Let's say the task is automatically aggregating customer support info. First step is collecting incoming emails Second step is summarizing each one Third ste

Let's say the task is automatically aggregating customer support info. First step is collecting incoming emails Second step is summarizing each one Third step is batching by issue/severity Notice how there is tolerance for deviance/error. "An error" looks like coding a ticket red instead of yellow, or slightly misrepresenting what a client said. The overall workflow can still be net positive.