Abstract for: A Systems Lens on AI‑Driven Job Redesign and Early‑Career Adaptation

Artificial intelligence is transforming early-career work in uneven, self-reinforcing ways. For 24–26-year-old ICT workers, algorithmic workflows and evaluation systems may either intensify workload and surveillance or foster autonomy and learning, allowing small early differences to compound into divergent career trajectories. Yet research rarely examines how AI adoption and organizational governance interact over time to shape workload, trust, and autonomy, it highlights the need for a system dynamics investigation of these evolving mechanisms. We employ a System Dynamics approach combining Behavior-over-Time (BOT) analysis and Causal Loop Diagrams (CLDs). BOT charts are based on historical data, while CLDs integrate insights from a systematic literature review and semi-structured interviews with ICT workers. The model links key variables grounded in both theoretical mechanisms and the lived experiences of employees using AI-driven tools, providing a robust framework to explore dynamic interactions in early-career work. Our s analysis reveals that AI adoption among early-career ICT practitioners in China follows bifurcating pathways shaped by reinforcing and balancing feedback loops. Reinforcing loops where Productivity, Skill Empowerment, and Trust Building drive trajectories of empowerment, autonomy, and positive AI experiences. The Work Intensification Balancing Loop eventually curtails reliance through overload and negative attitudes. Delayed feedback explains why early empowerment often mask mid-term burnout, it highlights the dynamic, non-linear nature of AI integration in careers. These findings suggest that AI adoption is governed by dynamic interacting with national policy and workplace practices. We identify three key organizational levers. First, enhancing algorithm transparency builds employee trust and reduces resistance to AI tools. Second, supporting continuous skill development helps workers integrate AI into daily routines and sustains productivity gains. Finally, dynamic workload management is critical: excessive task intensity can trigger overload and AI rejection, while balanced demands foster adaptive adoption trajectories over time.