The Specific Event

Disney recently encouraged employees to use AI chatbots to help craft their self-assessments and goal-setting documents for annual performance reviews. This is not a pilot program buried in one division. According to recent reporting, the push has extended broadly across the company's workforce. The move is notable not because AI is being used at Disney, but because of where it is being inserted: into the evaluation of human judgment, professional identity, and career-determining communication. This is a governance decision, and it deserves analysis as one.

Why the Performance Review Is Not a Neutral Document

Performance reviews are a coordination mechanism. They translate individual behavior into organizational signal. In hierarchical coordination, the self-assessment is specifically the moment where an employee demonstrates their capacity to reflect on their own work relative to organizational standards. It is, structurally, a metacognitive task. When Disney instructs employees to use a chatbot to assist with that task, the company is not simply offering a productivity tool. It is changing what the document is supposed to measure.

This connects to a problem I have been thinking about in the context of the Algorithmic Literacy Coordination framework: the difference between folk theories and structural schemas (Kellogg, Valentine, and Christin, 2020). A folk theory of the AI-assisted performance review might be "the AI helps me articulate what I already know about my performance." A structural schema reveals something more uncomfortable. If the goal of the self-assessment is to evaluate an employee's reflective and communicative competence, and that competence is now being partially delegated to a language model, the review measures neither the employee's original reflection nor the AI's output independently. It measures something in between, and the organization may not have a clear theory of what that something is.

The Governance Architecture Is Missing

What Disney has announced is an encouragement, not a policy. That distinction matters enormously. In organizational theory, the difference between an encouraged practice and a governed one is the difference between an informal norm and an institutional arrangement. Rahman (2021) argues that algorithmic systems create invisible constraints on worker behavior that are difficult to contest or even perceive. Disney's situation is, in some ways, the inverse: employees are being nudged toward using AI in a context where the institutional rules about that use are absent or underdeveloped.

Specifically, there are at least three unanswered governance questions embedded in this news. First, will managers be informed whether a self-assessment was AI-assisted? Second, will employees who choose not to use AI be evaluated against a different implicit standard, one calibrated to AI-polished prose? Third, what happens when two employees submit self-assessments that are structurally similar because both used the same underlying model? These are not edge cases. They are the predictable outcomes of deploying an AI tool in a high-stakes, comparative evaluation context without explicit governance.

The Competence Inversion Problem in Enterprise AI

There is a deeper theoretical issue here that connects to my research on adaptive versus routine expertise (Hatano and Inagaki, 1986). Performance reviews, at their best, are exercises in adaptive self-reflection. They require an employee to reason about novel situations, unexpected failures, and non-routine accomplishments. These are precisely the tasks where procedural AI assistance is least well-suited, because the value of the output depends on its authenticity as a record of situated judgment.

Disney's move effectively institutionalizes a form of routine expertise substitution in a domain that requires adaptive expertise. Employees who use AI to craft smoother, more structured self-assessments may produce documents that score higher on surface coherence while containing less actual signal about their adaptive capacity. Managers reading these documents face what Hancock, Naaman, and Levy (2020) describe as the AI-mediated communication problem: the receiver cannot easily determine whether they are responding to a human agent or to a machine-assisted proxy. The evaluation loop closes, but what it measures has shifted without anyone formally deciding to shift it.

What This Signals for Organizational Theory

Disney is not an outlier. It is an early, visible case of a pattern that will become common across large enterprises: AI tools inserted into legacy HR processes without redesigning the measurement architecture those processes depend on. The interesting theoretical question is not whether employees will use AI for their reviews - they will. The question is whether organizations will develop the structural schemas necessary to govern that use, or whether they will accumulate a growing stack of folk theories about what the documents still mean. Based on what we know about how organizations typically respond to technological change, the folk theory stack is the more probable outcome, at least in the short term.

Kellogg, K. C., Valentine, M. A., and Christin, A. (2020). Algorithms at work: The new contested terrain of control. Academy of Management Annals, 14(1), 366-410.

Hatano, G., and Inagaki, K. (1986). Two courses of expertise. Research and clinical center for child development annual report, 8, 27-36.

Hancock, J. T., Naaman, M., and Levy, K. (2020). AI-mediated communication: Definition, research agenda, and ethical considerations. Journal of Computer-Mediated Communication, 25(1), 89-100.

Rahman, H. A. (2021). The invisible cage: Workers' reactivity to opaque algorithmic evaluations. Administrative Science Quarterly, 66(4), 945-988.