A customer support lead uses the most capable model to draft hundreds of short, routine replies a day, such as order status updates and password-reset instructions, each checked by an agent before sending. Replies are good but slow to generate, and the team wants to speed things up. What is the most appropriate change?

Answer: A — Use a faster, lower-cost model for routine replies, keep the agent check, and reserve the top model for complex cases.

Optimizing the workflow means matching the model to the task: simple, high-volume replies suit a faster, lower-cost model, while the human check stays in place, as in (a). (b) gains speed by removing the safeguard. (c) defaults to the biggest model whatever the task. (d) discards a workflow that's working and only needs right-sizing.