Kantar and Microsoft build AI system to optimize ad content at scale

Kantar, a leader in brand intelligence, has spent 30 years refining methods to measure ad effectiveness. Historically, this involved consumer surveys and focus groups, amassing a database of 35 million interactions and 260,000 ads. The company later developed LINK AI, a machine learning model that scores ad performance quickly, replacing traditional research methods.

With the rise of social media, generative AI, and fragmented audiences, Kantar's clients sought deeper insights. They wanted to optimize ads for specific audiences and generate multiple variants. To meet this demand, Microsoft Frontier Company and Kantar's FDE studio created the Kantar LINK AI Content Optimizer, an LLM-powered system that scales and optimizes content by specifying improvements and predicting new asset scores.

The system uses Kantar's proprietary knowledge to create a learning loop. It defines rubrics for effective ads, scores them across dimensions, and provides targeted recommendations for improvement. The loop includes human feedback and experimentation to refine components, ensuring the optimized ads meet Kantar's standards. The platform runs on Azure, leveraging cloud-native orchestration, elastic scaling, and observability to reduce operational friction.

Ashok Kalidas, Chief AI Scientist at Kantar, emphasizes the shift from evaluation to actionable insights. 'It’s not enough to be a trusted provider of content evaluation, that doesn’t answer the ‘so what?’ question,' he said. The system now helps brands improve low-scoring ads, create personalized versions, and adapt to different platforms, all while scaling efficiently under load.