Liquid Death's media chief is running 17 Omega iterations in three weeks, a daily push/ease Slack, and Amazon analysis that collapsed from two weeks to two hours. Prime Day prep is now a day, and last year without the intelligence layer they "missed a ton."
Why it matters: This is the in-house bar for an AI-native commerce team: daily decisions, two people on a sizeable budget, no spreadsheet ritual.
At AWNY Meta started testing conversational campaign create, targeting edits, and budget changes inside the business assistant, with drafts, previews, and a human approval gate. Third-party agents now land work as Ads Manager drafts through an expanded ads MCP.
Why it matters: Treat this as a permissions review, not a feature tour. Decide which write actions an assistant is allowed before Q4 volume hits.
Docket's Gary Amaral told Humans of Martech that an 8X homepage-hero test moved Aura from inference to declared intent, and he now mines unanswered agent questions via MCP for the week's content plan. An Agent Qualified Lead is "replacing the vendor in Q1," not a 73-point score.
Why it matters: If a MAP still pays out on scored MQLs, the first agent experiment is rewriting the qualification sentence, not buying another scoring model.