Microsoft 365 Copilot Integrates GPT-4

Microsoft announced 365 Copilot in March 2023 — GPT-4 integrated directly with Microsoft Graph, giving the model access to your email, calendar, documents, and Teams conversations as live context for every interaction.

Microsoft Graph is the API that connects the Microsoft 365 data layer, including emails, calendar events, Teams messages, SharePoint documents, and OneDrive files. Microsoft 365 Copilot uses Graph to give GPT-4 access to your actual organisational context.

When you ask Copilot to draft an email based on a meeting, it can read your meeting transcript, your previous emails on that topic, and the related documents. This provides personalisation not from training on your data but from runtime access to it.

For instance, I have seen Copilot being used to automate the process of generating meeting summaries, which can save around 2 hours per week for each user. With tools like Azure Active Directory and Microsoft Intune, the IT team can easily manage access and deployment.

Moreover, the integration with Microsoft Graph allows for a high degree of customisation. This includes APIs like Microsoft Graph Toolkit and SDKs like Microsoft Graph SDK for Python. These tools can be used to build custom applications on top of Copilot, such as automated report generation and data analysis.

Microsoft 365 Copilot inherits your organisation's Microsoft 365 permissions and compliance settings. Copilot does not grant users access to data they cannot already access. If a document is restricted to a specific group, Copilot will not surface it to users outside that group.

The data processing happens in Azure, under Microsoft's existing enterprise compliance commitments including GDPR, ISO 27001, and SOC 2, which should provide some comfort to organisations with strict data governance requirements. With Azure Monitor and Azure Log Analytics, the IT team can monitor and troubleshoot any issues that arise.

In terms of trade-offs, the main consideration is the cost, with the $30 per user per month pricing model, which may be prohibitively expensive for small and medium-sized businesses, but for large enterprises, the productivity gains and cost savings can be substantial. Some estimates suggest a return on investment of around 300% in the first year.

The pricing for Microsoft 365 Copilot is $30 per user per month as an add-on to existing Microsoft 365 enterprise licences. This means for a 1,000-person organisation, that is $360,000 per year. The ROI calculation requires an honest assessment of how many knowledge workers will use it actively.

Early enterprise reports suggest that writing-heavy roles, such as marketing, legal, and communications, see the clearest productivity gains from using Microsoft 365 Copilot. I think this is because these roles can benefit from the AI-powered drafting and research capabilities.

Additionally, tools like Power Automate and Power Apps can be used to integrate Copilot with other Microsoft 365 applications, such as Dynamics 365 and SharePoint, which can further enhance the productivity gains and cost savings. With the Microsoft 365 Copilot API, developers can build custom integrations with other applications and services.

The buying decision for Microsoft 365 Copilot involves a cross-functional team, including IT, finance, HR, and legal, which can slow down the adoption process, but I think this is a necessary step to ensure that the organisation is getting the most out of the technology.

IT evaluates the security and compliance model. Finance evaluates the ROI. HR evaluates the workforce impact, and legal evaluates data residency and liability. This process takes time, which is why the enterprise AI adoption cycle in 2023 was slower than the press coverage suggested.

Furthermore, the adoption process can be facilitated by using tools like Microsoft FastTrack, which provides guidance and support for deploying and adopting Microsoft 365 Copilot. With the Microsoft 365 Copilot community, users can share best practices and learn from each other's experiences.

IT also needs to consider the infrastructure requirements, such as the need for Azure Active Directory Premium and Microsoft 365 E5 licences, and the potential impact on network bandwidth and storage, which can be significant, especially for large organisations with thousands of users.