Artificial intelligence has become increasingly difficult to avoid in the workplace, whether you particularly want to see it or not. An ever-present for many office workers is Copilot, the chirpy chatbot that aims to be your “AI companion” through just about every task you want to perform on a PC.

As well as being built into Windows, Microsoft Copilot is heavily integrated with Microsoft 365, working across applications like Word, Excel, Outlook, PowerPoint, and Teams. On paper, bringing AI assistance into these applications rather than requiring you to open a separate tab is a timesaver. But that all depends on one obvious question: is Copilot actually any good?

The answer is more complicated than you might think. For anyone accustomed to ChatGPT, Claude, or other leading AI tools, some aspects of Copilot might not seem that impressive. But while it might not be the cleverest chatbot in the building, being right there in the room with you can make it just as useful.

What is Microsoft Copilot?

Microsoft Copilot is Microsoft’s suite of generative AI tools for work and personal use. Used within Microsoft 365, Copilot aims to help users create, analyse, summarise and edit content in applications including Word, Excel and Outlook.

Users can pose queries to the ‘AI assistant’, which can work across documents, spreadsheets, email, meetings, and other Microsoft 365 content. These might include making changes or recommendations given the context of the files it has access to, and completing simple tasks within set guidelines.

Because of the different Microsoft 365 subscriptions and ways organisations can be configured, the version of Copilot you use and the features it offers can vary. Microsoft has also been changing the way its Copilot products are packaged and presented, so it’s worth checking your current license rather than relying on descriptions of what is or isn’t included.

It’s important to note the difference between Copilot and a general-purpose LLM like ChatGPT or Gemini. While these LLMs can be used as conversational tools with a broad range of capabilities, Microsoft Copilot has a more specific range of use cases. The version within Microsoft 365 for instance is designed specifically to operate within that environment.

While this means Copilot has certain limitations compared to other AI chatbots, it also has certain advantages. It can work with the information contained in your Microsoft 365 environment without you needing to paste or upload files for it to read, subject to your licence, permissions, and organisational configuration.

In Outlook, that might mean it can work with emails and calendar information. In Word, it can work directly with your documents, while in Excel it can analyse and modify your workbooks. That poses certain risks, but for a business that regularly utilises Microsoft 365 software, this integration can also offer huge value in terms of productivity over a standalone chatbot.

Is Copilot any good?

Copilot is good for certain tasks, and less good for others. That isn’t fundamentally any different to any other AI tool. Each has its own strengths and weaknesses: Claude for instance is great for coding and clear answers, but less good at flowing prose.

Whether Copilot is the right fit for you depends on what you want it to do. Some people find its writing generic compared to other LLMs, and that it provides less accurate or comprehensive responses. It can also struggle with multi-step processes compared to some other tools, which can better recall and utilise previous responses to inform new ones.

The company itself acknowledges this to a certain extent. AI can always produce incorrect or misleading information, and Microsoft warns users to review all generated results for accuracy. Its guidance for Excel, for example, explicitly notes that Copilot can make mistakes, misinterpret information, or produce inaccurate results.

AI is a tool, and different tools are better at tackling different problems. ChatGPT might produce a better first draft of an article, while Claude may be better at analysing a large document or spreadsheet. Another specialist tool meanwhile might produce better images, or transcribe a meeting more effectively.

Copilot only has to be good at the things you need it for, not to win every battle with its competitors. What Copilot is great at is interacting with the information and applications your business likely already uses. In Word, it can rewrite a passage without requiring you to copy and paste it into another application. In Outlook, it can summarise an email thread or help draft a response. In Excel, it can work directly with a workbook, creating and fixing formulas, charts, and PivotTables.

Microsoft is also constantly working to expand the AI models available within Copilot. Microsoft has indicated that some Copilot experiences now use models from both OpenAI and Anthropic, with model selection available in certain applications. In that sense, Copilot is starting to offer the best of both worlds, providing the complexity of answers and analysis offered by rival LLMs, while providing unique access to your Microsoft 365 files.

How to use Copilot in Excel

Excel is one of the most obvious use cases for Copilot. By combining AI with the complex spreadsheet software, Copilot can help analyse data, generate formulas, identify trends and outliers, create charts and PivotTables, and make changes to worksheets.

Microsoft now provides different modes for Copilot in Excel, including edit, plan and chat modes. Edit mode allows it to make changes to a workbook, while plan mode allows you to review its proposed approach before changes are made. Chat mode by contrast keeps the response within the conversation, rather than modifying the workbook.

