Every Leadership Skills List Mentions “Critical Thinking.” Almost Nobody Actually Protects It.

Scroll through any leadership development framework published in the last three years and you’ll find the same handful of buzzwords — critical thinking, emotional intelligence, adaptability — listed alongside communication and strategic vision, as though naming a quality were the same as building an organization that actually cultivates it. There’s a specific, well-researched capacity buried inside those vague buzzwords that deserves far more direct attention than it gets, because it’s genuinely the one leadership quality current AI systems cannot replicate — and a growing number of organizations are quietly training it out of their own people through the exact systems meant to improve performance.

The Quality With an Actual Name

Aristotle called it phronesis, usually translated as practical wisdom — the capacity to know which specific action is right in a particular, concrete situation, rather than simply applying a general rule correctly regardless of context. Psychologist Barry Schwartz, in a peer-reviewed analysis published in Research in Organizational Behavior, describes practical wisdom as the skill of improvising around standard procedures when a rule, applied literally, would fail to serve the actual purpose it was designed for — requiring not just knowledge of the rules, but the perceptual and moral skill to recognize the rare situation where following them exactly would be the wrong choice. This isn’t rule-breaking for its own sake. It’s the deeper competence of understanding what a rule is actually for well enough to recognize the specific case it wasn’t built to handle.

This is a genuinely different capacity than “good judgment” used loosely. It requires holding a rule and its underlying purpose simultaneously in mind, noticing when they diverge in a specific instance, and being willing to act on that divergence — with real personal accountability for being wrong — rather than defaulting to whichever choice is easiest to defend afterward.

Why This Sits Outside What AI Can Actually Do

This connects directly to a structural limit worth stating plainly. An AI system optimizes within the objectives and constraints it’s given — it doesn’t hold a genuine stake in whether those objectives were the right ones to pursue in the first place, and it faces no personal consequence for being wrong in a way that shapes how carefully it weighs a genuinely difficult call. Our piece on what humans should never delegate to AI covers this from the accountability angle directly — certain decisions require someone who can actually be held responsible for the outcome, and practical wisdom, as Schwartz describes it, is precisely the skill exercised in the moment that accountability gets tested: the decision to deviate from the standard process because the specific situation genuinely calls for it, made by someone who will personally answer for that choice either way.

This is a meaningfully different claim than “AI makes mistakes.” It’s that the entire mechanism of practical wisdom — weighing a rule’s purpose against a specific situation’s particulars, with genuine personal stakes in getting it right — isn’t something a system without stakes, consequences, or a personal stake in the outcome can meaningfully exercise at all, regardless of how capable it becomes at pattern recognition or language generation.

The Government’s Own Natural Experiment in Removing This Capacity

It’s worth grounding this in a real, large-scale, well-documented case rather than treating it abstractly, and the U.S. federal sentencing system provides exactly that. The U.S. Sentencing Commission’s own reports to Congress found that mandatory minimum sentencing statutes, by design, strip judges of the discretion to weigh the specific circumstances of an individual case, and concluded that rather than eliminating disparity and unfairness, mandatory minimums simply shifted sentencing discretion from judges, operating in open court with a documented reasoning process, to prosecutors making charging decisions with far less transparency and accountability. The rule-based system didn’t remove the need for judgment. It relocated that judgment somewhere with less visibility and less accountability attached to it.

This is Schwartz’s thesis playing out at the scale of an entire national policy: replacing a professional’s contextual judgment with a rigid rule doesn’t produce the fairness or consistency it promises. It just moves the actual judgment call to wherever the rule has a gap, usually to someone with less training, less transparency, and less direct accountability for getting it right — a pattern worth taking seriously well beyond criminal sentencing specifically, because it’s the same structural failure mode organizations create every time they replace a trained professional’s contextual judgment with an inflexible policy.

How Ordinary Organizations Do the Exact Same Thing

Most companies never pass a law, but they build the functional equivalent constantly: rigid customer service scripts that leave no room for a representative to recognize a genuinely unusual situation, zero-exception refund policies that treat every case identically regardless of context, and KPI systems that reward hitting a number regardless of whether hitting it, in a specific instance, actually served the goal the number was meant to measure. Each of these, examined individually, looks like a reasonable consistency measure. Collectively, following the exact pattern Schwartz describes, they train employees to stop exercising judgment at all, because judgment that deviates from the script carries personal risk with no organizational cover, while blind compliance, even when it produces an obviously bad outcome for a specific customer or situation, is always defensible.

