ChatGPT and Copilot: which to choose?

ChatGPT and Copilot are the undisputed leaders in LLM(Large Language Model) tools, and it's hard to choose one over the other. The growth of these tools has been impressive in recent months. It feeds ambitions, fantasies and anxieties of all kinds. My customers often ask me which one to choose. The question is legitimate, and it's easy to mix "cabbages and carrots" when answering it. I'm going to try to do so, without preconceived ideas, and with the aim of being as exhaustive as possible.

ChatGPT and Copilot battle to dominate the generative AI market. AI-generated image.

ChatGPT and Copilot: which one to choose - The different versions of each tool and their prices

Copilot

In order to get the most consistent comparison possible, the first thing to do is to identify the comparable versions between the two tools.

For Copilot, it's quite simple: there's a free version and a paid version.

The free version can be accessed anonymously at copilot.microsoft.com. This is a GPT4-based conversational agent connected to Bing.

It is also possible to connect to this platform with a Microsoft account (Professional, Education, Family or Live), which unlocks the image generation functionality (DALL-E 3 engine).

Finally, the "Copilot for Microsoft 365" version is the most complete version of Copilot (brief presentation here). It enables you to work with data from your organization's Microsoft 365 tenant, and provides integration within your tools. It is available as an add-on to the existing Microsoft 365 license at a price of €337.20 / year (€28.10 / month divided by 12) at the date of this article. There are, in that last sentence, 2 important subtleties:

  1. It's an add-on, so you'll need Microsoft 365 Business, Enterprise or Education licenses first. We'll see later that this is a technical prerequisite, but also an important condition for the user experience.
  2. Invoicing and commitment are exclusively annual. To date, there is no license for a monthly commitment only, and no trial license.

ChatGPT

For ChatGPT, there are 4 different plans(details available here):

  • Free: conversational agent with limited access (including LLM, image generation and file uploads);
  • Plus ($20/month, or €18/month at this date): access to preview features, higher access limits, access to the DALL-E engine for image generation, creation of custom GPTs ;
  • Team ($25/month annual commitment, or €22.5/month at this date; or $30/month monthly commitment, or €27/month at this date): higher access limits, data not used for LLM training;
  • Enterprise (on request): increased limits, access to support.

What I find interesting to compare are the "Copilot for Microsoft 365" and "ChatGPT Team" offers, which have relatively similar prices with advanced features.

ChatGPT and Copilot: which one to choose - Features

Infrastructure

Let's start with what's under the hood.

LLM and image generation engines are of similar levels (GPT-4 Turbo and DALL-E 3 for Copilot; GPT-4o and DALL-E 3 for Chat GPT Team).

Both are SaaS (Software as a Service) systems, meaning that the infrastructure is hosted in Microsoft (Azure) data centers for Copilot and OpenAI for ChatGPT. Here again, the infrastructure is comparable.

The 2 systems rely on a complex infrastructure. AI-generated image.

Security and confidentiality

For Copilot, authentication is managed by Entra ID (formerly Azure AD). Only a Microsoft 365 organization user account with the appropriate license will be able to connect to the organization's Copilot.

For ChatGPT, there's an authentication mechanism specific to ChatGPT, but also SSO mechanisms (Single Sign-On, i.e. you use the authentication mechanism of a third-party system to be identified on the service) offered by Google, Microsoft 365 and Apple.

In both cases, the organization's data is not used to train the models. This is an essential criterion, as it ensures that the confidentiality of the organization's data is preserved. Interestingly, this protection does not exist in the less expensive versions of either product.

Still on the subject of confidentiality:

  • For Copilot, all access to corporate content is via the Microsoft Graph. This means that any content (document, e-mail, etc.) to which the user is not supposed to have access will not appear in the information analyzed/restored by the AI;
  • For ChatGPT, the mechanism is significantly different. Documents are accessed via individual upload mechanisms directly in the prompt. Documents are thus duplicated from their storage space (Microsoft 365, Google Drive, etc.) to the OpenAI servers.

Use cases

The user experience is probably one of the elements that most distinguishes the 2 products, and probably the decisive criterion for choosing one over the other.

ChatGPT provides a conversational agent that is the single point of entry for all queries, and can be enhanced with connectors to other tools.

For Copilot, the philosophy is reversed (even if there is a chat on the Office portal, similar to what is done on ChatGPT). Interaction takes place within the applications of the Microsoft 365 suite, tool by tool. For example, PowerPoint documents are created directly in PowerPoint, with references to other files in the organization. In another example, automatic meeting summaries are made directly in Teams.

The choice between Copilot and ChatGPT is not an easy one. AI-generated image.

Customization

Both agents can be customized in a number of ways.

For Copilot, there are a dozen different customization options, more or less technical, which will influence either skills (new usage) or knowledge (extension of the search base). Many of these are still in preview version.

As for ChatGPT, I find the extensibility model much clearer and more logical. It's possible to create custom ChatGPTs, focused on a specific business area for example. It's also possible to integrate connectors to other tools, such as Google Workspace or Microsoft 365 (although the integration isn't very advanced, it's just a matter of being able to upload documents stored in these spaces). Finally, there's an API that lets you send prompts automatically (which, to my dismay, Copilot doesn't yet offer).

ChatGPT and Copilot: which one to choose - Strengths and weaknesses

Based on this rapid analysis, we can identify a number of strengths and weaknesses specific to each tool.

In my opinion, ChatGPT offers greater stability than its competitor. Another strength, in my opinion, is the customized GPTs. The third major strength of the tool is the API mechanism for automatically integrating prompts.

Copilot's main strength is its integration with Microsoft 365 tools. In my opinion, this integration is threefold. Firstly, on identity: the same supplier manages both identity and corporate data. Secondly, on use cases, which make it possible to go inside the tools to generate relevant, tool-specific information. And finally, on confidentiality, access is open to corporate data while respecting predefined access rights.

ChatGPT and Copilot: which one to choose - Conclusion

In conclusion, the choice between ChatGPT Team and Copilot for Microsoft 365 should be based on the criteria you value most.

If your goal is to have a tool that offers more personalization, then ChatGPT is the way to go.

If what's important to you is integration with the Microsoft 365 ecosystem you use every day, then Copilot is a more coherent answer.

As these tools are very recent, it's clear that many new developments will be introduced by publishers in the near future. And the relevance of this choice may also evolve according to these innovations.

Damien CELLE

Damien CELLE

Office 365 Consultant

Share

An article by

Damien CELLE

Damien CELLE

Office 365 Consultant

Want to go further?