To create Excel charts with AI, first define the question, metric, and rows to include. Ask the assistant for a chart specification and its source range, then compare the finished chart with the table. Attractive formatting cannot fix a missing month, a duplicated total, or a claim that the numbers do not support.
This guide uses fictional monthly order counts and an original chart illustration. It is an editorial walkthrough, not a live test of an AI product. File access, workbook editing, and chart creation vary by assistant. If your tool returns only instructions, you can build the reviewed chart in Excel yourself.
1. Choose the Question Before the Chart
For this example, ask: How many completed orders were recorded in each full month of the first quarter of 2026? The metric is a count of orders, not revenue, customers, or daily productivity. Each order is counted once in its completion month under the sample’s rule.
Use a simple bar or column chart to compare these three totals. Keep the months in calendar order. For a longer time series where the main question concerns the pattern over time, consider a line chart. Avoid a pie chart for this question: the reader needs month-to-month comparisons, not just each month’s share of a combined total.
Microsoft’s chart-type guide describes the options. AI can suggest a format, but ask it to explain how that format answers your question. Do not add a second metric or axis simply to make the result look more sophisticated.
2. Separate Complete Periods From Partial Data
The fictional source contains these four reporting rows. The owner confirms January through March are complete; April contains only the first five days. These labels and scope notes are part of the data.
| Period in 2026 | Completed orders | Coverage |
|---|---|---|
| January | 120 | Full month |
| February | 150 | Full month |
| March | 135 | Full month |
| April 1–5 | 30 | Partial month |
The chart for the stated question includes only January, February, and March. Retain April in the source with an exclusion note; do not delete the evidence. Do not replace the partial 30 with an estimated full-month value unless a separate forecasting task has been agreed and clearly labeled.
If you start from transaction-level files, check unique order IDs and completion dates before calculating monthly counts. A repeated monthly summary row is not an additional month’s activity. Preserve missing values as unknown; a blank count does not establish zero orders. Use the Excel file-combination guide and spreadsheet cleanup workflow to resolve source issues first.
3. Ask AI for the Chart and Its Evidence
Provide the checked table and definitions together. Share only material you are permitted to give the assistant. You can use this prompt with the fictional sample:
Create a chart specification for completed orders in January–March 2026 using the table below. Include only the three full months: January 120, February 150, March 135. Keep April 1–5, 30 orders, in the source but exclude it from this full-month comparison and disclose why. Use one series and a simple 2-D bar or column chart, with months in calendar order, a zero baseline, units of number of orders, and exact value labels. Do not plot a grand-total row, fill missing values, smooth the data, or forecast April. Return the exact chart data, source range if available, chart title, inclusion notes, and a two-sentence interpretation. Separate observed changes from unsupported explanations. If you cannot edit the workbook, give Excel steps instead of claiming to have created a file.
Check that the returned data still has exactly three records and values 120, 150, and 135. If the assistant creates a workbook, open the actual file and inspect the chart’s data selection. A preview image is not evidence that the workbook contains an editable chart linked to the correct cells.
4. Build the Chart in Excel and Inspect Its Scale
For a small manual example, put Month in A1 and Completed_orders in B1, then enter January–March and their counts in rows 2–4 of a sheet named ChartData. Keep the original four-row source on another sheet. Select ChartData!A1:B4, then use Insert → Recommended Charts or the chart-type menu to choose a simple 2-D bar or column chart. Microsoft documents the process in Create a chart from start to finish; menu details differ across Excel versions and platforms.
Give the chart the title “Completed orders, January–March 2026.” Label the numeric axis as number of orders, show all three values, and check the source range using the chart’s data-selection controls. For a horizontal bar chart, verify January is at the top, followed by February and March. Keep the actual time order even if a default layout reverses it.
Completed orders, January–March 2026
Fictional data · Number of orders · Full calendar months
This illustration uses a shared 0–200 scale. The bar lengths represent 120, 150, and 135 on that scale. The Office for National Statistics’ axis guidance recommends a zero baseline for bar charts because the lengths encode values.
If the baseline were cut to 100, the visible portions for January and February would be 20 and 50. February would appear 2.5 times as long even though 150 is only 1.25 times 120. A line chart can sometimes use a narrower, clearly labeled scale; that is a different visual encoding. Do not transfer that choice uncritically to bars.
5. Check What the Chart Actually Supports
The three included months total 405 completed orders. The movements are 120 → 150 → 135, so the counts rose and then fell. These statements can be checked directly:
- February is 30 orders above January: (150 − 120) ÷ 120 = 25%.
- March is 15 orders below February: (135 − 150) ÷ 150 = −10%.
- March is 15 orders above January: (135 − 120) ÷ 120 = 12.5%.
With the ChartData layout above, =SUM(B2:B4) returns 405, and =(B3-B2)/B2 returns 0.25. Format the latter as a percentage to display 25%. Check the denominator each time. Our percentage-change guide explains related interpretation errors.
Reject an AI caption saying “orders grew steadily every month”: March fell from February. Reject “April demand collapsed by 77.8%”: it compares a five-day count with a full month. Also reject “the campaign caused February’s growth”: the table contains no evidence about a campaign or causation.
These are calendar-month totals, not equal-length exposure periods or orders per day. They answer the stated volume question. To compare daily rates, you would need an explicit denominator, such as calendar days or confirmed operating days, and a separately labeled calculation. Three totals alone also do not establish a reliable forecast.
6. Keep the Chart, Data, and Caption Together
A supported caption for the illustration is: “Completed orders rose from 120 in January to 150 in February, then fell to 135 in March. April 1–5 is excluded because it is an incomplete month; these figures do not establish the reason for the changes.”
Before sharing, check the source range, filters, series names, month order, numeric axis, and exact labels. Confirm missing periods have not become zeros and a total row has not become a fourth bar. At the size readers will see, keep the title, values, and exclusion note readable. Include the underlying table or an accessible equivalent so the meaning does not depend on color alone.
Save the original data, reviewed ChartData sheet, chart, and inclusion notes together. When refreshing, verify whether April is now complete before adding it. For help choosing an assistant, see the AI data-analysis guide.
To present the checked results to a reader, continue with our one-page Excel report guide. It uses the same monthly-order example to connect the main finding, supporting figures, limitations, and a proposed next check.
Written September 14, 2026. Official documentation checked on that date. All order counts and the chart are fictional editorial examples; arithmetic was independently checked.