Staff whose managers reshape work around AI report more innovation, a 294-employee survey finds

The link ran partly through that confidence. AI self-efficacy was positively associated with innovation (b = 0.30), and the indirect effect of leader AI crafting through it was significant (0.06). The effect on confidence was stronger for employees whose tasks depended heavily on AI (b = 0.61) than for those who depended on it little (b = 0.13).

A second study, a vignette experiment with 382 business-school participants, repeated the confidence finding under controlled conditions. The survey sample came mainly from internet technology firms in three eastern Chinese cities, and innovation was self-reported, so the results may not carry over to other industries.

Study 1 recruited full-time employees through the Credamo platform who frequently used AI systems in daily work. Wave 1 measured leader AI crafting and AI task dependence; wave 2, two weeks later, measured AI self-efficacy and innovative behavior over the prior two weeks. Analysis used PROCESS Models 4 and 7 with controls including gender, age, education, position, tenure, industry and AI exposure time.

Study 2 recruited faculty and students from business schools at two Chinese universities, with 400 completing and 382 valid after attention checks (average age 22.36, 71% female). Participants imagined being a product operations specialist whose manager, Manager Zhang, either actively redesigned workflows with AI or rarely did, in a 2 x 2 design crossed with high or low AI task dependence. Manipulation checks were significant (leader AI crafting means 4.02 vs 2.44).

The authors recommend that organisations develop leaders' AI crafting skills, build employees' AI confidence through task-specific training and guided experimentation, and tailor efforts to how much each role depends on AI. For roles with low AI dependence, they suggest redesigning workflows to integrate AI more before relying on leader modelling.