The Quiet Cost of AI: When Your People Stop Trusting Their Own Judgment

We spend a lot of energy asking how fast AI is improving. I've started asking a different question: what is AI quietly changing about how my people think, decide, and trust themselves?

Because that's the shift I'm actually seeing. Not job losses. Not killer robots. Something quieter and, for a leader, more dangerous — a slow erosion of human confidence and reasoning.

The symptom nobody puts on a risk register

A capable engineer brings me a recommendation. It's good. But when I ask why — what the trade-offs were, what they'd have done differently — the answer gets thin. They moved fast. They just can't fully explain the path.

One leader I respect described it perfectly: "I know the answer… but I don't know if it's my answer anymore."

This isn't laziness, and it isn't dependency in the way we usually mean it. It's what some researchers call atrophied reasoning. When a tool does the analysis, the drafting, and the first-pass decision, the underlying muscle — the ability to reason from first principles and defend a call — gets used less. MIT research on over-reliance on automated decision aids has found the same pattern: short-term performance goes up while long-term learning and understanding quietly go down. You get faster and shallower at the same time.

For a security and technology leader, that should land hard. Our entire value in a crisis is judgment under incomplete information. If AI is making the organisation quicker but less able to reason, we're trading resilience for speed and not writing it down anywhere.

Why this is a leadership problem, not an AI problem

Here's the uncomfortable part. The answer is not to use less AI. I'm not anti-AI — I'm anti-unthinking adoption. The answer is to deliberately protect the human capabilities AI is eroding. And in 2026, three of those capabilities are becoming the real work of leadership.

Judgment, not control. As AI absorbs execution and analysis, the value of a leader moves upstream — to framing the decision well, holding ambiguity without freezing, and knowing when not to reach for the tool. Authority and even deep expertise matter less than they did. Judgment matters more. The most telling thing I hear from leaders who are adapting well isn't a productivity metric; it's "I trust my decisions again." That's the goal.

Discernment, not activity. When information is infinite and every tool makes it cheaper to produce more, the scarce skill becomes choosing less on purpose. I've watched good people confuse activity with impact — more meetings, more dashboards, more output — while the quality of their actual decisions flatlines. Discernment is a leadership capability now, not a personality trait.

Attention as a defended resource. You cannot reason well in a stream of interruptions. Protecting thinking time — real, uninterrupted, offline time — isn't a wellness perk. It's cognitive defense. If attention is where judgment is formed, then an always-on operating rhythm is a direct threat to the quality of every decision your team makes.

What I'm actually doing about it

A few practical moves, because principles without practice are just posts:

1. Pair every AI output with human reasoning. The question I now ask routinely: "What would you do without the tool, and why?" If someone can't answer, we're not ready to act on it.
2. Make decision quality discussable. We review how we decided, not just the outcome. Reasoning that can't survive a question doesn't ship.
3. Protect thinking time like a control. Meeting-free blocks, fewer notifications, and offsites designed around real decisions instead of status updates. I treat my team's attention as a finite organisational asset.
4. Develop people who can challenge the machine. The most valuable person in the room in two years won't be the one who runs the best prompt. It'll be the one who can look at a confident AI answer and say, "I don't think that's right — here's why."

The bottom line

AI doesn't replace leadership. It exposes it. The organisations that win won't be the ones that adopted AI fastest — they'll be the ones whose people still know how to think when the tool is switched off.

That's the capability I'm choosing to protect. I'd argue it's the one worth protecting most.

---

Draft notes:

• Sources behind this: Barry O'Reilly, "Six Counterintuitive Trends for 2026"; Egon Zehnder, "Leadership Outlook 2026"; MIT research on over-reliance on decision aids (verify the specific MIT citation before publishing — O'Reilly's inline stats aren't cleanly sourced).
• Possible LinkedIn hook (first 2 lines): "A capable engineer brought me a good recommendation last week. When I asked why, the answer got thin. That's the AI risk nobody puts on a register."
• Suggested tags for publish: #Leadership #AI #CISO #Judgment