This week in AI — 2 stories

This week's stories share a quieter common thread: AI moving from demo to infrastructure. You'll see what that looks like in two very different contexts - a cybersecurity firm that has quietly reengineered its entire threat-response pipeline around AI agents, cutting investigation time from 38 minutes to 89 seconds, and OpenAI making a significant commercial bet by embedding ads directly into ChatGPT's image-generation workflow alongside a full measurement stack. Together, they illustrate how the "what can AI do?" question is giving way to a more consequential one: what does it mean when AI is the product, the platform, and the pipeline?