A Japanese enterprise software company selling into the US market. Strong product, established at home, still building recognition with American buyers. The campaign needed to run across YouTube's full funnel — reach, view, and demand gen — which meant a large, format-diverse asset set rather than one flagship spot.
Two constraints shaped everything. The brand had visual standards but no defined rules for generated content. And the asset count was well beyond what a conventional shoot-and-adapt model produces on a campaign timeline.
A visual brand system, written before production. Palette, illustration style, character treatment, motion language, and a set of AI generation standards defining what qualifies as on brand — including a hard rule that logo, wordmark, and real product UI are never generated, only composited in post.
A production pipeline built against that system. Generation happens inside the brand rules rather than being corrected back toward them afterward, which is what makes the variant count achievable without the brand drifting.
The full funnel package. Reach, view, and demand gen assets across roughly a dozen format and ratio variants, developed from one creative spine and delivered organized for the ad platform.
One asset from the package was selected for head-to-head testing against an incumbent agency's version of the same brief.
Any company selling technical or enterprise product into a market where it isn't yet a household name, with more channel surface area than its production budget can conventionally fill.
The same brand system that governed the video funnel is what display units, social cuts, and campaign landing pages get built against. Extending a campaign into those formats is a scoping question, not a new discovery process — the architecture already exists.