AI Runs the Models. You Still Have to Own the Room.
The tools have gotten better at analysis. The room still belongs to whoever is actually in it
A model can process a negotiation's history, predict likely objections, and surface the data point that wins an argument, all before the meeting starts. What it cannot do is read the hesitation in someone's voice when they say yes to something they have not actually agreed to, or notice that the quietest person in the room is the one whose buy-in actually matters. Owning the room was never primarily an analytical skill. It is a relational one, and AI's growing competence at the first has done nothing to touch the second.
Preparation Is Not Presence
The temptation with any powerful analytical tool is to treat thorough preparation as equivalent to genuine readiness. A person who walks into a meeting having consumed a model's complete analysis of the other side's position may feel more prepared than they have ever been, and in one sense they are. But preparation built entirely from data has a particular blind spot: it cannot account for what changes in the room itself, the shift in tone when a difficult subject comes up, the moment someone reveals a concern the briefing never anticipated because it was never written down anywhere the model could read it.
Owning the room means noticing that shift in real time and adjusting, which requires a kind of attention no model can supply because it requires being there. This is not a case against preparation. It is a case against mistaking preparation for the thing itself.
What Gets Lost When Analysis Replaces Judgement
There is a specific failure mode worth naming: a person so reliant on a model's analysis that they lose the habit of reading a room independently, because the analysis has always been available to lean on instead. This is not a hypothetical. It resembles what happens to any skill that stops being practised because a tool now does the work. The skill atrophies quietly, and the person only discovers the loss the first time the tool is wrong or absent and they have nothing to fall back on.
Genuine expertise is not information plus confidence. It is judgement earned through the accumulated experience of having been in many rooms, having read many situations correctly and some incorrectly, and having built an instinct that a dataset alone cannot replicate. A tool that shortcuts the analysis without the person continuing to practise judgement produces someone who sounds prepared and is, underneath the preparation, less capable than they were before the tool arrived.
The Discipline This Requires
Use the analysis. Do not let it replace the discipline of arriving early enough to read the room before anyone starts speaking, of watching faces rather than notes during the parts that matter most, and of trusting an instinct that contradicts the model's prediction when the room itself is telling you something the model could not have known.
This is, at root, a question about where dignity actually resides in an encounter between people. A negotiation, a pastoral conversation, a difficult conversation with an employee, these remain encounters between persons, not between a person and a dataset about another person. The model can inform the encounter. It cannot become a substitute for showing up to it fully present, and the room, every time, can tell the difference.