AD FIELD NOTESAI advertising · the work behind the output

Case files / ai assisted film production

Lipton × Critical Mass: use AI to previsualize the shoot, then put real people in frame

Across Latin America, Critical Mass used Firefly to explore a large set of visual directions before a real-model production—an example of generative AI doing pre-production work rather than replacing the finished shoot.

Campaign period: May 2026Source published: 5 May 2026Archive entry prepared: 15 September 2026

Retrospective coverage, prepared in September 2026 from the original record.

AI reference imagery and a real photoshoot direction for Lipton Iced Tea, shown in an Adobe case-study visual.
Firefly references leading into a real Lipton production, as shown in Adobe’s case study. Lipton / Critical Mass / Adobe; image published by Adobe.

Reported execution: the model made the visual brief tangible

Adobe describes Lipton’s campaign as a multi-market launch across Latin America, with 25 stories under three campaign pillars and seven bottle designs. Firefly’s Structure Reference and Style Reference features helped the team explore composition, lighting, props and casting direction before production.

The case study says the team generated more than 100 storyboards, over 2,000 AI-driven images and 500 prompts. The final campaign used real models; generative imagery shaped the shared visual direction rather than standing in for the finished cast.

Our reading: pre-production is where the handoff becomes inspectable

The valuable artifact is not a polished synthetic person. It is the visible bridge between an exploratory image and a shoot plan: a creative team can show what it means by colour, light, motion and casting before asking a client to approve a production.

That makes the workflow easier to critique. A team can reject a reference, revise the brief or widen its casting conversation before money is committed to the shoot, while still keeping the people in the final work real and accountable.

Production lesson: keep the reference set attached to the decision

A generative previsualization library can become a fast way to hide taste decisions inside a pile of images. The better practice is to retain the selected references, the prompt or direction that made them useful, and the human reason for carrying each choice forward.

For a multi-market campaign, that record can become a lightweight visual constitution: what must stay consistent, what may localize, and what is only exploratory. The speed comes from comparison, not from treating the first plausible image as a final answer.

Evidence boundary: the performance claims remain partner claims

Adobe reports 142% of the original reach objective, a 12% complete-view rate against a 3% benchmark, and a 4.4% brand-association lift in Brazil against a 0.9% industry benchmark. Those figures are useful leads, but the public case study does not provide an independent audit or a separate first-air date.

This archive therefore uses the case-study month as a dated index anchor and separates the reported production workflow from the reported outcomes. It does not infer that the image generation alone caused the performance result.

Sources & limits

Adobe’s 5 May 2026 case study; reach, completion and brand-association figures are Adobe/agency-reported, and the page does not state a separate first-air date.

  1. Adobe — How Lipton and Critical Mass crushed advertising targets with generative AI — 5 May 2026.

Send a correction with the passage and supporting source.

The weekly field note

One useful note for the next brief.

Campaigns worth studying, the work behind them, and questions to bring into your next creative review. Delivered weekly.

How we handle your email