There’s a piece by Kevin Novak on what he calls “the confidence transfer”… the slow migration of trust from people to systems, regardless of which one is actually right. His example sticks with you: a 15-year analyst pushes back on an AI forecast, the room turns against her in a way she calls “almost physical,” her judgment gets quietly bypassed. She turns out to be right. By then nobody remembers she warned them.
Here’s the both/and. Deferring to the machine is often the rational call… on a lot of tasks these tools are fast, and their pattern-spotting can be genuinely more insightful than ours. And that same rational deference is quietly draining the human judgment you’ll need the day the machine is confidently wrong. Both true. The dependency is the whole point: the better it gets, the less you practice catching the rare, catastrophic miss.
This isn’t new, by the way. It’s automation bias at machine speed.
Monday step: in your next AI-assisted decision, measure how often your people actually overrule the output. If the answer is basically never, your human-in-the-loop is a rubber stamp, not a control.
Originally posted on LinkedIn.
RationalAI helps organizations turn ideas like this into results: AI strategy and advisory services. Previous article: Token Spend Passing Developer Salaries? Check What Has to Be True First.
