
What happened
In his annual letter, Satya Nadella said Microsoft organized into "Agents and Infra" and "Devices and Consumer" segments, offering over 11,000 models from OpenAI, Anthropic, Mistral and others.
Why it matters
Nadella frames cost against results as the "cost-to-outcome" frontier, and separating the harness from the model lets Microsoft pick the right model per job, which he says strengthens resilience.
WHO IT HITSEnterprise IT and procurement teams buying Microsoft 365 Copilot will face a model-selection layer that can swap generalist models without losing accumulated company-specific learning, while finance teams get a "cost-to-outcome" yardstick for AI spend.
Summaries like this, every morning.
Nadella's letter ties Microsoft's internal reorganization to a specific architectural bet. By separating what he calls the harness — context, memory and action space — from the model itself, Microsoft says it can route work to the "right model for the right job" across more than 11,000 models from OpenAI, Anthropic, Mistral, SpaceX and its own MAI family. The stated payoff is twofold: optimizing what he calls the "cost-to-outcome frontier," and business continuity, because interchangeable models reduce dependence on any single provider.
The customer-control argument follows from the same design. Nadella says customers should be able to "switch out a 'generalist' model without losing the 'company veteran' expertise accumulated in their learning system" — a way of making model choice a procurement decision rather than a lock-in. Copilot's move from synchronous chat assistants to asynchronous agents that complete long-running tasks, and its surpassing 30 million paid seats, are presented as evidence that this shift is already underway. Agent 365 extends the logic to management: agents handled the way companies handle people.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Burke Wealth Management's Q3 2026 investor letter said Amphenol's data center power business essentially picks…

A Seeking Alpha preview of Corning's Q3 results says AI demand is driving revenue optimism for the company, th…

Investing.com reports that all eyes are on Corning's earnings as its AI fiber growth faces a test

At the PRovoke Global Summit, McDonald's discussed rethinking its PR strategy in response to AI reshaping earn…

Honeywell Technologies launched a new AI platform at Dangote's refinery in Nigeria, according to Reuters

Booking.com is now having AI answer traveller questions ahead of the guest's arrival, a shift Hospitality Net…
