Using GenAI for short-term, exploitative tasks boosts incremental employee creativity, while exploratory GenAI use fuels radical innovation; employees' ability to integrate knowledge explains these effects and becomes more important as tasks grow complex.

Main Finding

• Distinct GenAI usage patterns differentially affect employee creativity. Exploitative GenAI use (refining and adapting existing knowledge) more strongly predicts incremental creativity, while exploratory GenAI use (seeking novel cross-domain inputs) more strongly predicts radical creativity. These effects operate through employees’ knowledge integration capability and are amplified when task complexity is high.

Key Points

• Theoretic framing: combines the exploration–exploitation framework as the core lens with self‑determination theory as a motivational background to explain how employees’ cognitive orientations toward GenAI shape innovation outcomes.
• Usage patterns:
  • Exploitative GenAI use: applying GenAI to verify, refine, and optimize existing solutions.
  • Exploratory GenAI use: using GenAI to acquire novel knowledge, experiment, and probe unfamiliar domains.
• Mediator: Knowledge integration capability (ability to synthesize, recombine, and reconfigure heterogeneous knowledge) is the central cognitive mechanism linking GenAI use to both incremental and radical creativity.
• Moderator: Task complexity positively moderates the effect of both exploitative and exploratory GenAI use on knowledge integration capability—GenAI’s empowerment effects are stronger under high‑complexity tasks.
• Empirical results (summary):
• Exploitative GenAI use → stronger positive association with incremental than radical creativity.
• Exploratory GenAI use → stronger positive association with radical than incremental creativity.
• Knowledge integration capability is a critical complementary mediator for these relationships.
• Task complexity strengthens the links from GenAI usage patterns to knowledge integration capability.

Data & Methods

• Sample: 381 matched employees from knowledge‑intensive firms in China.
• Design: Three‑wave time‑lagged survey (temporal separation to reduce common‑method bias).
• Key variables (survey measures described in paper): exploitative GenAI use, exploratory GenAI use, knowledge integration capability, incremental creativity, radical creativity, and task complexity.
• Analysis: Tested a moderated‑mediation model linking GenAI usage patterns → knowledge integration capability → incremental/radical creativity, with task complexity as a moderator of the first stage. (Authors report empirical hypothesis tests consistent with the four main results above; specific estimation details reported in the article.)

Implications for AI Economics

• Micro to meso-level innovation dynamics
  • GenAI is not uniformly productivity‑enhancing for innovation: its economic contribution depends on how it is used. Firms that want incremental process/product improvements should prioritize exploitative GenAI workflows; organizations seeking transformative, high‑impact innovation should promote exploratory GenAI use and cross‑domain experimentation.
• Human capital and complementary investments
  • Knowledge integration capability is the key human complement to GenAI. Investments in training, incentives, and team structures that strengthen employees’ integrative skills (synthesis, cross‑domain recombination) will increase the return on GenAI adoption.
• Task design and allocation
  • The economic value of GenAI rises with task complexity. Firms should route complex, open‑ended tasks to workers supported by GenAI and preserve routine tasks for either light GenAI automation or other efficiency measures. This suggests potential gains from task reallocation and job redesign to maximize human–AI complementarities.
• Policy and long‑run innovation growth
  • If widespread GenAI use becomes convergent and predominantly exploitative, there is a risk of homogenized idea streams that may dampen long‑run radical innovation and growth. Policy-makers and firm leaders should therefore encourage exploratory modes (e.g., R&D subsidies, protected time for experimentation) to sustain radical innovation.
• Labor market and managerial strategy
  • Demand will grow for workers skilled in knowledge integration and in managing human–AI co‑creativity. Firms may capture value by combining GenAI tools with organizational practices that preserve employee autonomy and stimulate exploratory behavior.
• Measurement and evaluation
  • Economic assessments of GenAI adoption should distinguish usage patterns (exploitative vs exploratory) rather than relying solely on adoption or frequency metrics; welfare/productivity estimates that ignore this heterogeneity risk misestimating GenAI’s impact.

Suggestions for further economic research: quantify macro implications of shifts in exploratory vs exploitative GenAI use for aggregate innovation rates and growth, examine industry and country heterogeneity, and estimate returns to investments in knowledge‑integration training as complements to GenAI.