Why Microsoft is moving from ChatGPT to Anthropic AI – the full story

Microsoft is set to integrate artificial intelligence models from Anthropic into its Office 365 Copilot suite, signalling a shift in strategy as its partnership with OpenAI encounters challenges. According to sources cited by The Information, the software titan will pay Amazon Web Services (AWS) to access Anthropic’s Claude models, which will handle more complex tasks such as automating Excel functions and creating PowerPoint presentations.

The decision comes after Microsoft’s internal trials reportedly revealed that Anthropic’s AI outperformed OpenAI’s models in certain high-level business functions. While GPT-5 will continue to power core Copilot features, Claude Sonnet 4 will take the lead in advanced productivity tasks. This dual-model approach reflects Microsoft’s evolving strategy to diversify its AI portfolio rather than relying solely on a single provider.

Microsoft’s Office 365 Copilot serves more than 430 million subscribers worldwide. Incorporating Anthropic’s AI promises to enhance the user experience, particularly in spreadsheet automation and slide generation, areas where early testers noted significant improvements. Copilot subscriptions remain priced at $30 per user per month, though analysts estimate the new integration could generate upwards of $1 billion in annual revenue, assuming modest adoption rates.

The move also occurs amid tense negotiations between Microsoft and OpenAI over the latter’s restructuring into a public benefit corporation. OpenAI CEO Sam Altman’s plan has delayed the company’s IPO, while investors such as SoftBank push for clarity on equity and model access. Despite the friction, Microsoft emphasises that OpenAI remains a key partner for frontier AI technologies, maintaining privileged access to its models at no extra cost.

Charles Lamanna, recently elevated to oversee Microsoft’s business applications, has been tasked with integrating Anthropic models into the Office 365 ecosystem. The addition of Claude models marks a continuation of Microsoft’s multi-vendor AI strategy, previously seen with GitHub Copilot adopting Anthropic technology for enhanced developer tools.

Analysts suggest that enterprises have been cautious about Office AI due to occasional bugs in AI-generated outputs. Microsoft’s adoption of Anthropic models appears aimed at addressing these limitations and improving reliability in complex tasks. Beyond Office 365, the company is also exploring alternatives for its consumer-focused Copilot app, highlighting a broader effort to diversify AI solutions.

This dual-AI strategy underscores Microsoft’s recognition of both OpenAI’s continued importance and Anthropic’s growing influence in enterprise productivity, while also navigating competitive pressures from startups developing workplace automation tools. The integration reflects a calculated move to strengthen its flagship products while hedging against risks associated with overreliance on a single AI provider.