Even Microsoft's CEO Satya Nadella Doesn't Blindly Trust AI Models, He “Interrogates” It

In addition to lauding Xbox while also expressing it to have a right business model, Microsoft CEO Satya Nadella has also offered a close look at how he personally works with AI models, saying he does not simply look at an AI-generated answer and accept it. Instead, he uses AI as something he can manipulate, question and reason over.

Nadella uses Excel to push AI beyond simple answers

In a recent episode of the Sources Podcast, Nadella explained that his approach is to actively work with the model rather than treat its output as the final answer (via Wccftech). “I just don’t see a model output and take it, I want to manipulate it, I want to interrogate it, I want to reason over it,” he said.

That comment is a good example of how Microsoft is positioning AI agents around existing productivity software. Nadella has repeatedly pointed to Excel as an important example, where users can work with an AI model inside the spreadsheet rather than simply asking a chatbot a question.

Nadella specifically discussed Microsoft’s Excel agent and how it can be used to work through more complicated tasks. He described selecting a formula inside a spreadsheet and asking the agent to examine it and create multiple scenarios. The idea is not simply to generate a result, but to let the user inspect and continue working with what the AI produces.

Microsoft has also described this approach as a combination of models, context and agents. Nadella has said the company expects AI systems to remain somewhat uneven in their capabilities, making the surrounding software and agent infrastructure important.