Microsoft chief executive Satya Nadella outlined how the company's AI agents are being used to scrutinize and manipulate data rather than simply accept model outputs. In a recent episode of the Sources podcast, Nadella described interrogating AI-generated spreadsheets and SEC filings to test assumptions and refine business decisions. He highlighted the Excel agent's capacity for causal reasoning, noting that he can open a complex financial model, examine the formulas behind a single cell, and request alternative scenarios. 'I don't just take a model output and move on,' he said. 'I want to manipulate it, interrogate it, reason over it.' This approach, he argued, enables deeper analysis of enterprise data and supports more informed capital allocation. Nadella also described a system that aggregates SEC filings and internal metrics into a semantic model, which is then processed by a coding agent to generate real-time dashboards tracking return on invested capital across business units. His comments follow a broader industry conversation about AI's role in streamlining executive tasks, including workforce planning and operational efficiency. The discussion reflects growing interest in how AI can move beyond automation to support judgment and strategic oversight in large organizations.