NAITSA Case Study | Creative Ops & Strategy

Case Study – AI Optimization

What started as a simple as to update brand tone and voice quickly became an opportunity for major communication efficiency. The NAITSA team had numerous people build out their existing site, creating drift in the branding. While auditing the site, it was quickly apparent that struggled to perform its job. Content was difficult to read. Linking was broken. So the effort transitioned into a CRO audit report and a custom AI agent to deliver consistent brand and voice no matter who produced the work.

The Approach

Website Audit

Using AI, the site was audited for conversion optimization. A full deck of recommendations was provided and briefed to the internal dev team.

Brand Guidelines

Through multiple workshops, a guideline was built to manage brand tone and voice and provide clear directions on the application of AI.

Custom Agents

A custom AI agent was implemented for the internal team to manage. It resulted in massive time savings, cohesive brand content and reduced page length.

Learnings

Working with the entire NAITSA team, it was apparent that everyone had different levels of experience with AI and used it in different way. Some were comfortable with end to end AI generation, while others wanted a guide in creating content for the site. The brand guildelines and AI agent had to be constructed to be approachable no matter the comfort in AI tools. People also had to be confident in the output quality, so prompting guidance was provided along with numerous tests to show a cohesive AI output.

Characters

Using a crochet style prompt, I built out each character through several references in different expressions, outfits and positions.

Props

There needed to be a blend of crochet and real-world props. Everything one way or the other tipped the setting either too literal or too clean.

Setting

Like props, there was a blend of crochet and real-world. For most settings, I had to build them separately from the characters then place them with Nanobanana.

Learnings

Producing shorter animations was much easier as it took fewer shots to deliver a decent output. The longer the video, the more you have to worry about retaining character detail. An important learning through this was the application of audio. Veo3.1 generated solid voice with good sync, but any background sounds or music was a problem. As I was cutting several scenes together, any music would break the continuity and I couldn’t remove it without also losing the voices. So I prompted out the background sounds and applied my own using Elevenlabs to generate things like sirens, organ music or when I needed a character voice over.