Microsoft Copilot: A Different Kind of AI

For a long time, I avoided Microsoft Copilot, focusing instead on AI tools that provided real value. With AI evolving so quickly, you have to choose your battles. ChatGPT and Claude were pretty sufficient for me. However, some of my clients were pushed into using Copilot by their IT teams, and they struggled to make it work. 

I needed to understand why—was Copilot the issue, or were the expectations misaligned?

What I discovered is that Copilot isn’t bad; it’s just a different type of AI that must be used differently. 

So, let’s take a look at how it compares to ChatGPT and Claude and what causes the disappointment.

The Deep Dive: What is Microsoft Copilot, and how to use it?

When people hear "Copilot," they often assume it's a single AI assistant. In reality, Microsoft has multiple Copilots, each designed for different functions:

• Copilot for Microsoft 365 → The one most finance teams use, embedded in Excel, Outlook, Word, PowerPoint, and Teams.

• Copilot for GitHub → AI-powered coding assistant.

• Copilot for Dynamics 365 (Finance, Sales, Customer Service, Security) → Industry-specific AI models.

• Copilot Studio & Copilot Agents → Microsoft’s attempt at customizable AI automation, similar in concept to ChatGPT’s Custom GPTs and Claude’s Projects, but with much less user control (so far).

Unlike ChatGPT and Claude, which are standalone general-purpose AI assistants, Copilot is built to work inside Microsoft apps. That means:

✅ It can pull data from Excel, Outlook, Teams, and SharePoint to provide context-aware insights.
✅ It maintains Microsoft 365 security policies, meaning it should not expose data to unauthorized users.
✅ It integrates seamlessly into existing Microsoft workflows, reducing friction.
✅ It uses OpenAI’s GPT-4o as its core model but also includes Microsoft-trained models for specific tasks.

However, Copilot lacks many features I rely on in ChatGPT and Claude.

Where Copilot Falls Short Compared to ChatGPT & Claude

Despite its seamless Microsoft integration, Copilot lacks key AI capabilities that I find invaluable in ChatGPT and Claude.

🚩 No Custom GPTs

• ChatGPT allows users to create AI models tailored to specific financial tasks, such as custom-built GPTs for FP&A, document automation, or compliance workflows.

• Copilot has no equivalent customization, and Copilot Agents (Microsoft’s alternative) are still largely untested.

🚩 No Project Feature

• The project's feature lets you structure AI interactions, keeping AI-generated insights organized over time.

• Copilot resets every time you start a new session.

🚩 Limited Choice of AI Models

• With ChatGPT, users can switch between different models based on their needs, optimizing responses for speed, accuracy, or reasoning capabilities. It currently offers over five different models, each with varying strengths in reasoning, input types, and output quality.

• Copilot is locked to GPT-4o, with an option to use o1 for deeper reasoning through the Think Deeper function—which is slow (30+ seconds per response) and limited in complexity.

🚩 Limited Agentic Capabilities

• ChatGPT’s Operator mode and Claude’s Computer Use feature enable it to autonomously browse the web, retrieve files, and interact with external tools.

• Microsoft Copilot does not have comparable capabilities. While Copilot integrates well with Microsoft 365, it does not perform autonomous actions.

🚩 Lower-Quality Outputs (Based on User Feedback)

• Multiple users report that Copilot’s answers are lower in quality and more error-prone than ChatGPT’s.

• It misinterprets financial data more often, requiring frequent fact-checking.

• Unlike Claude, which excels in document analysis, Copilot struggles with complex financial reports.

🚩 Longer Learning Curve & Less Intuitive Interface

• Copilot isn’t as user-friendly as ChatGPT or Claude.

• Many finance professionals struggle with its interface and need training to use it effectively.

• It lives in a sidebar inside Microsoft 365 apps—which is convenient but also makes it harder to manage conversations across multiple documents or projects.

Now that I’ve worked with Copilot myself, I understand why some of my clients were frustrated.

During my workshops, they first examined ChatGPT and Claude. These tools can analyze complex data, generate profound insights, and automate manual tasks. I presented them with exciting use cases of "what AI can do."

But then, their IT teams pushed them into Copilot, which wasn’t capable of those same tasks.

