The researchers identified three main dynamics at play. First, the ability to use AI to fill gaps in existing knowledge led workers to take on more and broader responsibilities, such as managers writing code to tackle problems they would previously have outsourced. Second, the ease of starting and finishing new tasks led workers to fit new bits of work into gaps between meetings or larger pieces of work that would once have been breaks or opportunities to chat to a colleague. And third, the ability to hand off tasks to AI agents led to a surge of multitasking, with workers setting a handful of pieces of work in motion and moving back and forth between them as they progressed.