Microsoft is taking steps to rein in employee spending on AI tokens, which had grown to thousands of dollars per month per worker. The company has begun directing employees to prioritize results for customers and business outcomes rather than simply maximizing technology use—a practice known as "tokenmaxxing."
To reduce costs, Microsoft selected GPT-5.6, a cheaper model, as the default option for internal use. The shift reflects a broader effort to make AI adoption more cost-efficient while maintaining productivity and business value.
