Utilizing a private instance of GPT, internal efficiencies increased 10X, daily questions grew to 20,000, and responses to more than 6.5 million on peak traffic days. The move also frees up significant time, enabling editors to focus on higher-value work.