
Is AI Making Us Smarter or Just Faster? A Conversation with Charles Good
Is AI making us better thinkers or simply helping us produce better work? Neuroscientist Paul J. Zak and Charles Good, president of the Institute for Management Studies, explore the growing gap between AI-driven productivity and genuine human capability. Their conversation examines why critical thinking, productive struggle, emotional intelligence, and hands-on learning remain essential for building expertise in an increasingly automated workplace.

Paul Zak, PhD
Founder
Share
Artificial intelligence is transforming how we work, but is it also making us less capable of thinking for ourselves? AI allows us to complete tasks faster, produce sophisticated analyses, and access information almost instantly. Yet these advantages raise an uncomfortable question: If AI does the thinking for us, how do we develop the expertise needed to know when it is wrong?
I recently discussed this question with Charles Good, president of the Institute for Management Studies and writes about judgment, learning science and the output–capability gap at Outlearn to Outperform where readers can also take his AI Capability Diagnostic. Charles argues that organizations face a fundamental choice: use AI simply to increase productivity or use it to build stronger human capabilities. Our conversation explored five important questions about learning, leadership, and the future of work.
Q: Is AI Making Us More Productive but Less Capable?
Paul: You've argued that AI can make people more productive while weakening their judgment. How do we know when AI is augmenting human capability versus replacing the experiences people need to develop it?
Charles: I call this the output–capability gap. AI can make your work look better almost overnight. The reports are sharper, the analysis is clearer and everything suggests you are improving. But here’s the catch. When AI takes over the thinking that once helped you develop sound judgment, your work may improve while your own capabilities stand still. Performance reviews rarely catch this because they measure what you produce, not what you're learning. The gap only becomes apparent only when AI gets something wrong, isn't available, or encounters a problem that requires judgment it can't provide.
So how do you decide what to delegate? I use a simple rule: If the task is administrative, let AI handle it. But if it involves the thinking and judgment you're paid for, work through it yourself first. Then use AI to challenge your assumptions, identify what you missed, and sharpen your reasoning. The goal isn't just to produce better work. It's to become better at doing the work.
Paul: Would it be fair to say that speed is the enemy of learning?
Charles: Often it is. Leaders are constantly in execution mode. They move from one task to another without taking time to reflect or learn from mistakes.
AI makes speed even more seductive. But cognitive science tells us that effective learning requires deliberate effort and what researchers call productive struggle.
Robert and Elizabeth Bjork's research on "desirable difficulties" demonstrates that learning that requires more effort often produces better long-term retention. You cannot shortcut the process of developing genuine expertise.
Paul: So mastery requires some struggle.
Charles: Exactly. The struggle is part of what makes learning stick.
Q: How Can Leaders Tell Whether Their Judgment Is Improving?
Paul: You've written that confidence and memory are poor measures of good judgment. What should leaders measure instead?
Charles: The uncomfortable part is that you can't always rely on people to recognize when it's happening. Those who lean most heavily on AI may be the least likely to notice their growing dependence. When the work looks polished and the answers come easily, it's natural to feel more capable, even when you're doing less of the thinking yourself. That's why simply asking people whether their skills are improving won't tell you much.
I've found two practices especially useful. First, write down your recommendation before turning to AI. Then use the tool to challenge your thinking, poke holes in your assumptions, and surface what you might have missed. Second is what I call the tool-off test. Once a month, take on something that matters to your work and don’t use AI. Write a decision memo, analyze a problem, or develop a recommendation from scratch. If you find yourself struggling with work that once came naturally, pay attention. Better output can hide declining capability. The only way to know is to test what you can still do on your own.
Paul: I agree, although the importance of the task matters. I'm perfectly happy letting AI produce a first draft of a PowerPoint presentation. Strategic planning for my organization is different.
Charles: Absolutely. I'm not arguing against AI. I'm advocating for a more deliberate strategy of where and when to use AI.
Q: Why Does So Much Leadership Training Fail?
Paul: Organizations spend enormous amounts on leadership training, yet much of it doesn't change behavior. What separates effective training from training people simply enjoy?
Charles: Training fails for the same reason AI use fails: consuming isn’t learning. I call a "cult of content." The assumption is that exposing people to more information will somehow produce better performance.
But consuming information isn't the same as learning.
Someone might proudly announce they've read 50 books this year. My question is: What do you remember, and what are you doing differently?
Research on memory shows that people forget substantial amounts of newly acquired information without reinforcement. Effective training requires retrieval practice, spaced learning, and opportunities to apply knowledge.
The most important part of training is often what happens afterward.
Paul: In my laboratory, we use a simple approach borrowed from medical education: "See one, do one, teach one."
If you can perform something and then teach someone else how to do it, you probably understand it.
Charles: That's closely related to the Feynman technique. Richard Feynman emphasized explaining complex ideas in simple language.
Naming something isn't the same as understanding it.
The goal of training should be observable behavioral change.
Q: Is Emotional Intelligence More Important Than IQ?
Paul: Emotional intelligence has become almost a prerequisite in discussions of good leadership. Is EQ genuinely measurable, and does it predict leadership performance?
