01 What happened
Microsoft says it worked with Kantar to build LINK AI Content Optimizer, an LLM-powered system that scores ads and recommends ways to improve them.
02 Key details
- The LLM-powered system scores ads, makes recommendations and helps generate variants for another round of scoring.
- Microsoft says the LINK AI output for Content Optimizer was simplified to six key dimensions, with scores added for individual scenes and images.
- According to Microsoft, LINK draws on 35 years of ad testing, with a database of 35 million consumer interactions and more than 260,000 ads.
- Microsoft reports that about 80% of requests failed in the first load test; prioritized queues and communication patterns with Dapr Workflow components raised the pass rate to 95% across thousands of requests.
03 Why it matters
The case shows how detailed asset scores and a combination of automated checks and expert feedback can support a repeated optimization cycle.
04 Who it matters to
Advertisers, advertising agencies and machine learning engineers.
Original sourceMicrosoft