Skip to content

How to use the Copilot function in Excel

What the Excel COPILOT function does, its syntax, uses for data analysis, how to write prompts for it and its limits at launch, with views from two partners.

Article · 4 minute read · 21 August 2025

What is the Excel Copilot function?

The Excel Copilot function is an AI formula in Microsoft 365. You type a prompt in plain language into a cell, and it returns a summary, a classification or another answer based on your data. Unlike a static formula, its output is dynamic: it refreshes when the underlying data changes.

It brings AI into Excel formulas, so you can find patterns in a dataset faster.

“The new CoPilot functionality in Excel is impressively versatile. While it may not directly enhance the core financial modelling techniques we rely on in project finance, it offers substantial value in accelerating data analysis … It’s a powerful tool for strengthening the analytical foundation behind our models.”

COPILOT function syntax, with an example

The syntax is simple:

=COPILOT(prompt, optional_context, …)

For example:

=COPILOT(“Classify this feedback“, D4:D18)

This formula asks Copilot to analyse the customer comments in cells D4 to D18 and sort them by sentiment. The results update as the data changes.

Microsoft says data processed with the COPILOT function is kept secure and is not used to train AI models.

“As a financial model auditor, I’m genuinely excited by the potential of CoPilot in Excel to streamline workflows and enhance productivity across financial modelling tasks … However, strict model audit procedures rely on full transparency and traceability of calculations. When AI-generated outputs lack visibility into the underlying logic or assumptions, it raises a critical question … Can you truly trust AI-generated calculations in financial models?”

Excel Copilot function examples for data analysis

The COPILOT function is built for practical business uses:

  • Summarise data quickly. Condense long survey responses or pick out the main trends.
  • Classify feedback or issues. Tag customer comments or support tickets automatically.
  • Generate ideas. Brainstorm keywords, messages or project outlines.
  • Simplify reporting. Turn complex formulas into plain-language explanations.
  • Create lists and tables. Build structured output directly in the spreadsheet.

How to write prompts for the COPILOT function

To get accurate results from AI in Excel formulas, follow these prompting tips:

  • Be specific about what you want, for example “Summarise in three bullet points”
  • Define the structure: a table, a list or a paragraph
  • Use clear action verbs such as summarise, rank or categorise
  • Include examples when precision or tone matters

The clearer the prompt, the more useful and consistent the output.

The COPILOT function used in an Excel worksheet, with a prompt in the formula and the AI output in the cells
Figure 1: The COPILOT function in an Excel worksheet.

Excel Copilot function limits at launch

At launch, the function has some constraints:

  • It only processes data in your worksheet, with no web or enterprise sources yet
  • It returns dates as text instead of Excel’s own date format
  • Usage limits apply: 100 calls every 10 minutes and 300 calls an hour

Microsoft plans to add richer date support, wider data access, better array handling and continued model improvements.

Which versions of Excel have the COPILOT function

The function is rolling out to Microsoft 365 Copilot subscribers:

  • Windows: Version 2509 (Build 19212.20000) or later
  • Mac: Version 16.101 (Build 25081334) or later
  • Excel for the web: coming through Microsoft’s Frontier program

Microsoft announced the function and its roll-out on the Microsoft 365 Insider blog, which has the latest details.

Why the Excel Copilot function matters for financial modellers

The function puts AI inside the formulas analysts already use, so a classification or summary that once took a separate tool now sits in the workbook and updates with the data.

For financial modellers it helps most with the work around the model: preparing reports, analysing raw datasets and handling feedback. The calculations a lender relies on still need formulas that anyone can trace and a model auditor can test. Our AI in Project Finance Modelling course covers where AI helps in live modelling work and where it does not.

Learn financial modelling in a room

Hands-on courses in person and online, taught to more than 20,000 people on one modelling standard.

AI in Project Finance Modelling

Where AI genuinely helps in live modelling work, and where it quietly fails. Hands-on with Copilot and Claude, practitioner-led. See dates and book today.

View full course details: AI in Project Finance Modelling

Financial Model Review in the AI Era

Financial Model Review in the AI Era: error classes, testing technique and where AI-assisted review earns its place. Practitioner-led, in person or online.

View full course details: Financial Model Review in the AI Era

Best Practice Project Finance Modelling

Build and sensitise robust cashflow models from scratch, faster in Excel. Hands-on, practitioner-led, in person or online. See dates and book your place.

View full course details: Best Practice Project Finance Modelling

Browse all training courses

More on this topic

Webinars

Is prompting Excel the wrong skill to teach your team?

Why freeform prompting gives inconsistent output when you use AI in Excel models, and how skills hold the AI to your house modelling conventions.

How AI is used in project finance modelling

How AI is used across project finance modelling today, the tools gaining ground, what stops teams scaling it and why the session focuses on Copilot.

Services

AI Enablement

Practical AI adoption for modelling teams: we assess how your team builds and reviews models, design AI workflows around that and coach the rollout.

Financial Model Audit

An independent opinion on the model behind your transaction, with a red flag report on any show-stopper issues in under 5 working days.

Learn to build these models yourself

We run public and in-house courses across the full financial modelling curriculum.