Charles: I think it matters more now than ever. As AI makes technical work easier to produce, the ability to work well with people becomes an even greater advantage. At senior levels, technical expertise is often a given. What separates effective leaders is how they build trust, handle difficult conversations, and bring out the best in others. These are the moments when leadership is tested, and no amount of technical expertise can compensate for getting them wrong.
I'm also skeptical of relying too heavily on emotional intelligence scores. Like measures of judgment, many depend on self-reporting, and people tend to rate themselves based on how they think they'd behave rather than what they actually do. I'd rather focus on behaviors we can observe. Can someone listen without getting defensive? Give honest feedback without damaging trust? Stay curious when challenged? These behaviors don't improve simply because someone understands their importance. They improve through practice, honest feedback, and a willingness to change how we respond in difficult moments.
Most leaders know they should listen without becoming defensive, remain curious when challenged, and handle difficult conversations with care. The challenge isn't knowing what good leadership looks like. It's being able to demonstrate it when emotions run high and old habits take over. Because knowing what to say is very different from being able to say it when it matters most.
Paul: Can emotional intelligence be trained?
Charles: Yes, particularly when we break it into component behaviors. Instead of simply telling someone to become more empathic, identify what empathic behavior looks like and give them opportunities to practice it.
The important question is whether their behavior changes.
Q: How Can Organizations Outlearn Their Competition?
Paul: As AI gives everyone access to similar information and analytical capabilities, what allows one organization to learn faster than another?
Charles: The difference is how deliberately organizations develop human capabilities alongside technology.
Organizations that automate everything will experience immediate productivity gains. But over time, their employees can lose the expertise needed to evaluate automated outputs.
This problem was identified decades ago by Lisanne Bainbridge in her paper Ironies of Automation.
As automation takes over more tasks, people become less practiced at performing them. Yet when automation fails, humans are suddenly expected to intervene in precisely the situations requiring the greatest expertise.
Aviation provides an example. Pilots need sufficient manual flying experience to respond effectively when automated systems fail.
AI creates a similar challenge across many professions. If people stop exercising judgment, eventually they lose the ability to distinguish good AI outputs from bad ones.
The organizations that pull ahead won't be the ones that automate the most. They'll be the ones that invest just as deliberately in developing human capabilities as they do in advancing technology. They'll treat human judgment and expertise as critical assets that must be practiced, measured, and maintained, rather than assuming those capabilities will always be there when needed.
Paul: My late colleague Peter Drucker described management as a liberal art. I think AI increases the importance of that perspective. Technical tasks become easier to automate, while human judgment, creativity, and relationships become more valuable.
But critical thinking requires expertise. You cannot evaluate an argument effectively without understanding the subject.
Charles: Exactly. Research comparing novice and expert physics students illustrates this distinction. Novices classify problems by superficial features, while experts recognize the underlying scientific principles.
Critical thinking is not simply a collection of generic rules. It depends on deep knowledge acquired through practice.
The Most Important Skill for the AI Era
Our conversation ended with a question I often ask about education: Why don't universities teach students the science of learning when they first arrive?
Charles believes that learning how to learn is the essential meta-skill.
He pointed to Microsoft's Satya Nadella, who championed moving from a "know-it-all" to a "learn-it-all" culture. Charles calls learning how to learn "the one skill that rules them all."
Charles also emphasized storytelling as a powerful learning tool. Stories organize information into meaningful narratives that people remember.
I agree. AI makes information abundant, but information is not expertise. Expertise requires practice, reflection, feedback, and the willingness to struggle with difficult problems.
The competitive advantage of the future belongs not simply to organizations that use AI most extensively, but to those that continue developing the human capabilities that make AI valuable.
The goal isn't to replace human intelligence. It's to make humans better at using it.
FAQs
1. What is the output-capability gap in AI?
The output-capability gap describes the difference between producing better work with AI and developing the skills necessary to perform that work independently. When AI handles too much of the thinking, employees may become more productive without improving their judgment or expertise.
2. Can relying too heavily on AI weaken critical thinking skills?
It can create that risk. When people consistently delegate reasoning and decision-making to AI, they may lose opportunities to practice the skills needed to evaluate its outputs. Using AI to challenge independently developed ideas can help maintain those abilities.
3. What is the tool-off test, and why is it useful?
The tool-off test is a practice Charles Good recommends in which professionals periodically complete an important task without AI assistance. It provides a way to assess whether their independent reasoning, problem-solving, and decision-making capabilities are being maintained.
4. What makes leadership training effective?
Effective leadership training goes beyond delivering information. It incorporates retrieval practice, spaced learning, opportunities to apply knowledge, and feedback. The ultimate measure of success is observable behavioral change, not simply course completion or learner satisfaction.
5. Why will emotional intelligence remain important as AI advances?
As AI makes technical and analytical work easier to produce, interpersonal abilities such as listening, building trust, navigating conflict, and responding constructively to feedback become increasingly valuable. These capabilities require real-world practice and cannot be developed through information consumption alone.


