The Executive Read, Part II: AI, Human Capability, and Performance

In Part I of The Executive Read, we asked a foundational leadership question: What are you really dealing with?

Part II moves from reflection to application. Marie Potempski examines what the answers to that question can reveal about performance, capability, accountability, authority, role clarity, and organizational structure.

A visible performance problem does not automatically reveal its cause. A missed deadline does not prove a lack of discipline. A polished document does not prove comprehension. Strong executive judgment requires leaders to separate observable facts from assumptions, investigate the conditions surrounding the work, and identify the correct point of intervention.

The episode also examines how artificial intelligence is changing the performance picture. Research involving more than 5,000 customer-support agents found that generative AI increased average productivity by nearly 14 percent, with gains of approximately 35 percent among less experienced and lower-performing workers. A separate six-month experiment involving 7,137 knowledge workers across 66 organizations found that active AI users spent approximately two fewer hours per week on email.

These gains are significant—but AI-assisted output and demonstrated human capability are not the same.

As AI becomes embedded in organizational systems, leaders must learn to evaluate two forms of performance: what a person can accomplish with technological assistance and what understanding, attention, judgment, and responsibility the person still retains.

This episode offers an executive-level framework for discerning whether AI is removing unnecessary friction, strengthening human capability, concealing a lack of understanding, or encouraging passive dependence. It also challenges professionals who intend to excel to develop the qualities technology cannot assume for them: sustained attention, sound judgment, emotional steadiness, ethical responsibility, teachability, and disciplined independent thought.

AI can improve speed, access, and presentation. Leadership must still determine whether the final answer is accurate, responsible, and fit for purpose.

IN THIS EPISODE

• Moving from the diagnostic questions in Part I to application
• Separating observation from interpretation
• How role clarity, capability, and working conditions affect performance
• Distinguishing development needs from structural constraints
• Research findings on AI and workplace productivity
• Why polished output does not necessarily demonstrate capability
• Focus, cognitive offloading, and critical engagement
• Assessing AI-assisted performance and retained human capability
• Executive formation in an AI-supported workplace
• Returning to the executive read

EXECUTIVE REFLECTION

Before responding to a performance concern, complete these statements:

What I know as fact is:

What I am currently assuming is:

The level at which the problem appears to be occurring is:

The evidence supporting that conclusion is:

The evidence that could change my conclusion is:

The human capability that must remain visible is:

The appropriate leadership response is:

CLOSING PRINCIPLE

Do not evaluate performance only by the quality of the final output.

Examine what produced it, what the individual understands, what the system permits, what the technology contributed, and who remains capable of carrying responsibility when conditions change.

Stratwell Hub™ develops disciplined decision-makers who value clarity over noise. 

Connect with Marie Potempski on LinkedIn for executive insights. Follow Stratwell Hub on Facebook for new episodes and updates. 

