Microsoft CEO Satya Nadella has penned a lengthy X post detailing Microsoft's position on AI models and what companies of the future should value most in AI.
Interestingly, Nadella argues that Microsoft is pushing for a multi-model strategy, and that its Foundry supports different models, agent frameworks, and other Frontier Tuning. Nadella outlines how businesses should operate in an increasingly AI-driven economy and says a business shouldn't adopt a single AI model and build its entire business around it. Instead, Nadella argues that companies should build their own learning systems on top of multiple models, combining employees' expertise with AI capabilities.
Nadella describes this setup as a balancing act between "human capital" and "token capital," with human capital being knowledge, judgment, relationships, ingenuity, and pattern recognition, while token capital represents the AI capabilities a company can develop using its internal knowledge, workflows, and accumulated experience.
According to Nadella, companies that want to operate efficiently should be able to fully replace the general-purpose AI model currently powering their systems without losing any valuable data, knowledge, or learning systems. In other words, a business should not have to start from scratch whenever it moves from one AI provider or model to another. Its institutional knowledge should remain within the wider system it owns, rather than being permanently tied to the underlying model.
Nadella said companies can achieve this by turning their workflows, domain knowledge and employee judgment into AI systems that become more capable with each use. Companies could also use private reinforcement-learning environments to improve models using real traces and feedback generated inside the organization.
Nadella compared this continuously improving system to a "hill climbing machine," where every completed workflow creates additional training signals that make the company's AI more useful.
"This loop becomes the new IP of the firm," Nadella wrote. "The companies that build this early will have an advantage that is hard to replicate, regardless of any new individual model capability."




