Judgement: The One Thing AI Can't Fake (Yet) 

 AI is not always dumb .... Nor is it always smart

Like many of you, we’ve been leaning into AI, using it to speed up research, simplify complex topics, and find tactical clarity. And most of the time, it’s very, very good.

And therein lies the rub: most of the time.

You see, we humans have spent thousands of years learning how to read our environments. In the old days, it was “is that a stick or a snake?” These days, it's more like “should I trust this consultant, researcher, or advisor?”

Over time, we humans have developed a rich set of filters for assessing people’s credibility . We don’t just evaluate what someone says, we read their body language, their tone, their cadence, even the hesitation before they answer.

We expect humans to act like humans. So, when a senior researcher gives us a report, we instinctively review it for subtle errors - not because we think they’re clueless, but because we know everyone has blind spots. 

But here’s where it gets interesting...

When AI generates something that sounds authoritative, we often skip that same scrutiny. We see it as “expert”, and forget that it’s also entirely capable of making really dumb mistakes - like confidently declaring the wrong year, or missing the obvious.

It’s not being malicious. It’s just not human.

Same goes for trust signals. When a human says “it’s going to snow,” we unconsciously parse the tone, body language, and confidence in the delivery. When AI says it? We get a flat, confident answer: “It’s going to snow”. No nuance. No signal strength. Just the words. (Some of my skiing friends will confidently predict snow anytime the temperature might drop below 5C.)

It’s up to us, then, to layer the judgment back in.

Over time, AI will get better at simulating these human cues - but it will take years (and generations of models). In the meantime, we’re at risk of making big decisions based on thin logic and incomplete signals. 

For owners and CEOs, the takeaway is simple: Don’t outsource your judgment. Use AI, it’s a powerful tool. But run its outputs through the same filters you’d apply to any new junior staffer: smart, fast, confident… and occasionally wrong in ways that really matter.

See you next time,

Andrew