It started while listening to an episode of Dave Stachowiak’s podcast featuring Dr Leanne ten Brinke, a psychological scientist who studies dark personality traits in the workplace. She shared a statistic that stopped me in my tracks: individuals with psychopathic traits appear in senior management at roughly three times their rate in the general population.

Why do corporate ladders consistently reward people who bring dysfunction with them?

The answer is surprisingly simple: we routinely mistake confidence for competence.

When we evaluate someone under conditions of uncertainty, real competence is difficult to measure on the fly. Our brains default to an easy shortcut: charisma, swagger, and assertiveness. People with dark personality profiles don’t struggle with self-doubt or imposter syndrome. They pitch terrible ideas with the same unyielding conviction as brilliant ones—and hiring committees reward the performance because they assume confidence signals capability.

The Illusion of Symmetry

On a walk recently, I took two photos of the same wild deer. From the side profile, his antlers looked textbook-perfect—balanced, imposing, and sleek. But when he turned and looked at me head-on, the asymmetry was obvious: one side grew noticeably wider and crooked.

The initial impression of flawless symmetry was simply a trick of the viewing angle.

We fall for the exact same trick with technology.

Recently, I revisited notes from The McKinsey Way, particularly around structured problem-solving. It got me thinking about how Generative AI models operate. When an AI generates an answer, it doesn’t stammer, clear its throat, or add an insecure “I think…” It presents information with pristine grammar, structured elegance, and absolute authority.

From one flattering angle, it presents an aesthetic of total perfection. But Large Language Models don’t actually “know” facts; they predict plausible sequences of text. The moment you look head-on—auditing the sources, checking the math, and testing the edge cases—the asymmetries and hallucinations become obvious.

When an LLM hallucinates, it does so with the exact same polished, unshakable confidence as a corporate charlatan. In both cases, we get seduced by fluency, assuming that because something looks symmetrical and articulate, it must be true.

Moving Beyond “Trust, but Verify”

This is where the classic management playbook needs an update.

In The McKinsey Way, structured problem-solving relies on rigorously stress-testing hypotheses against hard data. For decades, leaders have framed this balance using Ronald Reagan’s classic maxim: “Trust, but verify.”

However, in her conversation with Dave Stachowiak, Dr. ten Brinke pointed out why that rule fails when dealing with high-conviction deception—and why it fails just as badly with generative AI:

“As much kindness and compassion as we might show toward dark personalities, there’s one thing we can’t do: trust them… so the old rule of ‘trust but verify’ doesn’t hold. With them, it must be all ‘verify’ all the time.”

— Dr. Leanne ten Brinke

When you extend baseline trust upfront, you anchor on the assumption that the source is credible. Your brain immediately slips into confirmation mode rather than critical audit mode.

Whether evaluating high-stakes leadership decisions or integrating AI into your research pipeline, the operating model has to shift from “trust, but verify” to “verify by default.”

A Few Takeaways for the Work Ahead

• Separate delivery from substance: Charisma is a presentation skill; competence is an execution skill. Never evaluate an executive’s pitch or an AI’s summary based on how smooth it sounds.
• Audit the unsexy details: Treat polished presentations and AI drafts as working hypotheses. Check the math, look at the underlying data, and test the sources directly.
• Keep perspective: It’s easy to look at corporate dysfunction or AI chaos and feel cynical, but Dr. ten Brinke shared an empowering reminder: 80% to 90% of people do not have dark personality traits. The vast majority of people are honest, collaborative, and well-intentioned. The noise comes from a small, hyper-confident minority exploiting our cognitive blind spots.

Fluency is cheap, but rigour takes work. As our tools and workplaces get faster and more articulate, our most valuable skill isn’t speaking with authority—it’s learning to pause, see past the polish, and verify the truth.

How are you building verification habits into your own work—whether dealing with confident colleagues or AI tools? If you’ve found practical ways to cut through the fluency trap, I’d love to hear your thoughts and experiences in the comments.