An employee uses AI to complete a task in half the time. Another employee spends longer checking AI-generated work, identifies a serious error, and prevents it from reaching a customer. Which employee has demonstrated greater AI proficiency? The answer may seem obvious once the outcomes are known, but an organization's performance measures may tell a different story. If those measures emphasize how often employees use AI, how much work they produce with it, or whether they can demonstrate an efficiency gain, the employee who exercised better judgment could appear less proficient than the one who generated more output.
As organizations begin connecting AI proficiency to performance reviews, compensation, and advancement, they are also shaping what employees believe good performance looks like. The challenge is not simply determining whether employees can use AI. It is deciding which capabilities and outcomes the organization intends to reward, and whether its measures can distinguish meaningful contribution from visible activity.
AI Proficiency Is Becoming a Career Consideration
An October 2026 survey of 500 finance-industry executives, reported by Reuters, found that nearly a quarter of participating firms linked raises, bonuses, or promotions to AI proficiency. Thirty percent required employees to disclose AI use in formal performance reviews, while 42% reported introducing criteria to evaluate AI skills. Although these findings concern financial services rather than employers generally, they illustrate how AI is moving beyond an encouraged work practice and becoming part of how some organizations judge professional contribution.
That transition carries implications beyond the immediate evaluation process. Encouraging employees to experiment with a technology allows considerable flexibility in how they apply it. Making their careers dependent on AI proficiency introduces expectations that must be understood, demonstrated, and assessed consistently. Organizations therefore need to determine not only which capabilities matter, but also how those capabilities contribute to the work employees are responsible for performing.
When Performance Measures Reward the Wrong Behavior
Consider a manager evaluating three employees. One uses AI frequently to produce analyses and presentations, increasing the volume of visible output. Another uses AI selectively because their responsibilities involve sensitive information and decisions requiring substantial verification. A third uses AI to identify an error that would otherwise have led to an expensive operational decision. Each employee may be exercising sound professional judgment, yet a performance system that primarily rewards usage frequency or easily counted productivity could favor the first without establishing that their contribution was more valuable.
Organizations have long struggled to distinguish activity measures from performance outcomes. AI makes that distinction more consequential because it can substantially increase the amount of work produced while changing how that work is created, reviewed, and evaluated. An employee can generate more material without creating additional value, while another can prevent significant loss without producing a comparable volume of output. If the organization measures the wrong things, it may reward behavior that appears productive while undervaluing the judgment that protects performance.
Performance Measures Influence Behavior
Employees adapt to what their organizations measure and reward. When leaders emphasize time saved, employees look for opportunities to demonstrate speed. When AI usage becomes an evaluation criterion, employees have a reason to make that usage visible. These responses are understandable, but they can become counterproductive when the measure itself becomes more important than the outcome it was intended to support.
Employees may begin using AI for tasks where it contributes little value, devote insufficient attention to verifying outputs, or prioritize work that produces easily measurable efficiency gains. Managers may also encounter pressure to compare employees whose responsibilities involve substantially different opportunities and risks. The organization can consequently create conflicting expectations, encouraging greater AI usage while relying on quality, operational, and risk-management practices that require employees to limit or question that usage.
These contradictions cannot be resolved solely by encouraging adoption or providing additional training. They arise from how the organization defines successful performance and which behaviors its management systems reinforce. A measure introduced to accelerate AI adoption can therefore influence decisions, incentives, and working practices in ways leadership did not intend.
Different Roles Require Different Evidence of Capability
The capabilities associated with effective AI use vary considerably across professional responsibilities. An employee drafting routine communications may demonstrate proficiency through reliable and efficient production, while someone responsible for technical decisions or confidential information may demonstrate it through careful evaluation of outputs and appropriate restraint. In other roles, valuable AI capability may involve recognizing opportunities to redesign workflows, identifying limitations in automated recommendations, or preserving accountability when decisions pass between people and technology.
These distinctions make a single organization-wide usage measure an unreliable substitute for role-relevant evaluation. The CFA Institute's 2026 Finance Skills Pulse Survey also emphasizes the combination of AI competence, professional experience, and interpersonal capabilities, reinforcing the importance of considering AI-related skills alongside the broader requirements of professional work.
A meaningful assessment of proficiency should therefore reflect what employees are responsible for achieving and the judgment required to achieve it. Organizations risk creating contradictory expectations when they encourage employees to demonstrate visible AI activity without adequately recognizing situations in which careful verification, selective use, or a decision not to automate represents better performance.
The Organization Must Be Able to Defend Its Decisions
Compensation and promotion decisions influence employees' careers, their confidence in management, and their willingness to invest in developing new capabilities. Introducing AI proficiency into those decisions makes it important for leaders to explain what is being evaluated, why the criteria are relevant, and how employees can demonstrate the expected capabilities.
Differences in opportunity also matter. One employee may have access to approved tools, suitable assignments, and practical guidance, while another may work under technical restrictions or have responsibilities that limit appropriate AI use. Comparing their activity without considering those conditions could produce misleading conclusions about proficiency. A credible evaluation must distinguish an individual's capability from the circumstances under which that capability can be demonstrated.
The National Institute of Standards and Technology's Artificial Intelligence Risk Management Framework emphasizes organizational governance, defined responsibilities, and human oversight in managing AI-related risks. These considerations are relevant when employers determine how AI-supported work should be governed and how consequential judgments will be made. Employees can reasonably be expected to develop capabilities required by their roles, but the organization remains responsible for establishing an appropriate basis for assessing those capabilities.
What Leaders Need to Resolve
Before incorporating AI proficiency into compensation or performance systems, leadership needs to establish what contribution it intends to reward. The relevant outcome may be better work quality, improved decision-making, stronger professional judgment, greater efficiency, or some combination of these. Tool usage can provide useful information, but it cannot automatically establish those outcomes.
Leadership must also determine whether employees have reasonable opportunities to demonstrate the expected capabilities in their actual roles. Differences in responsibilities, access, guidance, and operating constraints should inform how expectations are defined and assessed. Finally, the organization needs clear accountability for consequential evaluations, including the ability to examine questionable conclusions when a measure does not accurately represent the work performed.
These are not reasons to abandon AI-related performance expectations. They are decisions that should accompany the introduction of those expectations, particularly when they influence compensation, advancement, or professional credibility.
The Larger Organizational Consequence
Performance systems do more than evaluate employees. They communicate what the organization values and influence how employees allocate effort, exercise judgment, and respond to competing demands. Introducing AI proficiency into those systems can reinforce valuable capabilities when expectations reflect meaningful contribution and responsible professional practice.
Poorly designed measures can have the opposite effect, encouraging superficial usage, creating conflicting incentives, and overlooking the judgment that makes AI-supported work reliable. An organization could consequently report progress in employee AI proficiency while unintentionally rewarding behaviors that contribute little to its actual performance.
Before making AI proficiency consequential to employees' careers, executives should be able to answer a more fundamental question: Are we rewarding better work, or merely more visible AI use?
References
1. Reuters, Nupur Anand, “Wall Street firms tie bonuses, promotions to AI skills, survey finds,” October 14, 2026. 2. Selby Jennings, USA AI Talent Pulse Report 2026, October 2026. 3. CFA Institute, Finance Skills Pulse Survey, May 2026. 4. National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework (AI RMF 1.0), January 2023.
AI-assistance disclosure: This article was developed with AI assistance. Responsibility for the final content and professional conclusions remains with Change Collaboration.