The issue isn’t that Copilot is bad—it’s that it’s different.

Copilot is not designed to be a versatile and customizable AI tool like ChatGPT or Claude. Instead, it’s a useful first step into AI adoption and can be valuable when used correctly—as an assistant for office-related tasks that finance teams perform in Microsoft products.

While ChatGPT and Claude outperform Copilot in AI reasoning and deep analysis, Copilot has one key advantage that they don’t:

✅ It’s fully integrated into Microsoft 365.

You can connect ChatGPT to Microsoft apps (and other tools like Google Suite, accounting platforms, CRMs, etc.), but you have to do it manually, app by app.

Copilot, on the other hand, is already connected to everything in your Microsoft ecosystem.

This seamless integration means Copilot excels at automating internal workflows—like summarizing meetings, drafting emails, or surfacing relevant documents from SharePoint—without requiring extra setup.

Copilot for Finance: Microsoft’s attempt to create function-specific copilots

Microsoft continues to expand Copilot’s capabilities by developing function-specific versions tailored for different industries. One example is Copilot for Finance, which is currently in beta testing.

So far, I haven’t found any meaningful user reviews. Based on Microsoft’s description, this appears to be an early attempt to tailor Copilot for finance-specific tasks. However, the suggested use cases aren’t particularly exciting: collections follow-ups in Outlook and reconciliations in Excel.

At this stage, it’s too early to say how useful these function-specific Copilots will be. Microsoft is clearly working on making Copilot more relevant to specialized workflows, but for now, we’ll have to wait and see how these tools evolve over time.

Final Verdict: Is Copilot the Right AI for You?

If your company isn’t fully on Microsoft 365, there’s no reason to choose Copilot.

Even if you are, Copilot shouldn’t always be your AI of choice.

Use Copilot if your main goal is to automate common office tasks—handling emails, summarizing internal data, generating reports in Word, or working within SharePoint and Teams. Copilot is a great fit for streamlining workflows inside Microsoft apps.

If you require advanced AI capabilities—such as complex financial modeling, AI-driven process automation, or custom workflows (as I’ve covered in previous newsletters)—you’ll likely need more powerful tools beyond Copilot.

The right choice depends on your tech stack and specific use case, which is why involving an AI consultant can be a smart first step. If you need guidance on selecting the best AI model for your company, feel free to reach out.

Where to Learn More

Microsoft doesn’t hold back on learning resources when it comes to Copilot. Whether you’re just getting started or want to explore advanced features, here are the best places to learn:

📌 Microsoft’s Official Learning Hub:
🔗 Microsoft Learn – Introduction to Microsoft 365 Copilot
A structured course covering Copilot’s core features, use cases, and how to integrate it into your workflows.

📌 Copilot Scenario Library:
🔗 Copilot for Finance & Business Use Cases
A collection of industry-specific use cases, including how finance teams can use Copilot for reporting, reconciliation, and document automation.

📌 Microsoft’s YouTube Channel:
🔗 Microsoft 365 YouTube – Copilot Tutorials
Short video tutorials showcasing Copilot in action across Word, Excel, Outlook, and Teams.

These resources are updated regularly, so keep an eye on them.

Closing Thoughts

Looking back, I’m actually glad I was pushed into exploring Microsoft Copilot.

I like challenges, and figuring out how to make Copilot work is my next one.

It’s clear that Copilot requires a different mindset, different training, and different security measures than the AI tools I was used to. But that doesn’t mean it’s useless—it just needs the right approach.

Now that I understand Copilot (and what it isn’t), I focus on helping finance teams make it work.

Stay tuned as I continue testing Copilot and finding ways to make it a valuable asset for finance workflows.

If you’ve had success—or struggles—with Copilot, I’d love to hear your experience!

We Want Your Feedback!

This newsletter is for you, and we want to make it as valuable as possible. Please reply to this email with your questions, comments, or topics you'd like to see covered in future issues. Your input shapes our content!

Want to dive deeper into balanced AI adoption for your finance team? Or do you want to hire an AI-powered CFO? Book a consultation! 

Did you find this newsletter helpful? Forward it to a colleague who might benefit!

Until next Tuesday, keep balancing!

Anna Tiomina 
AI-Powered CFO