Merge Mansion — Moral Dilemmas
Moral Dilemmas was an attempt at a premium UA film produced almost entirely with AI tools, at a fraction of the cost of a comparable live-action or animation piece. I initiated the project, owned it end to end, and led creative direction across concept, script, production, and variant testing.
The concept came out of audience research I’d kick-started earlier on what kind of stories actually engage our target audience. Three signals converged: moral dilemma stories perform well in mainstream TV and film, drama-led animation had been our strongest performance content, and the ReelShort-style short drama category was demonstrating that “every second is drama” drives downloads.
The film is built around a glamorous Mansion gala where every guest has their own drama spilling over the edges. Maddie moves through the crowd toward Grandma at the top of the staircase. Around her: secret affairs, a poisoned drink, a tug-of-war over an urn full of money, a cat burglar on the glass roof. The tagline: Let the games begin.
Three principles shaped how the script was built. The film works as a theme park story ride, with Maddies as a loose navigation device and scenes that function both inside the narrative and as standalone variants. It uses intentional incompleteness — moments cut short before resolving, leaving the viewer wanting to know what happens next. And it avoids continuity traps like long takes and resolution-dependent scenes, which is exactly where current AI generation gets awkward. The production constraint became a creative principle.
The film was produced by a small team using AI generation tools combined with traditional creative direction, editing, sound design, and post-production. A significant part of the work was internal — making the case that AI-driven production is a viable, durable approach rather than a shortcut. The proof had to come from the output.
The early test results showed strong top-of-funnel engagement — high view rates and competitive CTR — but conversion lagged. The diagnosis pointed to a gap between the dramatic promise of the ad and how the gameplay reads as a continuation of that drama.
Two things established beyond the campaign itself. AI-led production at premium quality is achievable now, not in two years, when the methodology pairs strong creative direction and traditional craft with AI’s actual strengths and limitations. And audience research, run as its own workstream, produces concept directions that brainstorming alone won’t.