A city communications office wants first-pass social captions for a recycling campaign before staff edit and post. Which workplace AI use fits that need?

For a city communications office producing first-pass social captions for a recycling campaign, the workplace AI fit is a generative assistant that drafts caption options staff must review before publishing. Generative models excel at producing multiple tone and length variants from campaign briefs, brand phrases, and known program facts, which compresses the blank-page stage for municipal communicators. Official city channels still require human ownership of accuracy, equity of messaging, accessibility language, and compliance with records and social-media policy, so the draft-then-review pattern matches how public-sector content teams already work.

An autonomous agent that directly auto-publishes generated captions to social channels posts directly to public channels without staff review, but auto-publishing skips the required draft-then-review step and does not provide first-pass drafts for staff to edit before publishing. A predictive machine learning model that scores draft captions for expected engagement ranks existing drafts by likely performance, but scoring drafts does not draft new caption options for the current recycling campaign. An extractive text model that condenses campaign guidelines into proper caption lengths shortens source guidelines to fit length limits, but condensing guidelines is not the same as drafting new caption language for this specific campaign.

Exam caveat: Domain coverage of everyday workplace AI emphasizes assistive drafting and summarization under human review, not autonomous publication on government channels. Do not equate "generative assistant" with "auto-post to the city’s account."

Operational check: Require a human publish gate, maintain an approved fact sheet the model must not contradict, log who approved each post, and spot-check drafts for tone consistency across neighborhoods before go-live.