Human-AI Symbiosis in Software Teams

The paper concludes that AI can reduce routine effort, help explore alternatives, generate candidate tests, and make repository knowledge more accessible but these benefits are conditional on managing incorrect outputs, security vulnerabilities, automation bias, privacy exposure, and integration overhead. It calls for measuring net delivery outcomes, defining decision rights before granting autonomy, and investing in training that strengthens critical evaluation, with future research needed in realistic team settings over longer periods