Generative AI Video Production in 3 Days: An Insurer's Case Study

The brief: production speed without brand risk

Generative AI video production is no longer a novelty for brand teams — the open question is how fast a regulated brand can actually ship it. When Wiener Städtische, an Austrian insurer, needed a full omnichannel campaign, the brief was blunt: move faster than a traditional agency sprint without ever trading away brand safety. The answer was a three-day production format built entirely around generative tools.

Most teams only meet the real {{link}} after the first cut is already locked. Wiener Städtische faced the opposite pressure: a strict internal quality bar that any generated asset had to clear before it could touch a media plan. Speed was welcome only if compliance kept pace.

The insurer had already watched generative pilots stall inside other organizations. The pattern was familiar — a flashy prototype, a few impressive frames, then a long quiet gap before anything reached a live channel. The mandate this time was different: finish a complete, brand-ready campaign, not a demo that impressed in a meeting and died in review.

For an insurer, brand-ready carries legal weight that a fashion drop never faces. Any customer-facing asset can become part of a complaint or a regulator's file, so the creative bar is set by the risk team as much as the marketing team. Generative tooling only earns a seat at that table if its output can be documented and defended after the fact.

Most teams only meet the real AI video production bottleneck after the first cut is already locked.

What the GenMedia Promptathon actually did

The work ran on an {{link}} rather than a string of disconnected tools. e-dialog, the agency partner, replaced a traditional workshop with a focused three-day sprint they called a GenMedia Promptathon, staged at Google CloudSpace in Munich with the client's project managers in the room.

More than 100 hours of conceptual preparation and strategic brand analysis preceded the sprint. That upfront work is the part teams skip when they treat generative video as a prompt box. Here it set the guardrails — character references, tone, and the compliance criteria every frame would later be judged against.

During the three days, an interdisciplinary team from creative and data science sat beside the insurer's managers and generated hundreds of hyper-realistic visuals and dynamic video assets. The client was not a passive spectator; project managers approved direction in real time instead of waiting for a finished reel to come back for sign-off.

The data-science presence mattered more than it sounds. Brand consistency for a financial brand is a trust signal, not a stylistic preference, so the team treated character and setting references as structured inputs rather than vibes. Embedding that discipline before generation is what kept the later sprint free of the endless regeneration loops that burn most AI video pilots.

The work ran on an AI-native creative pipeline rather than a string of disconnected tools.

Three days instead of eight weeks

Character consistency came from {{link}}, not from regenerating every frame. By locking references and workflows before generation, the team kept faces, settings, and product details stable across the hundreds of assets the sprint produced.

The result compressed a process that normally runs six to eight weeks into three days. Finished assets spanned three campaign themes in eight formats, including ten- to fifteen-second video and animated content mapped to every stage of the funnel — from awareness through to performance.

That timeline is the headline number, but the more useful lesson is what made it possible. The sprint did not cut corners on strategy; it moved the strategy upstream, into the 100 hours before anyone opened a generation tool, so the three days could be pure execution rather than discovery.

Traditional production would have spent much of those six to eight weeks in booking, shooting, and review cycles that generative methods collapse. The saving was not only calendar time but coordination cost: fewer external vendors, fewer location and talent dependencies, and a single environment where strategy and execution stayed in one room.

Character consistency came from reference-driven AI video, not from regenerating every frame.

Brand compliance was the real gate

Every asset shipped with an {{link}} that the client could inspect. Wiener Städtische's internal quality check was non-negotiable: visuals had to be indistinguishable from conventionally produced creative and ready for display, social, and Performance Max without a human retouch pass.

The campaign reported that all generated visuals passed the insurer's strict internal quality check and were not recognizable as AI-produced. That is the metric regulated brands actually care about — not how fast a frame renders, but whether it survives the compliance gate that protects the brand.

Compliance also shaped the tooling choice. The team ran Google Gemini on the Google Cloud Platform, a controlled environment rather than a scattered set of consumer apps. For a financial-services brand, where provenance and data handling are audit points, that control mattered as much as the output quality.

The not-recognizable-as-AI standard is the one clients should copy hardest. It reframes the goal from impressive demo to indistinguishable quality, which is the only bar a compliance or legal reviewer will sign off. Provenance tooling such as C2PA Content Credentials can then sit underneath that quality bar, labelling how each asset was made without undermining the creative.

Every asset shipped with an AI video creative audit trail that the client could inspect.

The client became a co-producer

The upskilling quietly built the foundations of an {{link}} for future work. Alongside the campaign, Wiener Städtische's team learned prompting methods and AI animation, so the insurer can now run future productions in-house rather than briefing them out every time.

This is the part that reframes the agency relationship. Generative tooling did not replace the client's capability — it transferred some of it. The insurer left the sprint able to maintain momentum between agency engagements, which changes how the retainer is scoped and how quickly the next campaign can start.

Industry data supports the shift. The 2026 CMO Barometer, drawn from 805 marketing leaders across 15 countries, finds that 68 percent see AI as the defining topic of the year and only 12 percent expect agencies to lead on AI-specific skills — brands increasingly treat AI as a capability they must own rather than rent.

Concretely, the retainer changed shape. Instead of paying the agency to own the entire craft, Wiener Städtische now holds part of the capability internally and calls the agency in for the hard, high-judgement parts. That split is more resilient than full outsourcing and cheaper than building the whole function from scratch.

The upskilling quietly built the foundations of an in-house AI video studio for future work.

Generative AI Video Production: What Commercial Teams Should Copy

The Wiener Städtische sprint is a useful template precisely because it is unglamorous. The breakthrough was not a model trick; it was moving strategy, brand analysis, and compliance criteria into a 100-hour prep phase so the generation window could be short and disciplined.

Three moves travel to other teams: brief the brand bar before the first prompt, lock references so consistency does not cost regeneration loops, and keep the client in the room so approvals happen at the speed of production. Teams that do this turn generative AI video production from an experiment into a repeatable sprint.

The broader market is moving the same direction. IAB projects U.S. digital video ad spend to surpass 80 billion dollars in 2026, with two in three buyers already using or planning agentic AI for video campaigns. The brands that win will be the ones who, like Wiener Städtische, treated speed and compliance as the same problem instead of opposing forces.

One caveat: the three-day format is not a license to skip prep. The sprint worked because 100 hours of brand and compliance work preceded it. Teams that open a generation tool first and worry about the brand bar later will reproduce the exact stall this approach was built to avoid. The speed is a payoff, not a shortcut.