The Finding That Should Trouble Anyone Studying Platform Coordination

A striking figure emerged this week in business reporting: 90% of executives say AI has not boosted productivity at their organizations, yet layoffs tied to AI announcements are continuing anyway. This is not a story about AI failing to deliver. It is a story about a specific and well-documented cognitive failure that organizational theory can name precisely: the awareness-capability gap operating at the executive level, producing decisions that are structurally decoupled from measured outcomes.

The Workday SVP quoted in concurrent coverage this week argues that effective AI usage can function as a "superpower" and that younger workers may hold a key advantage in realizing that potential. Both claims are worth scrutinizing. The productivity data suggests something more systematic is wrong than a generational skill gap. When nine out of ten executives report no measurable productivity improvement while simultaneously attributing headcount reductions to AI, the explanatory variable is not individual capability. It is schema failure at the organizational level.

Awareness Without Structure Is Not Literacy

Kellogg, Valentine, and Christin (2020) identified a persistent pattern in algorithmic work environments: workers develop awareness that algorithmic systems govern their outcomes, but this awareness does not translate into improved performance. The mechanism they describe is not ignorance of AI's existence. Executives clearly know AI exists. The failure is the absence of an accurate structural schema for how AI-mediated coordination actually changes the relationship between inputs and outputs. Knowing that AI "should" improve productivity is a folk theory. It describes a belief, not a mechanism.

Gentner's (1983) structure-mapping theory draws a relevant distinction here. Analogical transfer, the kind that produces genuine adaptive expertise, requires that learners map relational structure from one domain to another, not merely surface features. Executives who have absorbed the surface claim that "AI increases efficiency" are operating on feature similarity. They have not mapped the structural conditions under which that claim holds: the redesign of task sequences, the reallocation of human attention, the change in coordination overhead. Without that structural mapping, implementation proceeds on the wrong assumptions, and the productivity gains never materialize.

The Decoupling Problem Is Organizational, Not Individual

What the productivity data reveals is a decoupling between symbolic adoption and technical integration that organizational theorists have studied in other contexts. Firms announce AI adoption because the announcement carries legitimacy in capital markets and with boards. The operational reality, no measurable productivity improvement, runs parallel to the symbolic claim without intersecting it. This is not new behavior. It mirrors patterns identified in institutional theory when organizations adopt formally but implement loosely.

The more troubling signal in this week's reporting is that layoffs are proceeding regardless of whether the productivity premise has been validated. Rahman (2021) describes how algorithmic systems can create asymmetric accountability structures where workers bear the downside of system failures that management attributes to external causes. The current pattern extends that logic: firms are distributing costs associated with an AI transition that has not yet produced the benefits used to justify those costs. The workers being displaced are paying for a productivity improvement that 90% of their executives cannot actually measure.

Why the Workday Framing Compounds the Problem

The "superpower" framing that Workday deploys is analytically interesting because it locates AI value in individual user capability rather than in system design or organizational restructuring. This is consistent with what Hatano and Inagaki (1986) called routine expertise: the belief that procedural facility with a tool is the primary driver of outcomes. It is also precisely the framing that the ALC framework predicts will fail at scale. If AI value is primarily a function of who uses it well, then power-law distributions in outcomes are guaranteed. Most workers will not be the exceptional users, and the aggregate productivity figures will remain flat even as individual outliers perform well.

The argument that younger workers hold a structural advantage deserves a specific empirical question rather than acceptance as a premise: do younger workers have more accurate structural schemas for AI-mediated coordination, or do they simply have less anxiety about using the tools? Sundar (2020) distinguishes between machine experience and machine understanding. Comfort with a technology and accurate mental models of how it functions are different variables. The productivity data suggests that comfort is widespread and accurate structural understanding is not.

What This Week's Data Actually Tests

For anyone working on platform coordination theory, this week's reporting provides a natural experiment at organizational scale. The ALC framework predicts that schema induction, training that targets structural features rather than procedural familiarity, should produce better transfer than tool-specific training. The current enterprise AI deployment pattern is nearly a controlled test of the opposite condition: organizations have invested heavily in procedural familiarity through tool rollouts, training sessions, and licensing, while investing very little in structural schema development. The outcome, 90% reporting no productivity gain, is consistent with the framework's prediction. That is not a proof, but it is a pattern worth taking seriously as the empirical literature develops.