After Years of Working With AI, Can You Still Lead Without It?
When was the last time you reached for an AI tool without asking yourself whether you needed it?
For most of the leaders I advise, the honest answer is: I can’t remember.
What started as a tool for the occasional task has become something else. Leaders use AI now to write their emails, prepare for hard conversation, and guide decision-making. AI is increasingly integrated and, in some cases, inseparable from how people think, decide, communicate, and lead.
The leaders I talk to are clear about what they have gained. They openly shared that AI has made them faster and more knowledgeable. They have also shared a growing concern that they are losing a part of themselves through it and where the line is between effective AI use and AI overreliance.
Both things are true.
The same technology that expands our capabilities can also erode them.
Now I ask leaders: after years of AI use, can you still think, decide, and lead independently without it?
The Cost of Convenience: AI Overreliance
Generative AI sits in a different category from the technologies that came before it. People don’t just use it. They argue with it. They confide in it. They consult it the way they might consult a trusted colleague. Decades of research on human-computer interaction shows that people apply social rules to systems that talk back, even when they know better.
Research on leaders suggests a similar process is now unfolding with AI: As leaders work closely with the technology, its capabilities begin to feel like extensions of their own competence. That creates a particular risk: AI overreliance.
Whatever becomes part of how you work also becomes hard to work without.
In human relationships, dependency is mutual. The leader–AI relationship works differently. Whatever a leader’s individual inputs contribute to a model’s training, the system does not depend on that particular leader to keep thinking carefully. It produces fluent output regardless. Nothing in the exchange requires the leader to maintain their independent capabilities. Over time, those capabilities can weaken behind an increasingly polished surface. I call this AI erosion.
What Is AI Erosion?
I view AI erosion as the gradual weakening of independent judgment, attention, and skill as a leader becomes dependent on AI. As leaders increasingly depend on AI to think, communicate, and decide, the human judgment that underpins effective leadership can begin to atrophy.
What protects against AI erosion isn’t less AI use. It’s the deliberate cultivation of boundary practices: routines that keep the parts of leadership that should remain yours from drifting into the model. Boundary practices are how leaders use AI deeply without becoming dependent on it in ways that cost them something.
The 3 principles that follow are boundary practices for the 3 dimensions of leadership most exposed to AI erosion: how you listen, how you think, and how you decide.
3 Practices for Leaders To Prevent AI Overreliance
Principle 1: Listen for What AI Cannot Hear
When was the last time you struggled with a hard conversation because you hadn’t yet “run it through ChatGPT”?
If that question is easy to answer, you’re watching AI erosion in your relational life and risking AI overreliance.
Listening is one of the hardest human skills. It takes effort, attention, and the willingness to be present with what someone is saying rather than what you expect them to say. It also degrades with disuse.
Listening is also how leaders stay close to ground truth. Every organization has what I call a reality delta: The gap between what leaders believe is happening and what is actually happening on the ground. AI can widen the gap. It produces clear, polished summaries that feel authoritative but that flatten the reality of how work actually unfolds.
Most leaders are good at brainstorming. Fewer are good at painstorming: surfacing the misaligned priorities, anxieties, inertia, and noise (PAIN) that frustrate people in getting good work done.
Here are 3 practices to enact this principle:
Make painstorming a regular habit.
Add a pain check-in to retrospectives and rollouts. Ask: What feels most important? What worries are showing up? What is hard to act on? What is getting lost in the noise? The most useful answers tend to come in a sigh, a hedge, or a story that runs longer than it should.
Listen for context.
In your next one-on-one, when you ask how things are going, listen for context. Stories are how people experience their work: messy, specific, sometimes contradictory. To close the reality delta, go beyond surface-level updates and ask for context: Walk me through the last time this happened. Tell me about the day you noticed it.
Track what you avoid.
Notice conversations where you felt reluctant to proceed without notes, scripts, or AI preparation. If the count is non-trivial, take it as a signal that the listening muscle is weakening.
Principle 2: Stay With the Hard Question
Think about the hardest question facing your organization right now. How long can you sit with it before you reach for a prompt?
