50+ platform-specific questions drawn from the Databox Prompt Library, plus the framework that separates answers you can act on from answers that sound right. TL;DR Most AI questions about marketing data fail because they lack three ingredients: a time window, a named metric, and a comparison baseline — not because the AI is incapable. In Databox’s “Time to Insight” survey of 66 marketing and analytics teams, 64% said answering a basic business question typically takes one to three days. The bottleneck is almost never the tool. It is the question. Platform-specific questions produce platform-specific decisions. Cross-platform questions, like connecting GA4 traffic behavior to HubSpot pipeline outcomes to Google Ads spend, produce the decisions that actually move revenue. Most teams never get there because they stop at the per-platform view. Every question in this article comes from the Databox Prompt Library, organized by integration and ready to copy. Every prompt is built with all three ingredients already present. Introduction Monday morning. A Director of Demand Gen spots a 20% drop in MQLs from the previous week. The analyst queue is two days deep. The paid media team wants a campaign decision by noon. She screenshots her Google Ads dashboard, pastes it into ChatGPT, and asks: “Which of these campaigns should I pause?” The model answers immediately: campaign names, reasoning, and a recommendation. It sounds like an analyst. She acts on it. What she does not know: the screenshot clipped the date range. The ROAS figures are from the wrong period. The…