EN
Sector: Healthcare
Task: AI Powered Assessment
Project Highlights:
A healthcare company had 48 hours to evaluate 250+ long-form videos in order to pivot to a new focus audience for their upcoming marketing goals.
Using an AI-powered, automated workflow that blended Python, Gemini, and Google Workspace, we created a scalable, repeatable video audit system.
We generated clear, campaign-specific recommendations with customized scoring and tone-adjusted summaries.
The final delivery was a polished, multi-tab Google Sheet report for easy navigation, quick insights, and strategic decision-making, as well as a repeatable, scalable framework for future audits.
The Challenge:
In today’s content-rich environment, many organizations juggle massive libraries of video assets — this client was one of them. So when they faced an extremely tight timeline to perform an evaluation against their new marketing goals, they called Assemble to help make sense of the madness. With a vast archive spread across YouTube, Vimeo, and internal folders, we had to determine which of their 250+ videos (many over 1-hour in length) were the right fit for their new campaign. So this wasn’t just any review — it was a shift toward a highly specific target.
The real kicker? We had 2 days to do it.
The Approach:
With just 48 hours and a limited budget, a manual audit was out of the question. They needed a fast, scalable, and smart solution that could sift through hours of content to deliver clear, actionable insights.
So we architected an automated, AI-powered workflow that blended Python, Gemini, and Google Workspace into a seamless evaluation system. Unlike typical back-and-forth interactions with Gemini’s browser chat, our Python-based app ran completely unattended, processing video after video without requiring human input.
First, we created a shared Google Sheet and collaborated with the client to define specific evaluation criteria. This included video & audio clarity, production quality, audience alignment, and content reusability. We then mapped out scoring logic and answer formats — everything from “Yes/No” toggles to short summaries and 1-5 ratings — so that AI output would be consistent, clear, and campaign-ready.
With that, our Python tool automatically pulled metadata from YouTube via API, indexed local B-roll and Vimeo files, and prepped everything for analysis. But the real magic happened next. Gemini didn’t just read the transcripts — it evaluated the actual video and audio content.
Using tailored prompts, we could:
- Analyze each video against every criterion
- Deliver answers in the exact format the client needed
- Assign a score out of 100 to gauge suitability
The system allowed us to track progress and pause/resume analysis, avoiding API quota issues and saving time by skipping completed videos. While all the facts were there, it lacked empathy. So we then ran everything through Gemini again to soften any overly direct or harsh feedback — especially for lower-scoring videos — so the results felt constructive and client-friendly.
The Big Wins:
In just 48 hours, we helped the client:
- Review hundreds of videos
- Get clear, campaign-specific recommendations
- Make informed creative decisions without blowing the budget
The final deliverable was a clean, multi-tab report with a summary of all videos, scores, and quick-glance insights; individual tabs for each video, with category-by-category analysis and clear scoring visuals; and easy navigation and color-coded styling for a clean, professional presentation.
What’s even better is that this system is now a repeatable framework for future content audits, so our client is now set for future audits. It was a great opportunity to show how we can engineer high-volume creative solutions that blend AI efficiency with human expertise to help our client scale smarter — and fast.