The quality of the result depends heavily on how well your spreadsheet is structured. If you have a neatly organised table containing monthly sales figures, for example, you might not get much change out of asking what the spreadsheet tells you.

Instead, you’re more likely to find success asking Copilot to analyse this year’s sales data by region, and create a chart with the percentage changes of the regions with the strongest growth. Microsoft recommends being specific about the columns or ranges involved, and refining prompts when the initial result is not quite right.

Copilot is also useful for the less glamorous side of Excel. It can help create formulas, explain how formulas work, apply formatting, and filter or sort information. The important caveat is that you still need to understand the numbers.

If Copilot creates a formula that looks plausible, but is ultimately wrong, Excel isn’t going to fix it for you. For tasks like financial reporting, forecasting, or other high-consequence work, Copilot still needs to be treated as an assistant, rather than an accountant.

How to use Copilot in Outlook

Email is one of the areas where AI can provide an immediate and obvious productivity benefit. For most people, reading and replying to emails often involves many of the same stock responses. The trick lies in summarising email content, identifying which emails can be answered in the same terms, and which ones demand a more considered approach.

Copilot in Outlook might be most useful as a way to summarise emails, pulling key information out of massive nested email threads. But it can also help draft messages, give you advice on the content and tone of your drafts, help push incoming messages into specific inboxes, and even help with scheduling.

Microsoft also provides a chat interface that can interact with relevant inbox and calendar information, although the precise capabilities depend on licensing and organisational settings.

Again though, it’s important to reiterate that while AI-generated email can sound polished, it can still be subtly wrong. Copilot might misunderstand the importance of a particular detail, use an inappropriate tone for a certain contact, or confidently state something that you never agreed on. In short: always read the message before sending it.

How to use Copilot in Word

Copilot in Word probably feels the most familiar for anyone acquainted with LLMs like ChatGPT. It can create drafts, rewrite existing material, summarise documents, answer questions about their contents, and help restructure your writing.

Copilot is particularly useful for overcoming the ‘blank-page problem’. Instead of starting a report from scratch, you can give Copilot the subject, audience, desired structure and source material, and ask it to produce a first draft. You can then work through that draft yourself, checking its claims and improving the areas where the AI sounds staid and robotic.

It can also be useful for transforming existing material. For example, you could ask it to turn a technical document into a customer-facing summary, shorten a lengthy report for senior management, or rewrite a section in a more formal style. You could even use it to reword an old blog to freshen it up for reuse (though you should be careful about this from an SEO perspective!).

Microsoft also allows users to reference other documents, emails and meetings where the relevant capabilities and permissions are available. This is where Copilot begins to look more interesting than a conventional chatbot because it can potentially connect the task with the context in which the work is actually taking place.

Again though, quality control matters. Copilot is very good at producing something that looks plausible at first glance, but which lacks specificity to your organisation, and can contain mistakes. If a document contains legal, financial, technical, or regulatory information, every important statement needs to be checked against the source material.

Should your business use Microsoft Copilot?

For most organisations already invested heavily in Microsoft 365, Copilot is well worth considering. Using Copilot doesn’t mean you need to stop using ChatGPT or other AI tools. Different tools have different strengths, and those strengths will continue to change as the underlying models develop (as many Claude converts will tell you).

The biggest argument for using Copilot is integration. It brings AI assistance into Word, Excel and Outlook more seamlessly than any other tool, with the potential to speed up repetitive work like document prep, getting insights from spreadsheets, or summarising emails.

The best results from any AI tool always come from good instructions (or prompts, in AI speak). Give it good instructions that provide the context it needs, and be proactive about asking it to show its workings. Use the plan or chat modes in Copilot before allowing it to make substantial changes, and check the output for any errors.

For businesses considering Copilot, it’s also worth thinking about your IT estate. Copilot is most effective within a well-managed Microsoft 365 environment, where things like permissions and access rights are configured properly, and data is managed and structured in a way that helps Copilot do its job.

–

The technology behind Copilot will continue to improve, and Microsoft’s integration of increasingly capable models into 365 and other software is likely to make it more useful over time. But businesses don’t need to wait for a mythical point at which AI becomes infallible when the practical opportunity is already here.

Start using Copilot for repetitive work like first drafts, summaries and admin, but keep people responsible for judgement, accuracy, and decisions. If getting the answer wrong would be expensive, don’t leave it to your copilot – take the wheel.