I’ve been circling a set of ideas for months, trying to understand what today’s AI actually means for managers and for the organizations we lead. It’s finally clicking into place.
The short version:
these systems don’t think — they talk.
And sometimes they talk beautifully… while being quietly, confidently wrong.
That’s not hypothetical.
The new KaBLE evaluation research shows that LLMs collapse completely when asked to reason about beliefs, values, or identity — the very domains where managers and teams make their most judgment-heavy decisions.
I’ve written a longer piece exploring what this really means for leaders and why it’s more dangerous than it looks.
Here’s the opening section.