The Sherlock Company automates creative content at scale with Vertex AI and SADA Services

Scaling creative automation in the era of AI

For global brands in media, entertainment, and beyond, visual storytelling is essential to audience engagement. Companies must generate personalized and localized creative assets at an unprecedented scale—without compromising artistic integrity.

Yet managing the creative process across thousands of media variations is complex and labor-intensive. Traditional workflows struggle to keep pace with the sheer volume of assets required to deliver customized experiences across digital platforms. Here’s where The Sherlock Company blends human expertise with automation to help brands manage and deliver on their global creative vision and operations. 

“We are a creative operations agency using services and software to help brands create, localize, manage, and deliver their global content at scale” says Simon Clark, President and CEO of The Sherlock Company. “Our proprietary platform enables us to produce a massive volume of creative assets while maintaining the nuances of design integrity.”

Business challenge

Optimizing thumbnail selection for hyper-personalization

As the demand for hyper-personalized key art and promotional visuals surged from the media and entertainment industry, The Sherlock Company needed a more efficient way to extract high-quality images from full-length video content—without relying on time-consuming manual processes.

Sherlock’s clients—including major streaming services and media giants—wanted to use AI to drive deeper engagement through hyper-personalized key art, but lacked domain expertise. Traditionally, selecting thumbnails for a film or TV series required editors to scour hours of footage, freeze frames, capture high-quality stills, and format them for use across platforms.

“It was so laborious that, in many cases, it simply wasn’t done,” says Clark. “Studios typically rely on preapproved marketing images selected long before a movie’s release. But as demand for personalized visuals grows, brands need a scalable way to extract compelling, on-brand stills from video content.”

Sherlock needed an AI-driven solution to automate thumbnail selection while ensuring each still met creative and storytelling criteria. The company also wanted to develop in-house expertise in AI and machine learning, ensuring a solution that was scalable, repeatable, and adaptable to future creative automation needs.

Solution

Building a multistage AI pipeline with Vertex AI

With these goals in mind, Sherlock collaborated with SADA, An Insight company to develop an AI-powered proof-of-concept (POC) using Google Cloud’s Vertex AI platform and Gemini large language model (LLM).

“We didn’t have deep expertise in AI model selection, but SADA’s recommendations were critical,” says Clark. “They proposed Gemini for its ability to analyze video metadata and generate scene recommendations, paired with Vision AI for post-processing.”

SADA helped Sherlock architect a multiphase AI pipeline comprising:

• Phase 1 (Gemini-powered scene analysis)

• An LLM that analyzes video scripts, metadata, and supplemental data (e.g., actor prominence) to identify compelling scenes
• Custom prompts refined during the POC to ensure outputs aligned with brand guidelines (e.g., “highlight protagonist closeups” or “prioritize action sequences”)
• Phase 2 (frame extraction and Google Cloud Vision AI filtering)
  • Vision API analyzes video segments flagged by Gemini to extract stills
  • Gemini analyzes stills to automatically reject frames with blurriness, off-angle faces, or low visual appeal
• Phase 3 (dynamic template integration)
• Sherlock’s team-built UI to allow clients to adjust prompts and templates dynamically, ensuring flexibility for future campaigns

“SADA’s phased approach showed us that trying to force one model to do everything would fail,” says Clark. “Breaking it into stages let us maintain creative control while scaling exponentially.”

The decision to use Gemini from Vertex AI’s model garden—instead of third-party or self-hosted alternatives—was driven by its seamless integration with Google Cloud tools, cost efficiency, and alignment with Sherlock’s existing infrastructure.

“Gemini allowed us to ground recommendations in contextual data, like character importance or iconic scenes, while Vision AI filtered out unusable frames,” explains Sam Archibald, AI Security Expert at The Sherlock Company.

Preserving creative integrity from design to implementation

SADA experts designed and configured a dedicated Google Cloud project optimized with Cloud Storage, Cloud Run, and Vertex AI services. Beyond development, they facilitated knowledge transfer by hosting workshops post-POC, enabling Sherlock’s team to integrate the pipeline into their proprietary platform independently.

“SADA didn’t just build a POC—they gave us a blueprint,” says Rachel Siecke, VP of Global Operations at The Sherlock Company. “Within six weeks, we had a working prototype. Two weeks later, we integrated it into our UI and started demoing to clients.”

Impact

Shifting from POC to production

Now in full production, Sherlock’s AI-driven workflow ingests video files via Cloud Storage, processes them through the multistage pipeline, and delivers curated stills ready for localization. Sherlock’s UI allows clients to:

• Adjust AI prompts based on regional preferences (e.g., emphasizing specific actors for international markets)
• Monitor pipeline performance in real time
• Export assets directly to marketing platforms

“This process proved that personalization and multiple creative variations can be seamlessly incorporated into our workflow,” says Clark. “It’s opened up a whole new area of potential—not only for us but also for our clients.”

Cutting production time from one day to 10 minutes

Previously, selecting and approving thumbnails required highly trained specialists, consuming up to one day per video. With the new AI-driven workflow, this process now takes just minutes, allowing Sherlock to offer hyper-personalized key art as a scalable service.

“With our application, we can process a load of movies and get high-quality outputs in about 10 minutes,” says Clark. “Before, a person could take a whole day to watch an entire video, select stills, ensure they weren’t blurred, and then get them approved. Now, we can generate multiple variations every day with minimal effort.”

Redefining the economics of creative production

Traditionally, creative agencies and content teams assumed that producing more variations meant higher costs. With AI-driven automation, Sherlock can now generate unlimited creative assets without dramatically increasing expenses.

“This capability could be a paradigm shift in how creative is built,” says Clark. “We could have 10 different versions of a piece of content without significantly raising costs, giving clients far more options and flexibility.”

Transforming the learning experience

Working with SADA and Google Cloud didn’t just result in a new solution—it equipped Sherlock’s team with the knowledge and confidence to explore AI-driven innovation independently.

“Using an analogy, SADA taught us how to fish,” says Archibald. “It wasn’t just about delivering an end product; it was about learning how to do AI workflows ourselves in the future.”

This deeper understanding of AI also led to greater engagement with the broader AI and cloud technology ecosystem. Sherlock was invited to Google Cloud’s Partner Summit, where they connected with industry leaders and gained insights into Google Cloud’s AI roadmap.

“We’re a relatively small company, but at that conference, Google Cloud spent a lot of time with us,” says Clark. “It reinforced that if you’re working on something innovative, you can get the support you need, regardless of your company’s size.”

Achieving key outcomes

Overall, by working with SADA, The Sherlock Company was able to take the AI proof-of-concept that SADA engineered into a production system that enabled them to:

• Reduce thumbnail selection time from one day to 10 minutes, with AI-driven automation
• Establish scalable content generation with unlimited variations at minimal added cost
• Gain deep expertise in AI, empowering future innovation
• Create a foundation for ongoing AI-driven projects with SADA and Google Cloud
• Strengthen industry credibility through hands-on AI implementation and participation in Google Cloud’s AI ecosystem