Personalizing Google Shopping for Every City Block
I led production of an AI system that wrote and shipped hyper-local holiday ads for Google Shopping, block by block across New York City, in near real time. I owned scope, budget, timeline, partner coordination, and delivery.
- Client
- Google Shopping
- Role
- Group Executive Producer, Addition
- Year
- 2025
- Focus
- Generative AI, Agent Orchestration, Delivery Operations, Vendor Management
Overview
The brief.
Show that Google Shopping is the best way to find the right products during the most competitive shopping season of the year. The answer was hundreds of ads, each tuned to a specific NYC neighborhood and a moment. Production had to run like a system, not a campaign.
The work.
We built a pipeline that utilized agents to combine Google Maps data, Search trends, and Gemini to write ads specific to each neighborhood, and in some cases, each block. An agentic layer assembled and exported finished video automatically, adjusting each asset for query length and product. The biggest trends and products got a full Times Square treatment, showcasing the hottest products in 3D billboards, with backgrounds generated in Gemini and animated in Flow.
My job was making that system shippable. I scoped a build where the final asset count wasn’t fixed, budgeted across engineering, creative, and production, and held a launch date locked to the holiday season. I was responsible for managing day-to-day producers, engineering, design, content strategy and a team of vendor partners.
Improvements I drove.
Hundreds of variations can’t go through a normal review cycle. Along with proprietary Google data, our agents were trained on hand-crafted examples, so output stayed on-brand before anyone reviewed it. We created a human-in-the-loop system to provide needed approvals from ideation to copy to visual identity and final assets.
A custom review interface surfaced the strongest options for the team to approve or send back. Approvals shifted from reviewing everything to picking from the best, and that speed let the campaign react to live moments like the Macy’s Thanksgiving Parade as they happened.
Client partnership.
The project team included team members from Addition, R/GA, 3D vendor & post house, and the media partners in charge of the Times Square billboard and LinkNYC screens. I kept scope, dates, and approvals aligned across all three so live-moment content could move from generation to street without stalling.
Outcome.
Hundreds of hyper-local messages ran across hundreds of NYC locations. Google covered the campaign on Think with Google and the Google Cloud blog, framing it as proof that AI is “augmenting” human creativity rather than replacing it.