Microsoft recently added a new function to Excel: =COPILOT()
This new AI-powered function lets you type a question or instruction directly into your spreadsheet and receive intelligent responses…
=COPILOT("List the names of the Mac Operating Systems since 2007")
=COPILOT("Extract the email address", A1:A10)
But before you get too excited, there are some important caveats you need to be aware of.
What You Need to Access COPILOT
Currently, the COPILOT function is only available to users on the Beta Insiders Channel and you’ll need a paid Copilot licence to use it – either the $20-a-month Copilot Pro licence or the $30-a-month Copilot Business licence.
That is on top of your Microsoft 365 subscription.
Note: If you have a Microsoft 365 Personal or Family subscription with the standard 60 AI credits per month, this function currently won’t work for you.
How the COPILOT Function Works
The function is surprisingly straightforward. You simply type =COPILOT(“your question or request”) and it generates responses using AI.
For example, =COPILOT(“List the names of the Mac Operating Systems since 2007”) creates a dynamic array that spills down your spreadsheet with the results.
The function can also take a second, optional argument – a cell or range of cells. So you could write =COPILOT(“What is the average”,E1:E5) to analyse specific data, not that you’d need AI to calculate the average of a set of numbers…it’s an example.
The Reality Check: Where COPILOT Falls Short
Here’s where things get concerning. Microsoft themselves warn that COPILOT can give incorrect responses and recommend avoiding it for numerical calculations. They explicitly state you should use native Excel formulas like SUM and AVERAGE for tasks requiring accuracy.
Even though it returned the correct result with my “What is the average” demo, when I asked it to calculate total expenses for a specific lawyer from a dataset, the COPILOT function gave me one answer, but when I used the tried-and-tested SUMIFS function with the same dataset I got a completely different (and correct) result.
This is particularly problematic because if you’re using AI, you’re likely doing so to save time or because you don’t know the correct formula to use. As long as the result looks “in the ballpark,” you might accept it without verification.
Where COPILOT Actually Shines
Despite its limitations with calculations, COPILOT excels in several areas:
Data Extraction: I used it successfully to extract email addresses from inconsistent contact data where traditional text functions would struggle due to varying formats.
Categorisation: It can analyse text and categorise it effectively. For instance, reviewing employee performance summaries and determining whether they’re “Positive,” “Negative,” or “Neutral.”
Research Tasks: Need stadium names for football teams? COPILOT handled this perfectly.
Data Insights: Perhaps most impressively, it can generate executive summaries from large datasets. I asked it for “5 key insights from this data to give to the CEO” and received genuinely useful analysis points.
The Bottom Line
COPILOT represents an interesting step forward in making Excel more accessible to non-technical users. However, it’s crucial to understand its limitations, particularly around numerical accuracy. Use it for research, categorisation, and data insights, but stick to traditional Excel functions for calculations that matter.
As with any AI technology, we’re still in the early days. The function will undoubtedly improve over time, and your feedback through the thumbs up/down buttons helps Microsoft refine it.
If you have access to COPILOT, I’d encourage you to experiment with it, but always verify critical calculations using proven Excel functions. It’s a tool that shows promise for specific use cases, just not the universal Excel assistant some might expect.
Want to see the COPILOT function in action? Check out the video below