CoPilot: Generative AI meets Windows

Generative AI hype has overwhelmed the technology sector for the last nine months. It was Microsoft's alliance with OpenAI that kickstarted this buzz. The launch of ChatGPT and Bing Chat at the start of the year opened the eyes of the world to the full potential of Artificial Intelligence. Tech firms have spoken of nothing else since. During this time, Microsoft have announced multiple generative AI products across much of their enterprise software portfolio. This week, those capabilities finally reached Windows with the release of Microsoft CoPilot.

CoPilot for Windows

CoPilot is not the first time that Microsoft have tried to embed an AI based virtual assistant into their flagship operating system. It was only earlier this year that Microsoft removed the last vestiges of the ill fated Cortana. Many of the capabilities of the old Cortana integration are available in Microsoft CoPilot.

The ability to find files or tweak settings based on natural language prompts is a key pitch for CoPilot in Windows, just as it was for the old virtual assistant. In effect, CoPilot can act as a supercharged search function, which is useful but is probably not something that people will need day to day. That's what killed the equivalent Cortana functionality. Not enough people needed it to justify the cost of the backend cloud services.

The new CoPilot button in Windows does also offer a direct integration with Bing Chat, but it's unclear how many people will actually use it. In recent years, Microsoft have made a big effort to unify the search capabilities across Bing, Office and Windows. The idea is that you should be able to get the same results across all channels. This sounds nice in theory but has also been controversial. Not everyone appreciates seeing their OneDrive or SharePoint files in Bing search results, even if it is presented securely.

Microsoft 365 Chat

CoPilot does have one significant advantage over Cortana that will drive adoption. It is a Generative AI after all. The real pitch for CoPilot is its integration with Office, and the associated cloud services in Microsoft 365. Here too, Microsoft are building on top of existing capabilities, but they're also able to offer something new. A chatbot built right into Microsoft Office with all the capabilities of ChatGPT. They've even given the chatbot its own branding: Microsoft 365 Chat.

Much of the discussion around CoPilot has focused on its ability to help draft or redraft documents and emails. Yet, the Office team has been working on similar capabilities for a while. Features such as Editor in Word or Designer in PowerPoint could already redraft individual slides or paragraphs. These capabilities are already helpful, if somewhat inflexible, and will be massively enhanced by the integration of a Generative AI.

However, for many executives, the ability to transcript meeting recordings or summarise content will be more useful. Microsoft has previous experience with auto-generated meeting transcriptions and highlighting actions in emails across much of its Office 365 portfolio. Outlook has had a suggested actions feature for years, an add-on which has gradually become more useful over time. Generative AI adds an extra layer to this capability, providing the ability to summarise an email thread as well as greater accuracy for video transcriptions.

Cost Control

The use cases for CoPilot in Microsoft 365 are compelling. The pricing is far less so. At $30 per month, I can see many companies choosing to pass on this innovative new technology just due to cost. The equivalent Google Bard functionality doesn't have a steep price tag, which could prove to be a significant competitive advantage to Google Workspace.

Microsoft does have one advantage, which could help them argue against cost concerns. Few technology companies are as trusted by enterprise, which should overcome many of the compliance and security hurdles that slowed the adoption of Generative AI in a business context this year. Much like Bing Chat, Microsoft doesn't use the conversations of business customers to train the underlying ChatGPT AI model.

It's still early days, and business usage of Generative AI is still in the experimental stage. Not all of the envisioned use cases for Generative AI will achieve mainstream adoption. It's also not clear how much companies or individuals will be willing to pay just for Generative AI functionality. This has impacted the market for virtual assistants such as Alexa or Siri, growth of which has stalled in recent years. We're already seeing limited usage of the technology for content writing and translation, but mostly only among financially stretched small businesses. At the moment, the output of a good human copywriter is still vastly superior. This may change over time.