Level 2 taught one professional to do one agency task well with AI, under supervision and with a record. Level 3 is about the operation: the same work run by many people across many clients, where the account director can no longer read every caption, email, landing page and report before it ships. It is written for account directors, creative and content directors, heads of social, media and PR, studio and production managers, marketing operations staff, and in-house marketing managers who are accountable for work they cannot personally read.
Modules one to four build the operating frame. Module one shows how volume and handoffs turn a correction rate into a monthly count, how busy reviewers stop looking until a wrong draft becomes a published post, a sent email or a changed bid, why shared tools make writers sound alike, and why an owner with stop authority and a measured baseline come before any redesign. Module two maps a content, media or PR workflow as it actually runs, with client approval loops, freelancer and vendor handoffs, the steps where the law already expects a control, and a step-level baseline of time, rework and errors. Module three builds a four-tier grid from consequence, reversibility, claim type and channel, anchored in the FTC's substantiation factors, and attaches review depth, records and re-tiering triggers to each tier. Module four writes a procedure a reviewer can enforce, designs its review step against automation bias, turns endorser training and monitoring into procedure, and tests it on a new team member.
Modules five to seven govern what the tools can reach. Module five chooses the brand guidelines, approved claims with their substantiation, reviewed disclosures and past work that enter a client corpus, tags each by client, claim, evidence, expiry, jurisdiction and source, records provenance, and controls poisoned files with owners, review and removal. Module six shows how retrieval still goes wrong, through set-aside orders, vacated rules, expired claims, the wrong client or jurisdiction and poisoned web pages, and answers with expiry tags, currency checks, deterministic limits and logging. Module seven keeps each client's briefs, customer lists, results and unreleased news out of every other client's work, with least privilege, phishing-resistant sign-in, a record of vendor data terms that change, contract checks before data reaches ad tech, and access logs that do not copy the content.
Modules eight to eleven run the work at scale. Module eight keeps a metric register that says what each figure means and what any industry audit covers, makes holdouts routine, reports search changes across a portfolio honestly, and reviews every figure in a pitch or a description of the agency's own AI. Module nine operates endorser training and monitoring across a creator roster as the agency's own control, watches review sources for the six practices the federal reviews rule prohibits, keeps AI-drafted replies under the brand's disclosed identity, and reports the enforcement record accurately. Module ten records source type, oversight and ingredients for generated assets, explains what a credential does and does not prove, builds the New York synthetic performer check into the studio line, and records human contributions for registration. Module eleven runs outreach for many clients with consent records that name the seller and the number, one revocation log on the ten-business-day clock, calling windows and do-not-call lists, and sampling of sends and calls.
Modules twelve to fourteen keep the operation honest over time. Module twelve builds a watchlist of official sources, records proposals and pre-publication reports as open items, tracks dated state changes in Connecticut, California and Colorado, treats platform policies as contract terms, and records each check. Module thirteen draws quality samples by risk tier, applies an eight-code error taxonomy with omissions taken seriously, sets thresholds in advance, designs against reviewer fatigue, and explains why a detector score is never a quality method. Module fourteen agrees escalation paths and stop conditions with the client before launch, reads what the AI productivity studies did and did not measure, chooses measures that reveal failure and resist gaming, and prepares the capstone.
Statements of authority say whom they bind and when they were checked; no state's rule is presented as national, voluntary frameworks are never taught as law, and proposed rules are taught as proposals. The level ships with a printable workbook and nine templates, a final examination that draws forty scenario questions from a reviewed bank, and a capstone Workflow Redesign of one agency workflow built on fictional clients, campaigns and figures.
Everything here is professional education. It is not legal, privacy, advertising-compliance or intellectual property advice, and it does not replace counsel, a platform's own policies, which are contract terms, or the law that applies to a campaign. Completing the level earns an independent educational certificate issued by AI Coalition Network with a public verification page. It is not a licence, a platform certification, a professional designation or a regulatory approval, and it carries no professional education hours.