Uber employees can now rehearse presentations and get feedback from an AI trained to think and respond like CEO Dara Khosrowshahi before meeting with the real executive. The tool, called "Dara AI," reveals how companies are using AI to simulate executive decision-making patterns - and raises questions about what this means for corporate culture.On one hand, this is genuinely clever. Getting feedback before high-stakes meetings makes sense. Not everyone has easy access to senior leadership, and when you do get face time, you want to make it count. An AI that can simulate the CEO's thinking style, predict his questions, and identify weaknesses in your pitch could be valuable preparation.On the other hand: we're training employees to optimize for an algorithm. Instead of crafting compelling arguments, they're gaming what the AI predicts the CEO will respond to. That's a subtle but significant difference, and I'm not sure it leads to better decisions.The technology behind this is straightforward. Train a large language model on the CEO's emails, meeting transcripts, public statements, and decision patterns. Fine-tune it to replicate his communication style and business philosophy. Give employees an interface to test their pitches and get AI-generated feedback. From an engineering perspective, it's impressive but not revolutionary.What's more interesting is what this says about corporate structure and power dynamics. In theory, "Dara AI" democratizes access to executive thinking. Employees at any level can test ideas against what the CEO would likely say. That could surface good ideas that otherwise wouldn't get heard.But it also reinforces a very specific approach to decision-making: the CEO's judgment is the ultimate arbiter, and the goal is to align your thinking with his. There's no AI clone of front-line drivers or customer service reps that executives consult before making decisions. The information flow is one-directional, from the top down.I'm also curious about accuracy. How well does "Dara AI" actually predict Khosrowshahi's real responses? If it's highly accurate, that raises interesting questions about whether executive decision-making is more algorithmic than we think. If it's not accurate, employees are preparing for a simulation that doesn't match reality, which could be counterproductive.The bigger trend here is AI-mediated human interactions. We already have AI scheduling assistants, email tone analyzers, and interview prep tools. "Dara AI" extends that to simulating specific individuals. That has obvious applications beyond corporate settings - imagine practicing difficult conversations with AI versions of family members or colleagues.But something gets lost when we mediate everything through simulations. The point of presenting to the CEO isn't just to get approval - it's to engage in dialogue, respond to unexpected questions, and potentially change his mind. If everyone optimizes their pitches based on what "Dara AI" predicts will work, they're converging on similar approaches rather than bringing diverse perspectives.There's also the question of how Khosrowshahi himself feels about this. Having an AI clone of you that employees interact with more often than they interact with you directly is... weird. It could be flattering or creepy, depending on your perspective. And what happens when the AI gives advice that contradicts what the real CEO would say?My take: "Dara AI" is a fascinating experiment that reveals both the potential and the pitfalls of AI in corporate culture. It could genuinely help employees prepare better. But it also risks replacing authentic communication with optimization theater. Whether that's good or bad probably depends on whether you think the CEO's judgment is generally correct and should be replicated, or whether you think diverse, unoptimized perspectives lead to better decisions.Either way, it's a very Black Mirror moment - impressive technology that makes you slightly uncomfortable even as you acknowledge its utility.