This connects to a related theme worth drawing out directly. Our piece on the psychological safety that makes team rituals like retrospectives and postmortems actually function covers a closely related mechanism: Google’s own research found that teams where mistakes were “held against” people performed worse, precisely because that dynamic teaches people to protect themselves rather than exercise genuine judgment. An organization that punishes a well-reasoned deviation from policy exactly as harshly as a careless mistake is teaching its people the same lesson mandatory minimums taught prosecutors — that visible discretion is riskier than invisible, defensible compliance, even when compliance produces the worse outcome.

Why This Matters More, Not Less, in an AI-Integrated Workplace

This capacity is becoming more valuable precisely as AI absorbs more of the routine, rule-following layer of work. Our piece on the skills becoming more valuable in an AI-integrated workplace found that analytical thinking and judgment under ambiguity top employer surveys precisely because AI is absorbing the well-defined, procedural tasks that used to fill a large share of most jobs — leaving the remaining human contribution disproportionately weighted toward exactly the kind of context-sensitive judgment call practical wisdom describes. An organization that’s simultaneously automating routine work and training its remaining people to avoid discretionary judgment is optimizing itself directly against the one capability the surrounding labor market increasingly rewards.

This also connects to a broader professional development theme. Our guide to building genuine AI fluency alongside real depth of expertise argues that the professionals holding up best combine domain depth with the judgment to apply it flexibly across shifting contexts — which is, in practice, close to a modern description of practical wisdom itself, applied to career strategy rather than moral decision-making specifically.

What Leaders Can Actually Do About It

Protecting this capacity requires deliberate structural choices, not just verbal encouragement to “use good judgment,” since Schwartz’s own research is explicit that incentive systems and rigid rules actively crowd this quality out rather than remaining neutral toward it. Creating explicit, visible permission to deviate from standard procedure when a specific situation genuinely calls for it — provided the person can clearly articulate their reasoning afterward — signals that discretion carries less career risk than defensible compliance, reversing the exact incentive mandatory minimums created for prosecutors. Evaluating decisions based on the quality of the reasoning behind them, not just the outcome, protects people who make a genuinely well-reasoned judgment call that doesn’t pan out, which is precisely the distinction Google’s research found separates psychologically safe, high-performing teams from ones where people learn to protect themselves instead.

A Concrete Example of the Difference in Practice

It helps to see the distinction applied to something ordinary rather than abstract. A customer service policy states that refunds are only issued within thirty days of purchase, no exceptions, and a representative bound rigidly to that script denies a customer who’s clearly a long-time, loyal buyer, asking about a defective product they only just got around to using, thirty-four days after an unusually chaotic month in their own life. The rule was written to prevent abuse of an open-ended refund window, not to punish a genuinely loyal customer over a four-day gap explained by a real defect. A representative exercising practical wisdom recognizes that the specific situation falls outside what the rule was actually designed to guard against, and makes an exception, explaining the reasoning clearly in case anyone asks later.

An AI system handling the same interaction, optimized to enforce the policy as written, has no mechanism for recognizing that distinction unless someone explicitly programmed an exception for it in advance — and if they did, that exception is really just another rule, not a demonstration of judgment in the moment a genuinely novel situation arose. This is the actual gap worth internalizing: AI can execute increasingly sophisticated rules, including rules designed to look flexible, but the moment of recognizing that a situation falls outside every rule anyone thought to write down in advance is precisely where practical wisdom operates, and precisely where it can’t currently be automated away.

A Quick Audit of Your Own Organization

A useful, honest exercise: think of the last time someone on your team deviated from a standard process because the specific situation genuinely called for it. Were they rewarded for the good judgment, punished regardless of the outcome because they’d broken protocol, or did the deviation simply go unnoticed either way? Most organizations, examined honestly, will find the second or third answer is closer to the truth than the first, which is exactly the condition Schwartz’s research identifies as quietly training practical wisdom out of an organization’s people, one defensible, rule-following decision at a time.

Frequently Asked Question

Conclusion

Practical wisdom — the capacity to recognize when a rule’s literal application would betray its actual purpose, and to act on that recognition with real personal accountability — is the leadership quality current AI genuinely cannot replicate, because it depends on stakes, consequences, and context-sensitive moral judgment no optimization process actually holds. The U.S. Sentencing Commission’s own findings on mandatory minimum sentencing show what happens at scale when an institution tries to engineer this capacity away with rigid rules instead: the judgment doesn’t disappear, it just relocates somewhere less visible and less accountable.

Most organizations build smaller versions of the same trap through rigid scripts, zero-exception policies, and incentive systems that reward compliance over reasoning — and as AI increasingly absorbs the routine, rule-following layer of work, the organizations that deliberately protect and reward genuine judgment, rather than accidentally training it out of their people, are the ones positioning themselves for exactly the capability the surrounding labor market is already rewarding.