Launch decisions, recapped and drafted
An agent in the campaign channel that recaps what was decided, drafts the customer announcement from the thread and keeps the brand voice rules
The problem
During a launch the campaign channel fills with decisions, copy changes and status updates faster than the marketing lead can read, and other teams keep asking the same policy questions about logos, sponsorships and customer stories. The announcement gets written from memory at the last minute and each draft reinvents the house style.
How the agent works
Claude Tag is added to the campaign channel; no connection is needed for recaps and drafts, and a Google Drive or Notion connection lets it answer policy questions from the brand guidelines and name the document it used. The lead asks for a recap bounded by a time window (what got decided, what is still open, what is waiting on marketing), and when a thread settles what the announcement should say, asks in that thread for a draft with a named format, length and messaging document to match. Brand voice rules such as sentence case headlines and banned words are saved once with a 'remember for this channel' message, so every later draft in the channel follows them. A standing role can answer intake requests from the brand and process docs and route anything uncovered to the right owner with a one line summary. Claude posts the draft in the thread; a person edits it and sends it from the email or social tool, since Claude does not send anything. Claude Tag is in public beta.
Worked example
Anthropic's Claude Tag recipes for marketing teams show the exact messages: 'catch me up on this channel since Monday' for the recap, 'turn this thread into an announcement draft I can send to the customer list, keep it under 200 words and match the messaging doc linked above' for the draft, a sponsorship policy question that asks which document says so, and a channel memory entry that fixes sentence case headlines, 'sign up' as the verb and 'signup' as the noun. The same page shows a weekly campaign metrics digest, which needs a BigQuery or Snowflake connection, and lead questions against HubSpot with read scopes. The documentation advises keeping memory entries short and linking the full style guide instead of pasting it. No customer results are published.
Get a new agent use case every week
Join the Agent OS Club: every week a new use case with the exact prompts, skill files and setup steps, so you can build your own team of AI agents to run the work for you. It is how The Agentiv runs, on a team of 40 agents and growing. USD 15 a month, cancel any time.
Join the Agent OS ClubBuilding this for an organization? Learn more at theagentiv.com