Organizing a disordered Excel spreadsheet often takes longer than its initial creation. To test whether Microsoft’s Copilot could alleviate some of this burden, I provided it with a purposely unstructured workbook and a concise instruction. However, instead of auto-correcting issues, a lot of its focus was spent on determining which problems to leave untouched.
I Kept My Request Deliberately Simple
A Good Assistant Shouldn’t Need Extensive Instructions
Instead of meticulously crafting a detailed request, I aimed to assess how advanced Microsoft’s AI has truly become. Copilot is integrated within Excel, which means it can access the workbook directly without needing a detailed explanation of the data structure.
Another reason for a straightforward prompt was the common critique of AI tools, which often require excessive time to refine requests. This can negate any time saved by using the assistant in the first place. Thus, my request to Copilot was:
Organize the spreadsheet while keeping the original data intact. Enhance the formatting, correct inconsistencies when appropriate, highlight anything requiring manual review, and provide clear explanations for each major modification. Avoid making assumptions if the correct value isn’t clear.
The workbook contained a multitude of chaotic data points, such as inconsistent formatting, varied spellings, impossible dates, and unnecessary blank rows. While Microsoft recommends using structured tables for optimal performance, I wanted to evaluate how Copilot managed an unstructured dataset.
The Review Sheet Proved Most Valuable
The Assessment Outperformed the Edits
Upon reviewing the workbook and checking for date accuracy and calculation errors, Copilot concluded the cleanup process as “complete.”
A standout feature was the “Issues – Manual Review” worksheet. Rather than offering a brief overview, Copilot compiled an entire sheet listing issues that needed human attention, categorizing them efficiently.
Most notably, it provided clear explanations for each flagged item. For example, impossible dates were labeled with notes like “Month 13 does not exist.” Mismatches in calculations displayed both recorded totals and expected values, while ambiguous dates were given multiple interpretations instead of blind guesses.
This level of detail would significantly streamline the manual review process. Despite my vague initial request, it generated a surprisingly comprehensive audit and clarified its reasoning for choosing not to make certain adjustments.
The Spreadsheet Still Required Significant Work
Many Obvious Issues Went Unaddressed
When I first examined the supposedly “cleaned” spreadsheet, I was surprised by the amount that remained unchanged. While some changes, like the removal of excess whitespace and improved number formats, were evident, numerous blatant formatting issues persisted.
Inconsistent fonts and font sizes continued to clutter the workbook, and existing cell background colors remained untouched, complicating Copilot’s conditional formatting.
While it claimed to have standardized date formats, many still varied. Given that formatting is generally straightforward, I anticipated Copilot would identify clear inconsistencies.
Currency formatting improved for number entries, but instances of text such as “USD 12.00” were merely flagged rather than converted. Even minor layout choices, like centering some columns while left-aligning others, seemed arbitrary rather than enhancing readability.
In various instances, Copilot appeared to prioritize the wrong elements. The header row, which was already well presented, received more attention than the significantly messier sections of the workbook.
Copilot Knows When To Avoid Guesswork
At Times, It Was Excessively Cautious
Copilot demonstrated a commendable ability to refrain from making risky assumptions. For instance, it didn’t alter status values like “Complete,” “Completed,” and “Done,” recognizing that they might hold different meanings. It also refrained from changing various order ID formats, which seemed valid in their own contexts.
While these choices were sensible, others appeared overly cautious. It avoided capitalizing state names under the premise that it required domain knowledge, even though simply changing “arizona” to “Arizona” is unlikely to misinterpret the data. Similarly, it left inconsistent capitalization in customer names untouched, even when corrections seemed obvious.
This caution extended to unclear dates. Copilot rightly flagged entries like “08/01/2025,” as they could represent either January 8 or August 1 depending on locale. However, it also flagged “13/05/2023,” even though “13” cannot be a valid month. In that scenario, it likely could have corrected the error without issue.
This conservative approach minimized the potential for mistakes, but resulted in several tasks that perhaps could have been resolved automatically.
Would I Use Excel Copilot Again?
Not for Cleaning Spreadsheets
This evaluation answered many of my initial questions.
Did my simple request limit Copilot’s effectiveness? Possibly—but that was intentional. While Copilot generated an impressively thorough review sheet, it did not deliver as well on the actual cleaning. A longer prompt might yield better results, but that trade-off becomes questionable if the time spent crafting the prompt equals that of fixing the spreadsheet.
Is it faster than Excel’s built-in features? Not really. Standard tools such as Find and Replace, Power Query, and traditional formatting methods remain my go-to solutions for a majority of issues encountered in this workbook.
Can I trust its outcomes? Only after personal verification. While the review sheet was genuinely helpful, some questionable decisions in the cleaned version caused concern.
As a tool for reviewing spreadsheets, Copilot exhibits potential. However, for actually cleaning them, it still has progress to make. If I had to manage another messy workbook, I wouldn’t rely on Copilot for repairs. Instead, I’d likely request an initial audit and then utilize Excel’s native tools for the corrections.
Copilot Is Not a Universal Solution
This test did not convince me that Copilot is the ideal solution for spreadsheet cleaning. However, this doesn’t mean I would dismiss it entirely. I’ve found it surprisingly effective for more creative and analytical tasks, from crafting dashboards and reports to suggesting complex formulas worth keeping. For the moment, though, I would still reach for Excel’s built-in cleanup tools before considering Copilot for another disorganized workbook.
