What should a company show before saying AI made people more productive?

An NBER working paper based on a survey of nearly 750 corporate executives finds a gap worth keeping visible: firms’ perceived AI productivity gains are larger than their measured gains. The authors suggest revenue may take time to show up. That may be true. It also means a manager should not turn an optimistic survey answer into proof that a team got time back. Before a company says AI made people more productive, I’d want the task, the before-and-after measure, the period measured, and what happened to the review and cleanup work. A revenue gain months later and a support team leaving on time are both good outcomes; they are not the same one. What would make that claim useful enough for workers to trust this week?