Fifteen customer replies sat in a folder before lunch. Every one had been written in seconds.
The AI on this team drafted customer responses almost instantly. Then every draft waited in a shared folder for a supervisor to review it. By noon the pile held fifteen. Customers waited hours for messages that had taken seconds to write.
Instead of improving the AI, the team looked at the pile and asked why everything was waiting.
The answer was one blanket rule (opens in a new tab): a supervisor reviews every reply, however routine. So they changed the rule. Routine responses matching standard templates went out with a spot check; full review stayed for everything else. The pile dropped from fifteen to three, and response times fell by hours.
They never touched the AI. One rule change cleared the queue.
The same thing in a kitchen
A chef preps twice as fast as anyone in the building. Every plate still waits at the pass for the same manager to check it. Fifteen plates lined up, steam fading, chef watching.
The chef got faster. The kitchen didn’t.
Where to begin
Yesterday’s question (opens in a new tab) was whether the process should look different at all. Today’s is where to start.
Start at the queue: the place where AI output piles up waiting for a human. Drafted emails waiting for approval. Reports waiting to be read. Proposals waiting for a manager to press send.
Look for the bottleneck after the AI step, not before it. Almost everyone looks before, because that’s where the tool is. The queue is where the redesign begins.