A neat chart can describe the wrong slice of a table. Copilot in Excel can create charts, and Microsoft explicitly advises reviewing its output. The practical review starts with the source cells rather than the colours or the title.
Try a six-row example
Imagine a volunteer group recording hours over six weekends: 3, 4, 0, 5, 2 and 6. The total is 20 hours. A chart that omits the zero weekend can still look convincing, yet it tells a different story about continuity.
Write down the intended question before asking for the chart: “Show hours for all six weekends, including zero.” Then compare the category labels, selected range and every plotted value with the table. This is an illustrative dataset, not a measured test of Copilot.
Read the chart as a claim
- Does the title describe the actual selection?
- Are the units hours rather than people or percentages?
- Are blank cells treated differently from recorded zeros?
- Did a filter remove rows you expected to include?
Keep the source table available alongside the chart. If the model adds a trend explanation, review that separately: correctly plotted numbers do not automatically justify a causal story. Product availability depends on the account and supported Excel setup, but these checks also work for a chart made without AI.