AI prompts for Excel

Review this workbook for errors and risks. Work through it systematically and cite the cells you're referring to.

Check for:
1. Formulas that break the pattern of the row or column they sit in.
2. Hardcoded numbers inside formulas where a cell reference should be used.
3. Broken or suspect references (#REF!, #VALUE!, links to closed workbooks).
4. Ranges that miss rows or columns — especially SUM ranges that stop short.
5. Circular references and inconsistent rounding.
6. Totals that don't tie to their components.

For each issue: the cell, what's wrong, why it matters, and the corrected formula. Rank by materiality. Flag anything requiring my judgment rather than fixing it silently. Do not change anything yet — list findings first.

Explain how this workbook works, as if briefing someone who has to take it over.

Cover:
1. The purpose of each sheet and how they connect.
2. Which cells are inputs/assumptions vs calculations vs outputs.
3. The main calculation chain — what drives what.
4. Any unusual or fragile construction I should be careful with.
5. What would break if someone inserted a row or changed a key assumption.

Cite specific cells. Finish with the five things a new user most needs to know.

This formula isn't returning what I expect: {cell_reference}

Expected result: {what_it_should_do}
Actual result: {what_it_returns}

Explain what the formula is actually doing, step by step, why it produces this result, and give me a corrected version. If there's a simpler or more robust way to write it, show that too and say why it's better.

Reconcile {schedule_or_range} against {source_range_or_sheet}.

1. Compare the two and identify every difference.
2. For each difference: the amount, the line it sits on, and the most likely cause (timing, misposting, missing entry, duplicate, classification).
3. Tell me which single document or record would confirm each one.
4. Total the differences and confirm whether they net to the variance I'm chasing.

Cite cells. Do not adjust anything — this is a diagnostic pass.

I want to restructure this workbook: {describe_the_change}

Before making any change:
1. Tell me every formula and dependency that would be affected.
2. Flag anything that would break, and what the knock-on effects would be.
3. Propose the safest sequence to do it in.

Then wait for my confirmation before changing anything. When you do make changes, preserve existing formula logic and explain each edit with the cells involved.

Build a {schedule_type} schedule in a new sheet, following the structure and conventions already used in this workbook.

Requirements: {requirements}
Period: {period}

Match the existing formatting, formula style, and layout conventions. Use cell references rather than hardcoded values so it's maintainable. Add a short notes column for assumptions. Label anything that needs a professional's sign-off with [VERIFY].

Test how sensitive this model is to its key assumptions.

1. Identify the assumptions that most drive the outputs.
2. For each, show the effect on {output_cell} of a reasonable move up and down (state the range you used).
3. Tell me which assumption the result is most fragile to.
4. Flag any assumption that looks unsupported or unusually optimistic.

Present as a clean sensitivity table and cite the cells you used.

Draft file documentation for this schedule, suitable for a reviewer.

Document:
1. What the schedule does and how it's built.
2. The source of each key input (cite the cell and where the figure came from).
3. Any judgment or estimate applied, and the basis for it.
4. What a reviewer should check, in priority order.
5. Open items and anything needing confirmation [VERIFY].

Write it so a reviewer — or a future CRA auditor — can follow the reasoning without me in the room. Be precise about what is fact vs assumption vs judgment.

Clean up the data in {range}. Do the following and tell me what you changed:

1. Trim leading/trailing spaces and fix inconsistent capitalization in text columns.
2. Convert numbers stored as text into real numbers.
3. Standardize the date format in {date_column} to YYYY-MM-DD.
4. Flag (don't delete) duplicate rows and blank required fields.
5. Split {column} into separate columns where it holds more than one value.

Put the cleaned result in a new sheet, leave the original untouched, and list anything ambiguous for me to resolve.

Create a PivotTable from the data in {range}.

Rows: {row_field}
Columns: {column_field}
Values: {value_field}, summed
Also add: {value_field} as a percentage of the column total.

Sort descending by total, format the values as currency with no decimals, and add a clear title. Then tell me the three things the summary shows that I'd want to look at.

Write an Excel formula that does the following: {what_you_need}

Data layout: {describe_columns_and_ranges}

Give me the formula, explain what each part does in plain language, and tell me what would break it (blank cells, text where numbers are expected, ranges that grow). If a newer function (XLOOKUP, FILTER, LET) makes it cleaner, show that version and note which Excel versions support it.

Apply conditional formatting to {range} to make exceptions visible:

- Highlight negative values in red.
- Highlight any value more than {threshold} different from the prior column.
- Highlight blank cells in a required field.
- Highlight duplicates in {key_column}.

Use subtle fills, not loud ones — this is a working document. Then summarize how many cells were flagged in each category.

Build a variance analysis comparing {current_range} to {prior_range}.

Include columns: line item, prior, current, change in dollars, change in percent. Sort by absolute dollar change, largest first. Format negatives in parentheses.

Then, below the table, write two or three sentences on what the biggest variances suggest — but do not speculate about causes the data doesn't support. Where a cause is unclear, say what document would confirm it.

Create a {chart_type} chart from {range} showing {what_to_show}.

Title it clearly, label both axes with units, format the value axis as {format}, and keep the styling clean and professional — no 3D, no heavy gridlines. Place it on {sheet_or_location}.

Then tell me in one or two sentences what the chart actually shows.

Compare {range_a} against {range_b} using {key_column} to match rows.

Produce a new sheet with three sections:
1. Matched items where the amounts agree.
2. Matched items where the amounts differ (show both and the difference).
3. Items appearing in only one of the two ranges.

Total each section and confirm whether the differences explain the overall variance between the two ranges.

Turn the list in {range} into a working tracker.

Add columns for: owner, status (dropdown: Not started / In progress / Done / N/A), due date, and notes. Apply data validation to the status column so only those values are allowed. Conditionally format overdue items in amber and Done in green. Add a summary at the top showing counts by status.