Editorial
White Paper
The Future of Leadership Development: AI-Driven Manager Enablement
AI’s most consequential limitation is not a technical one. It’s structural. Large language models are trained to predict the most probable next output: to find the center of the distribution, weigh the consensus and optimize for what the data suggests is most likely to be right. This is what makes them extraordinarily capable. It is also what makes them, at a fundamental level, incapable of judgment.
What Judgment Requires
Judgment is not calculation. It is the capacity to weigh competing considerations that cannot all be quantified, to act on moral intuition, to take a position the evidence doesn’t fully support because experience tells you something the data doesn’t. These capabilities depend on emotion — not as a contaminant of rational thinking, but as its engine. Fear signals risk. Discomfort signals misalignment. The felt sense that something is wrong, even when the numbers look right, is not a failure of reasoning. It is reasoning of a kind that AI cannot perform.
The consequences of this limitation become visible when we examine where AI’s consensus-seeking tendency actually leads. Ask a generative AI to invent a new animal, and it will produce a plausible hybrid of existing ones: a flying mammal, a scaled amphibian, some recombination of the familiar. Ask a seven-year-old the same question and you will get something genuinely original, because children have not yet learned to optimize for the probable. The difference is not intelligence. It is the freedom that emotion provides to be wrong, to look foolish, to bet on something that doesn’t fit the pattern.
This dynamic has direct organizational consequences. Research by Doshi and Hauser, published in Science Advances, found that while generative AI raises the floor for less creative individuals, it narrows the ceiling for everyone: stories produced with AI assistance were more similar to each other and to the AI-generated prompts they drew from. A 2026 analysis of AI’s cognitive impact found that billions of people using the same underlying models risks a standardization of thinking patterns: what the authors describe as homogenization of the collective imagination, precisely when organizations need contrarian insight most.
Computational modeling work by Weisberg adds a further dimension: populations with more contrarians, people willing to hold and act on dissenting views, produce wider ranges of knowledge than conformist populations. Contrarianism generates positive externalities. The herd cannot discover what the outlier can.
Editor's Note: Miss the first article in this four-part series? Read now: The Human Advantage: The Question No One Is Asking
The Learning Design Problem
One of the article's authors has spent much of his career applying the discipline of instructional design: using learning science to help people retain and apply knowledge in ways that actually change behavior on the job. The hardest part of that work is rarely the practice itself. It is contending with the widespread misperceptions that subject matter experts and organizational stakeholders hold about how people actually learn.
Those misperceptions are already visible in how AI handles learning design. Prompt an LLM to build a training program from a content document and it will produce something that reflects the most common conventions of how training looks: heavy text, bullet-pointed slides, expert-led instruction, regardless of whether any of those conventions predict successful learning outcomes. The AI reproduces what training usually looks like. It cannot exercise judgment about whether that is what training should look like.
A Case Study of Consensus Thinking
The danger comes not from any single bad AI recommendation, but from the atrophying of the organizational capacity to push back on consensus. And from leadership cultures that never developed that capacity in the first place. Consider what happened inside a large media and entertainment organization when a compensation team set out to solve a real and legitimate problem: employees weren’t advancing fast enough, and the career structure wasn’t giving them enough room to grow. The solution was analytically coherent. The team built a job architecture with twenty-two distinct levels, enough granularity, in theory, to offer meaningful upward movement every year or two without requiring a role change or a significant salary investment. On paper, the logic held. In practice, it was immediately and universally understood to be broken.
What the model couldn’t account for was what promotion actually means to a person. Ideally, a promotion isn’t just a level change. It’s validation — a signal that what you’ve done has earned you something real. Strip that meaning out, reduce a promotion to a bureaucratic reclassification rather than a genuine acknowledgment of growth, and you haven’t solved the advancement problem. You’ve made it worse and added insult to it.
The people closest to that human reality (the managers who had to sit across from employees and explain why this year’s promotion felt like nothing, the facilitators who heard every story about how absurd those conversations had become) understood the failure immediately. That knowledge was not represented in the room where the decision was made. The team was talented, but the culture didn’t reward dissent. So none of it surfaced until the damage was done.
This is the irrational human advantage rendered in negative: not the moment when human judgment saved the day, but the moment when its absence didn’t announce itself until it was too late. AI would have given that compensation team a twenty-two-level architecture and called the problem solved. The judgment that the solution was wrong, that it violated something real about how people experience recognition and advancement, required emotional knowledge the model cannot hold.
Webinar
Jun
17
Engaged, But Not Convinced: What's Eroding Beneath Your Strongest Engagement Numbers
Perceptyx EX Benchmark '26: leadership trust, change confidence and conviction gap truths at large enterprises.
Register
Webinar
On demand
Retention Is a Communication Problem: New Data from 6,000+ Employees
This session gives leaders a clear diagnosis and practical path to treat retention as an infrastructure problem.
Watch Now
Webinar
On demand
Do Learning Programs Really Work? How to Turn Education Into Engagement In Healthcare
See how leaders are using learning programs to build trust with healthcare professionals and create measurable engagement.
Watch Now
Webinar
On demand
Fuel Your Front Line: Workforce Trends Shaping 2026
Get expert insight into the tools and tactics top frontline teams will need in 2026.
Watch Now
Webinar
On demand
Fix the Disconnect Between Projects and Profit
Your margins are hiding in disconnected data. Let's find them.
Watch Now
Webinar
On demand
AI in Customer Service: Faster Resolutions, Happier Customers
Don’t let rising demand burn out your team. See how to build a smarter, more resilient support org.
Watch Now
The Human Bulwark
The implication for organizations is not that AI input should be distrusted. It is that AI input should always travel with a human who is close enough to the actual human experience of the decision to know what the model can’t. That person is not a check on AI’s technical accuracy. They are the check on its most dangerous tendency: the tendency to produce the most probable answer in a situation where the most probable answer is exactly wrong.
In our next post, we introduce a framework for understanding precisely what AI disrupts in the people asked to work alongside it — and what leaders can do to restore it.
Editor's Note: How are other people thinking about the intersection of human capability and AI?
- The Cognitive Economics of AI — AI promised to free up our time, but it's making us busier than ever. Here's why protecting time to think is a leadership imperative.
- The 'Delusion Gap' Between AI Vision and Reality — Leaders need a vision for what AI can do for their company. But they also need a realistic understanding of what it takes to make that vision a reality.
- Designing Human and Technical Architectures for AI-Powered Collaboration — When your data can talk, so can your people.
Learn how you can join our contributor community.