marie@stratwellhub.com

Transcript

SPEAKER_00 0:10

Welcome to the Clarity Podcast by Stratwell Hub. I am Marie Potemsky, founder and host. In part one of the Executive Read, we began with a foundational question. What are you really dealing with? We looked beyond issues and considered whether problems sits at the level of the individual, the role, the relationship, the authority structure, or the larger system. You have now had some time to reflect on the questions that were posed. Today let's roll up our sleeves and get to work. What might your answers mean? How do they translate into performance? Capability, responsibility, and the proper point of intervention. And as an artificial intelligence becomes increasingly integrated into organizations, how do we distinguish between the performance of the person, the performance of the technology, and the combined performance of both? The objective is not merely to identify a problem. That's not what we're here to do today. It is to learn how to read a situation accurately enough to respond well. That is where leadership begins. The questions from part one do not automatically establish the cause of a problem. They identify what must be investigated. A missed deadline does not necessarily by itself prove that someone lacks discipline. An incorrect decision does not automatically prove that the person lacks capability. In conflict certainly does not necessarily mean that two people are simply difficult, and a polished document does not prove that the person who submitted it understands the content. The visible result is evidence, but is not the complete diagnosis. The disciplined leader separates what happened from the meaning being assigned to it. What do I know? What am I assuming? What evidence supports my interpretation? And what evidence might challenge it? This pause is not indecision. It is command. It prevents an emotional reaction, a preferred narrative, or an incomplete impression from becoming an organizational decision. Performance is affected by more than willingness. Research has repeatedly connected role ambiguity with poorer job performance. When people do not understand their responsibilities, execution can deteriorate. Research has repeatedly connected role ambiguity with poor job performance. When people do not understand their responsibilities, priorities, authority, or standards, execution can deteriorate. Goal setting research has also demonstrated that specific, challenging goals generally produce stronger performance than vague instruction to simply do one's best, provided people possess the necessary knowledge. Ability, commitment, and feedback, that qualification does matter. A leader cannot merely announce a demanding objective and then interpret every difficulty as resistance. Before reaching that conclusion, we need to examine the conditions. Was the expected result clearly defined? Was the standard understood? Did the person know which decisions they were authorized to make? Were their priorities consistent? These questions do not weaken accountability, they will protect it. When responsibility is placed accurately, capable people are not blamed for structural disorder, and poor performance is not hidden behind structural excuses. The objective is rightly ordered responsibility. A person can understand an assignment and still lack the knowledge or skill required to complete it. Evidence of a capability gap could include repeated technical errors, difficulty performing essential steps independently, or clear improvement after instruction and supervised practice. But development must be purposeful. Training should not be used simply because it's easier than confronting the actual problem, as we often see in many companies today. Instruction cannot permanently correct conflicting priorities, unavailable information, inadequate systems, unclear authority or procedures that obstruct the work, nor should support come as indefinite substitutes for personal responsibility. A leader must determine, is the person learning? Is the person applying what has been learned? Is performance becoming more reliable? Can the person now operate with less supervision? Development should increase capability, confidence, and responsible independence. It should not create permanent dependence. The executive question is therefore not merely could this person have worked harder? It more likely is what evidence demonstrates whether the results were shaped by effort, capability, clarity, authority, judgment, or the conditions surrounding the work? Artificial intelligence makes this diagnosis more complex. AI can create real productivity gains. In a study involving more than 5,000 customer support agents, access to a generative AI assistant increased average productivity by almost 14%. For the least experienced and lower performing workers, the improvement was approximately 35%. For the most experienced and highly capable workers, the effects were much smaller, and in some cases slightly negative. That distinction is important. AI does not affect every person in the same way. It can transfer practical knowledge to a less experienced worker and help that person operate closer to the level of a stronger performer. But it may offer less value to someone who already possesses advanced judgment, expertise, and efficient methods. A separate six-month field experiment involves 7,137 knowledgeable workers across 66 organizations. Among the employees who actively use the AI tool, time spent on email decline by approximately two hours per week. These are legitimate values. AI can remove friction, accelerate routine work, organize information, and create space for higher value thinking. But the additional space is valuable only when the person knows how to use it. AI assisted output and independent human capabilities are not identical. A person may produce stronger work because technology has removed repetitive or low value effort. That's the augmentation, but a person may also produce work that exceeds his or her actual understanding. That is concealment. The final document may appear polished, the analysis may sound authoritative, the presentation may be professionally organized, but can the person explain the reasoning, identify the assumptions, recognize an error, defend the recommendation when challenged, adapt when the circumstances change. Leadership cannot evaluate competence only by examining the finished output. It must examine the human command behind the output. The stronger the technology becomes, the more important this distinction also becomes. AI allows people to transfer portions of memory, drafting, organization, information retrieval, administration, and analysis to a machine. This is often called cognitive offloading. Used properly, cognitive offloading can preserve attention for judgment, creativity, planning, and decision making. Used carelessly, it can also weaken active engagement. A 2025 study surveyed 319 knowledge workers and collected 936 real-world examples of generative AI use. The researchers found that greater confidence in the AI's ability to perform the task was associated with less reported critical thinking efforts. I'll repeat myself here. The researcher found that greater confidence in the AI's ability to perform the task was associated with less reported critical thinking effort. By contrast, people who had greater confidence in their own ability to evaluate or complete the task reported greater critical engagement. This does not prove, beyond a reasonable doubt, that AI inevitably damages thinking. It tells us, however, that the posture of the user matters. Is the individual actively directing the tool or passively accepting what appears? Is the person bringing knowledge, judgment, and questions to the process or merely waiting for an answer? This is blind thinking. The ability to focus is not demonstrated by staring at a screen or producing more words. It's demonstrated by sustained contact with reality. Can the person remain with the problem long enough to understand it? Can they resist the urge to accept the first convenient response? Can the person notice uncertainty without immediately escaping it? Ultimately, can the person hold competing information, regulate frustration, and continue thinking clearly? Those are leadership capabilities. As AI becomes embedded into organizations, performance should be evaluated at two levels. The first is assisted performance. Can the person use AI efficiently, ethically, and accurately? The second is retained capability and can the person describe the problem before asking AI to solve it? Explain why the final answer is sound. Can the person identify information that AI may be missing? Verify consequential facts against the reliable sources. Recognize when the output conflicts with professional knowledge, organizational standards, or reality. Can the person complete essential parts of the task independently, remain attentive when the work requires sustained thought rather than immediate output? And can the person accept responsibility for the decision rather than attributing it to the technology? These questions are not designed to catch people failing. They are designed to reveal what must be strengthened. A serious organization is not merely identify weakness, it develops people who are willing and able to grow. The person who can examine an error without collapsing into defensiveness has developmental capacity. The person who can receive correction, adjust, and return with stronger work is demonstrating aptitude. And the person who remains teachable while accepting responsibility is becoming very dependable. The standard should not be that every task must be completed without AI. That would ignore a legitimate and increasingly valuable tool. The standard is that AI should augment human capability without making human capability impossible to observe. For routine and low risk workers, great delegation of technology may be appropriate. For consequential decisions, however, stronger human understanding, verification, and situational awareness and control must be intact. The higher the risk, the greater the requirement for demonstrating independent judgment. This is also a training ground for those who want to excel. Excellence will not belong merely to those who use AI fast. It will belong to those who can combine technological fluency with disciplined attention, sound judgment, emotional steadiness, ethical responsibility, and the ability to think independently. AI might help a person reach an answer, but it cannot form the person who must carry the responsibility for that answer. Formation still requires practice, a willingness to be corrected, restraint before action, and the humility to acknowledge what is not yet known. So returning to questions for part one. What are you actually observing? What evidence supports your interpretation? What are you assuming? Are you evaluating the person's capability, the technological capability, or the combined performance, or both? Can the individual explain the reasoning? Can they verify the facts? Identify an error? Can they reproduce the essential thinking? And has AI strengthened human performance or made the underlying human capability more difficult to see? The executive read requires more than reviewing the final results. It requires understanding what produced it. As AI becomes more present in our systems, the responsibility to discern accurately does not diminish. It actually increases exponentially. The objective is not to resist AI. It is to introduce it without losing the attention, discipline, understanding, aptitude, and judgment the organization still depends on. AI can improve speed, but it is not comprehension. AI can improve presentation, but fluency is not accuracy. AI can help produce an answer, but leadership must still determine whether the answer is true, responsible, and fit for the purpose. Thank you for joining me here today. This has been the Clarity Podcast by Stratwell Hub. Remember, protect your standards, strengthen your clarity, and lead strategically.