Microsoft is also training sales teams to position OpenAI, Anthropic and Google as rivals, reinforcing a broader push to sell its MAI models as a fully integrated enterprise AI stack.
Microsoft is starting to route selected Microsoft 365 workloads away from OpenAI and Anthropic models and onto its own MAI lineup, while also sharpening its sales pitch against OpenAI, Anthropic and Google as direct competitors in enterprise AI. After unveiling seven in-house Microsoft AI models at its Build 2026 conference in early June, Microsoft began on July 7 migrating parts of Excel and Outlook to its MAI stack, sending tens of thousands of prompts per week through internal systems rather than paying third-party inference costs. During an internal planning session around July 14-15 for fiscal year 2027, Executive Vice President Jay Parikh described a strategy centered on selling Microsoft’s full end-to-end AI system rather than individual components. Copilot EVP Jacob Andreou reportedly told staff that Anthropic’s Claude was “slower and less accurate” inside Microsoft Office applications, as sales teams were given talking points portraying OpenAI, Anthropic and Google as offering less integrated solutions. The shift comes after Microsoft amended its partnership agreement with OpenAI in April, lifting an exclusive access arrangement that had given Microsoft preferential rights to OpenAI’s models. Microsoft’s flagship MAI-Thinking-1 model runs on 35 billion parameters with a 256K context window, and the company says it can deliver up to 10 times better cost efficiency in tuned enterprise workloads while matching or outperforming GPT-5.5 and Claude Opus 4.6 on key benchmarks. The move highlights both a cost-control strategy and a more explicit competitive break from the earlier narrative of Microsoft and OpenAI as closely aligned partners.