What work should a team stop doing before it adopts AI?

Harvard Business Review argues that AI transformation needs a redesign of work, not a headcount target. That becomes pretty plain in a small team. You cannot ask people to keep the same targets, learn a new tool, check its output, and patch the mistakes in the gaps between everything else. Before rollout, I’d want one thing named and taken off the list: a weekly report that shrinks, a meeting that goes away, or a queue the tool owns with a person on call for the exceptions. Give people paid time to learn it while the stakes are low. Otherwise “adoption” just means somebody carries the old job plus the AI cleanup after hours. What would you remove from the team’s week before calling an AI rollout fair?