One entry point. A lot of opinions. One decision for the advertiser.
After growing Business AI adoption ~70×, one problem remained: advertisers still had to find an AI agent inside a list of creative enhancements.
It was also the wrong home. Creative enhancements changed the ad before someone clicked. Business AI changed the ad and what happened after the click.
With the agent heading to every advertiser, it needed an entry point of its own.
It started with a sprint alongside the team that owns ad creation, then turned into months of competing requirements for the new card.
I kept coming back to one principle: if advertisers can’t control it, don’t make them decide on it. That cut the card down to one choice and gave us room for the part that actually needed explaining: what an AI might say in their name.
Because its prompts and responses change, I couldn’t show advertisers exactly what they’d get. Instead, I designed around examples: sample prompts, an example response, and a preview of how they could appear in the ad.
The new card beat the old entry point in a split test and was chosen for every advertiser.
