Microsoft's internal AI models aim to lower costs as engineers tackle AI‑assisted design

Microsoft is routing many routine Office queries in Word, Excel and Outlook to its own in‑house MAI model family, reducing reliance on external providers such as Anthropic and OpenAI. By handling high‑volume, low‑complexity requests with cheaper internal models, the company expects to cut per‑query inference costs and improve margins for products like Copilot, though consumer pricing may not change immediately.

At the same time, engineers are integrating AI agents into model‑based design workflows. While AI can generate code, tests and models quickly, it can also introduce hidden friction by overlooking engineering constraints, physical assumptions and verification needs. Experts emphasize the need for structured agentic workflows that keep engineers in the lead, ensuring AI‑produced artifacts are disciplined, reviewable and aligned with system intent.