AI Shouldn't Make Your Decisions. It Should Make Them Better.
The most dangerous thing about AI in the boardroom is not that it gets things wrong. It is how confidently it gets them wrong — and how quickly we mistake fluency for judgement.

Oleksii Novodvorskyi
Managing Director · Nashville, TN

In the last eighteen months I have sat in more meetings where someone read aloud a paragraph from a language model and then said "so that is the answer" than I can count. The paragraph was always fluent. It was frequently wrong. And the speed of it — the way a confident answer arrived before anyone had finished framing the question — did something to the room. It shortened the thinking.
AI is extraordinary at the part of a decision that is about assembling information: summarising ten documents, drafting the options, surfacing the precedent nobody remembered. It is dangerous at the part that is about weighing what those options cost real people in real conditions. The first part feels like work. The second part is the work.
The teams I see using AI well have drawn a line, quietly and explicitly, between synthesis and judgement. They let the model do the synthesis — the first draft, the comparison, the "what does the data say" — and they reserve judgement for the humans who will live with the consequences. They treat the model's output as the opening of a meeting, not the minutes of one.
The teams using it badly do the opposite. They paste the question in, paste the answer out, and move on. The model becomes a way to skip the conversation that would have exposed the assumption nobody checked. A decision that used to take three days of argument now takes three minutes of reading, and the organisation cannot tell the difference until quarter-end, when the thing they decided quietly fails.
A few practices have served me well. Never let an AI-generated recommendation reach a decision meeting without a human who owns it — someone whose name is on the outcome and who has read the sources, not just the summary. Ask the model for the options and the trade-offs, then close the laptop and argue. Use AI to pressure-test your thinking by asking it to make the strongest case against your preferred answer; if it changes your mind, you have learned something, and if it does not, you understand your own position better.
Most importantly, watch the speed. When a decision feels suddenly easy because the model wrote a confident paragraph, that is precisely the moment to slow down. Confidence is not correctness. Fluency is not judgement. The value of AI in a leadership context is not that it decides for you — it is that it gives you a sharper first draft of the problem, so the time you save on assembly gets spent on the thinking that actually matters.
The leaders who will benefit most from this technology are the ones who were never in love with their own first drafts anyway. AI gives them more drafts to reject, faster, on the way to a decision they can defend. The ones who will be hurt by it are the ones who were looking for permission to stop thinking. The tool amplifies whichever habit you bring to it.