Hard questions require leaders to lead through uncertainty. You carried the question with you and developed as leader in the process. Discomfort was part of the work.
Judgment is a skill, and like most skills it atrophies through disuse. Something subtler also happens as using AI becomes your default mode for problem solving. You stop developing the capacity to work through hard problems.
Here are 3 practices to enact this principle and protect your human judgment:
Block weekly time for unassisted critical thinking.
30 minutes. A notebook. No prompt to react to. Work through the hardest question facing your team using only your own reasoning. The point is not the answer. The point is keeping the muscle in use.
Designate categories you refuse to delegate.
Protect 2 categories by default. On ethics, AI can surface stakeholder impacts and map tradeoffs, but the weighing of values is yours. On people decisions, form your own read of a candidate first. Once you’ve seen what the model thinks, the concept of anchoring means it’s remarkably hard to unsee it.
Stay longer than feels comfortable.
When you reach for AI mid-conversation or mid-decision, pause for a minute What you find on the far side of that minute is often closer to judgment than what arrives in the prompt window.
Principle 3: Lead From Conviction
When was the last time you defended a position that cost you something? Not a small disagreement. A real defense, of a value or a judgment you held, when the easier choice was to let it go.
If the answer is hard to find, conviction may have started to thin without your noticing. AI doesn’t cause that on its own. But it makes the easier choice always available, and the easier choice is what conviction has to stand against.
Conviction has 2 anchors: the values you hold and the self-direction to act on them. Management scholars call this a protean orientation, a concept introduced by Professor Tim Hall at Boston University, to describe individuals who are self-directed, adaptable, and values-driven.
In a review of 3 decades of research on the protean orientation, we found a phenomenon we describe as the protean paradox — people who are self-directed and values-driven tend to contribute more to their organizations. Conviction is what drives contribution in leadership.
A leader’s conviction cannot be generated by AI. The model can produce a recommendation. It cannot stand behind one.
Here are 3 practices to enact this principle:
Write a one-page document of your operating principles and paste it into your AI tool’s system prompt.
Take 30 minutes to write what you stand for as a leader, the calls you will not delegate, and the decisions you insist on making yourself. Then paste it into the custom instructions of the AI tool you use most. Now your principles are present every time you prompt. You don’t have to remember to read them. The tool reads them for you. Update the document once a year, or when something forces you to.
Before any consequential decision, take 60 seconds to write your call before you open AI.
Not a memo. Just 2 sentences in a notebook or a notes app: what you would decide right now, and why. The exercise costs you a minute. The information it gives you over a quarter tells you how AI is sharpening your judgment or replacing it.
Protect a regular recurring block for uninterrupted thinking, without devices.
One recurring block. Phone in another room. Laptop closed. Use it for the hardest open question you are carrying. Research on mindfulness in leadership finds that even modest contemplative practice builds greater self-awareness and clear decision-making.
AI, Human Judgment & the Leaders Who Protect Both
Remember the 3 boundary practices to prevent AI overreliance:
- Listen for what AI cannot hear.
- Stay with the hard question.
- Lead from conviction.
These aren’t arguments against AI. They’re conditions for using AI deeply without becoming dependent on it in ways that compromise integrity.
Build the practices early. AI erosion is self-reinforcing: The less capable you become at independent work, the harder it is to notice. Start while you can still feel the difference between what you produce on your own and what you produce with AI in the loop.
If you lead a team, the most useful thing you can do is build boundary practices into the way the team works. Protected time for unassisted thinking. AI-free working sessions on consequential decisions. A development model that evaluates judgment, not just output.
The leaders who will do well in the AI era aren’t the ones who use AI most. They’re the ones who’ve protected themselves against AI overreliance and can still listen, think, and stand without it.
Ready to Take the Next Step?
The transition to AI is as much a human challenge as a technical one. See how our AI and leadership solutions equip leaders for the operational demands of AI adoption and the hopes, fears, and trust dynamics that determine whether it succeeds.
