Generative AI and the Redesign of Knowledge Work

Abstract

Generative AI tools are entering offices faster than organisations can decide how to use them. We follow six professional-services teams for nine months, combining observation, interviews and analysis of work logs. Teams that agreed shared norms for checking outputs, and redistributed tasks according to skill rather than seniority, saved time without a drop in quality. Teams that left use to individual choice produced uneven results and reported more rework. Junior staff gained new responsibilities but had fewer chances to learn through routine tasks, a trade-off managers rarely planned for. The study argues that value from generative AI depends on redesigning roles, review practices and training together. We offer guidance for leaders who want to introduce these tools while protecting skill development and accountability, including clear rules about disclosure, review and responsibility for errors that reach clients.