How does Microsoft Copilot compare with ChatGPT for Office work?

Two months ago, Microsoft opened Copilot to individuals inside its Office apps. Copilot Pro, launched on January 15 at $20 per user per month, brings Copilot into Word, Excel, PowerPoint, Outlook and OneNote for people with a Microsoft 365 Personal or Family subscription. That is the same monthly price as ChatGPT Plus, so a lot of people who write reports and build spreadsheets are now asking which one to pay for.

They are built for different ways of working. Copilot lives inside the document. ChatGPT lives in its own window, and you bring the work to it.

What each one actually is

Copilot Pro is a subscription layered on top of Microsoft 365. It adds a Copilot pane to the Office apps on PC, Mac and iPad, priority access to OpenAI's GPT-4 Turbo model when demand is high, faster image creation in Designer, and a tool for building custom Copilot GPTs that Microsoft says is coming soon. Copilot in Excel is still a preview and works in English only. It does not include Copilot in Teams, which belongs to the business product, Copilot for Microsoft 365, at $30 per user per month.

ChatGPT Plus is OpenAI's $20-a-month plan for individuals. Its data tool, introduced to Plus users last July under the name Code Interpreter, writes and runs Python code in a sandbox (a sealed-off computing space) to do the math, charts and file conversions you ask for, then lets you download the results.

There is also a free Copilot at copilot.microsoft.com and a free ChatGPT, but neither works inside your Office files.

One task, two tools

Here is a worked comparison. You have a Word draft of a quarterly report and an Excel sheet of monthly sales for four regions. You need a one-page summary with the key numbers, a chart and three recommendations.

A prompt that works well in either tool:

Using the sales data, give me: total sales per region for the quarter,
the region with the largest change from last quarter (with the percentage),
and one sentence on what might explain it. Show your calculation.
Then draft a 250-word summary for managers, plain language, three
recommendations as bullet points at the end.

Expect Copilot to save the most time on the writing steps, because nothing leaves Word. ChatGPT is more transparent on the analysis, because you can open the code it ran and check every figure. Either way, check the numbers yourself against the spreadsheet before the report goes out; both tools can make arithmetic or reading errors.

Where each one fits

Copilot is the better fit if your work starts and ends in Office documents, if formatting matters, and if you would otherwise spend time copying text between windows. ChatGPT is the better fit for data work you want to verify, for files that aren't Office documents, and for general questions, brainstorming and learning that have nothing to do with a specific file.

Businesses have a third question, which is where their data goes. Copilot Pro is for individuals; Microsoft points organizations to Copilot for Microsoft 365, which can also draw on work email, meetings and chats. OpenAI's equivalent for small teams is ChatGPT Team, at $25 to $30 per user per month, with GPT-4, DALL-E 3 and the same file tools; OpenAI says it does not train its models on Team data.

When is Copilot Pro worth it?

Copilot Pro is worth it when three things are true. You already pay for Microsoft 365 Personal or Family, since Copilot Pro is an add-on to it. You spend several hours a week writing or editing in Word, building slides in PowerPoint, or working through email in Outlook. And you would use the assistant in place, on the documents you already have, rather than starting from a blank chat.

It is harder to justify if you mostly need Excel analysis, since Copilot in Excel is still a preview in English only, or if you work in Google Docs or other tools. It also overlaps heavily with ChatGPT Plus: both are built on OpenAI models, both cost $20 a month, and few people need both. If you pay for ChatGPT Plus already and copying text into Word doesn't bother you, keep what you have.

A practical approach is to try one month on a real project. Keep a short log of each time Copilot saved you a step, and each time you went back to ChatGPT or did the work yourself. At the end of the month the log will tell you more than any comparison table.