Rise has been running buying agents on a supermarket account since June and reconstructs every change from a timestamped audit log. Dept now routes prompts through an AI gateway so staff cannot pick the model; Brainlabs treats token review as existential, and PMG's Alli For You caps daily tokens and forces a human check when someone hits the ceiling.
Why it matters: Ask every partner running agentic buying which model the request actually hits, and whether the log reconstructs the change or just the token bill.
Tinuiti moved AI visibility from a tenth of creator-strategy weight to a quarter, and it now touches 60% of the roster. Dept hands creators exact product names, chapter markers around real questions, and a clean transcript; Go Fish's Lauren Lyster says long-form gets cited and Shorts almost never. Rate cards have not moved.
Why it matters: Rewrite the next creator brief around the category questions models already cite, and pick voices that already show up in those answers.
Mile Marker, Jellyfish, Lifesight, and M+C Saatchi told AdExchanger the $1B run rate is real cash, credits excluded, but spend stays experimental without competitor benchmarks, incrementality APIs, existing-user suppression, and prompt-level trends. M+C already skews toward commercial-intent conversations. Lifesight wants about six months of cross-channel lift before it leaves test.
Why it matters: Keep ChatGPT Ads on a labeled test line until suppression and incrementality land. Early-mover credits are not a core channel.