A polished campaign draft is not approval to publish.

Marketing teams feel AI first as speed: a campaign brief becomes a landing draft, three subject lines, and a localisation pass before lunch. The literacy gap is not “write better prompts.” It is claims, provenance, data boundaries, and a human before publish.

Here is the Friday scramble. A contractor pastes a customer quote and a competitive claim into a chat tool, asks for “on-brand” copy, then queues the result for the CMS. Brand judgment never got a turn. Neither did legal’s claim checklist. The draft looks finished — so it almost ships.

A polished campaign draft is not approval to publish. FluencyRank is workplace practice, not a brand certificate.

What Marketing & Communications should practice

• Claims and provenance — every number, superlative, and “customers say” line needs a source you can show
• Translation and localisation — meaning and tone checks, not only fluent wording
• Customer and confidential data — what never enters a prompt or third-party tool
• Verification before publishing — human approval before the CMS or ad queue
• Iteration with brand judgment — when to regenerate, when to rewrite by hand, when to stop

Anthropic’s fluency work is useful without inventing conversion statistics: discernment and iteration matter more when the artifact looks ready to publish. The Commission’s AI literacy Q&A is useful the same way — measures should fit role and context. Neither source asks you to mint a marketing AI certificate.

FluencyRank’s Marketing & Communications role path puts those habits into short workplace scenarios — campaign claims, redaction, and verify-before-publish — with organisational policy reminders in the challenge. That is decision support for champions and L&D, not a guarantee that every asset will clear legal.

Practice scores reflect engagement on the platform. They are not a comprehensive assessment of brand risk or a substitute for your approval workflow.

If your marketers already generate faster than they verify, start with a demo pack that forces the claim check — then keep humans in the publish queue.