Microsoft CEO Satya Nadella argued that a learning loop is more powerful than an LLM, stating that the real competition is whether an organization learns from what it builds, not which model it picks. A learning loop is a system that improves with use by capturing what happened, learning from it, and improving, unlike most enterprise AI which remembers nothing about a business. Nadella contrasted this with the "human in the loop" approach, which he described as a checkpoint that does not enable learning. He illustrated the concept with a sales team: without a learning loop, an AI drafts proposals that require edits repeatedly, but with a learning loop, the system captures every edit and learns the company's actual sales logic, building proprietary intellectual property. Nadella wrote in an X memo that companies should compete on learning loops that encode judgment, workflows, and expertise, rather than on model selection, and warned against "token-maxing," or using the most powerful model for every task.