather than products, an approval pathway fast enough that people use it, decision rights, and stop authority with written restart conditions. Module three builds the decision record that lets a successor partner, peer reviewer or inspector understand a decision nobody explained to them, and keeps it alive through evaluations, partner departures and rule changes.
Module four maps the obligations. It separates the duties that come from state accountancy law, from membership of a professional body and from practice before the IRS; assigns each service line to the standards that follow its entity type and to the data rules that reach it; works through whether an AI-specific law could reach a firm or its clients, by a method that records facts rather than assuming coverage; and grades every line by how well the firm actually knows it.
Modules five to seven cover what the firm buys. Module five assesses a supplier before client data reaches it, across data handling, security, testing evidence and change management, reads a service organisation report for its type, categories, system and period, and tells an evidenced answer from an evasion. Module six reads evidence claims at their real size: what was measured, by whom, on what data and with which model, why examination and benchmark scores are not client-work results, and how to design the firm's own acceptance test before calling anything validated. Module seven writes the decision into the contract: data use and training on inputs, security terms and subprocessors, notice of changed terms and changed models, incident notice fast enough for the firm's own duties, and deletion, sunset, exit and acquisition.
Modules eight and nine handle security and fraud. Module eight places AI systems inside the written information security programme rather than beside it, with the inventory, the risk assessment, the safeguards applied where AI changes their reach, agent limits, and an incident plan that preserves prompts and logs before containment deletes them. Module nine assumes a voice, a caller identifier or an email can be faked, and builds procedures that do not depend on spotting one: out-of-band verification of people and payment instructions, phishing-resistant authentication for the accounts impersonators want, the signs of client data theft and who to report to, and training for staff and clients.
Modules ten to twelve cover the client relationship and the outside reader. Module ten reads engagement letters against the firm's actual AI use, separates the duty to inform a client from the duty to secure confidentiality, works through when tax return information needs consent, compares stricter state board rules such as California's and Texas's, and handles fees and restrictions the firm cannot meet. Module eleven takes independence and objectivity on the facts: hosting and information systems services for attest clients, business arrangements with tool providers, and a conclusion written facts first. Module twelve designs verification that is independent in person, source and method, makes AI use visible to a quality reviewer, and builds an evidence pack while describing accurately what regulators have and have not said.
Modules thirteen and fourteen turn the level into a decision. Module thirteen leads adoption among trained sceptics: which objections deserve to be kept, the two opposite reliance failures to lead against at once, and measuring adoption without making use the target. Module fourteen sequences the programme against the dated milestones that genuinely bind this firm, sorts pending rules that may never arrive, states costs, benefits and risks with an evidence grade on every line, and writes the report a partnership can vote on.
Statements of authority are labelled throughout as law or rule, professional standard, professional guidance, best practice, emerging practice, or an AI Coalition Network recommendation, with whom each binds. PCAOB and AICPA requirements are kept apart by entity type, proposals are labelled as proposals with the date checked, and no state's rule, professional body's code or voluntary framework is presented as binding on everyone. Where the authority is genuinely unsettled the level teaches a method of reasoning rather than a confident answer the sources do not support. The level ships with a printable workbook and nine templates, and the examination draws forty scenario questions from a reviewed bank.
Everything here is professional education. It is not tax, accounting, audit or legal advice, it does not replace the professional standards, SEC and IRS rules or state board requirements that govern accounting, tax and attest work, and learners must check the rules that apply to their own licence and practice. Completing the level earns an independent educational certificate issued by AI Coalition Network with a public verification page. It is not a CPA licence, carries no professional education hours, and does not satisfy any state board of accountancy requirement.