Anyone can spin up charts with an AI prompt, but most dashboards end up as eye candy that tells you nothing. Before you type “make me some graphs,” answer three questions: who’s looking, what decision they’ll make, and the one key takeaway. In the example using the WHO Ambient Air Quality Database (PM2.5, PM10, NO2 measures across 7,000+ cities from 2010–2022), the audience is everyday people asking “Is my city’s air getting cleaner?” Under that, you drill into regional comparisons, trend direction, and action steps—the data shows who improved, the AI explains why (e.g., China’s Blue Sky Policy). Pick chart types that match your questions. Use line charts for trends (PM2.5 over time), bar charts for rankings (regions by average PM2.5), scatter plots for correlations (PM2.5 versus NO2). Avoid pie charts with too many slices, 3D effects, dual axes and spaghetti charts. Reference the From Data to Viz decision tree to zero in on the right form, then prompt with exact specs—colors tied to WHO severity tiers, annotated reference lines, sorted bars. Design with intention. Stick to a maximum of five meaningful colors (green ≤5 μg/m³, blue 5–15, orange 15–35, red >35). Strip out borders, shadows and clutter. Label directly and start bar axes at zero. Lay out KPI cards at the top, primary charts in the top-left, supporting views underneath, and a detail table at the bottom. Add context markers like a dashed line at the 5 μg/m³ WHO guideline or a vertical line marking the 2020 COVID-19 lockdowns. Finally, craft a narrative arc—establish normal levels, highlight the pollution problem (93% of cities exceed safe limits), reveal the biggest improvements, then point to next steps for your own city.