How should leaders and employees divide responsibility during an AI pilot?
Leadership sets the purpose of a pilot, the resources and the limits on what it may do. The employees working inside the process show where the task breaks down and whether the change helps. Both sides need a feedback route in place before the pilot starts.
Teresa Carlson described AI transformation as a change that needs to move from the top of an organization and from the bottom. For a pilot, that splits the decisions leadership has to make from the information employees can give.
Imagine an AI agent that drafts answers to internal policy questions. Leadership has to decide which policy documents count as approved sources and who owns the final answer. The testers need an easy way to report a missing source, a confusing answer or a question that should've gone to a person.
Testers should keep each original question together with the draft answer and the correction, and the process owner reviews those examples with them. "The answers are sometimes wrong" is hard to act on. An example that shows the exact policy paragraph the AI agent missed gives the team something to fix.
Don't ask employees to make up for an unclear design in silence. Give them a way to change it.
What to try this week
Before your next pilot starts, name the person who decides what happens after a tester reports a problem, and set up the place those reports go.