AI at work has a double edge: among finance professionals it raises autonomy-driven innovation but also fuels technostress and disengagement; visible, supportive leadership around AI strengthens the upside while blunting the downside.

Main Finding

AI usage has a dual, opposing effect on employee behavior in finance-sector workplaces: it increases innovative work behavior by raising employees’ sense of self-determination, while simultaneously increasing work-disengagement behavior by generating technostress that leads to emotional exhaustion. Leaders’ AI symbolization (leaders’ visible endorsement/use of AI) moderates both paths — amplifying the positive, autonomy-enhancing path and buffering the negative, exhaustion-inducing path.

Key Points

• Theoretical framing: Stimulus–Organism–Response (SOR) model integrating self‑determination theory, technostress concepts, Conservation of Resources theory, and Social Information Processing theory.
• Dual mediators:
  • Cognitive/empowerment path: AI usage → sense of self‑determination (autonomy, competence, relatedness) → innovative work behavior.
  • Affective/depletion path: AI usage → technostress (overload, intrusion, complexity) → emotional exhaustion → work disengagement.
• Moderation by leadership: Leaders’ AI symbolization (explicit leader signals of AI support/use) strengthens the AI → self‑determination link and weakens the AI → emotional exhaustion link.
• Net implication: AI is a “double‑edged sword” at the micro level — can raise innovation but also raise withdrawal/disengagement risks unless organizational leadership and design reduce technostress and support autonomy.
• Conceptual contribution: Integrates cognitive and emotional mechanisms to explain simultaneous opposing behavioral outcomes of AI usage and identifies a managerial boundary condition (leader symbolism).

Data & Methods

• Sample: 285 finance professionals.
• Design: Four-stage longitudinal survey (four waves) — temporal separation used to support mediation/moderation tests.
• Constructs measured (as described by authors): AI usage (generalized intensity of AI use for work), sense of self‑determination (autonomy/competence/relatedness), emotional exhaustion (burnout dimension), innovative work behavior, work disengagement behavior, leaders’ AI symbolization.
• Analytical approach (as reported): mediation and moderated‑mediation analyses within the SOR framework to test the two parallel indirect paths and the moderating effect of leaders’ AI symbolization. (Authors report longitudinal modeling to reduce common-method bias; specific estimation details are in the paper.)

Implications for AI Economics

• Human–AI complementarity is conditional: AI can increase firm-level innovation/productivity by enhancing employee autonomy and competence, but benefits are conditional on organizational context (leadership signaling, training, design). Economic models of AI adoption should include organizational frictions and managerial signaling as determinants of realized productivity gains.
• Hidden costs of AI adoption: Technostress and emotional exhaustion are real labor costs that can reduce labor supply intensity and increase disengagement/turnover. Cost–benefit analyses of AI investments should account for these psychological/behavioral externalities (e.g., lower effective labor supply, reduced quality of effort).
• Role of leadership as a multiplier: Leaders’ visible support for AI acts like a complementary investment (organizational capital) that increases returns to AI capital. Endogenous leadership behavior and organizational culture should be modeled as moderators of AI diffusion and productivity spillovers.
• Policy and firm strategy implications:
  • Firms should invest not only in AI tools but in leadership communication, training, and AI designs that enhance autonomy and reduce complexity to capture positive innovation effects.
  • Labor-market policies and retraining programs should consider emotional/psychological frictions as barriers to effective re‑skilling and AI adoption.
  • Measurement: empirical studies estimating AI’s effect on productivity and wages should incorporate measures of technostress and managerial signaling to avoid biased estimates of complementarities.
• Research directions for AI economics:
  • Quantify long-run impacts on wages, employment churn, and firm-level TFP when accounting for moderated human responses (leadership, training).
  • Heterogeneity across task types and AI functions (assistive vs. surveillance/monitoring vs. decision automation) — different risk profiles for technostress and autonomy effects.
  • Incorporate organizational signaling and psychological costs into diffusion models and into equilibrium models of firm competition where AI adoption decisions have strategic complementarities.

If you want, I can extract the paper’s reported effect sizes, tests, or robustness checks (if provided) and translate them into back-of-the-envelope welfare or productivity estimates for firms.