430. Cody Coleman. Coactive AI.
Dr. Cody Coleman
September 24, 2026
Listen to full episode :
The promise of AI in advertising isn't just faster workflows. It's the ability to understand content the way people do.
In this episode, Gabriella Mirabelli speaks with Dr. Cody Coleman, founder and CEO of Coactive AI, about why the next frontier of AI isn't simply generating content. It's understanding it.
For decades, advertisers and publishers have spoken different languages. Brands describe campaigns in terms of emotions, narratives, and intent. Publishers organize massive content libraries using genres, keywords, and metadata. Translating between those two worlds has always been manual, time-consuming, and often imprecise.
Multimodal AI changes that equation.
Rather than relying on titles or tags alone, today's models can analyze video, audio, dialogue, imagery, and other signals together, making it possible to identify specific scenes and moments that align with a campaign's objectives. That creates new opportunities for contextual advertising, content discovery, and media operations at a scale that was previously impossible.
Technology alone isn't enough.
Cody explains why general-purpose foundation models often fall short in specialized industries like media and advertising, and why organizations need benchmarks that measure performance on real-world media tasks instead of generic AI capabilities. He also discusses MediaPerf, an effort to establish meaningful standards for evaluating AI systems in media and marketing.
The conversation also explores a critical question facing every organization deploying AI: where should humans remain in the loop?
Rather than replacing human judgment, the most effective AI systems amplify it. Human expertise defines intent, reviews edge cases, validates results, and teaches systems what success looks like. That partnership between people and AI is what transforms a capable model into a production-ready solution.
Along the way, the episode discusses:
Why contextual advertising has struggled to scale until now
How multimodal AI understands content beyond metadata
The role of benchmarks in accelerating AI innovation
Why "good enough" AI accuracy often isn't good enough for enterprise use
How organizations can balance automation with governance and human oversight
Why successful AI deployments increasingly rely on multiple models working together rather than a single foundation model
what changes when a four-to-six-week manual review process compresses into minutes
where organizations need to rebuild internal governance to use these tools well
Dr. Cody Coleman is the Founder and CEO of Coactive AI, an AI infrastructure company helping enterprises search, understand, and activate large-scale content and asset libraries. Under his leadership, Coactive has raised over $54 million from investors including Andreessen Horowitz and Bessemer Venture Partners, and works with enterprise customers including Comcast NBCUniversal, Fandom, and Thomson Reuters. In addition to his work at Coactive, Cody serves on the Consumer Technology Association (CTA) Board of Industry Leaders and the CTA AI Council.
Cody is also a co-creator of DAWNBench and MLPerf, and a founding member of MLCommons, a nonprofit organization dedicated to accelerating machine learning innovation for everyone. His research spans performance benchmarking of hardware and software systems, multimodal AI infrastructure, and computationally efficient methods for active learning and core-set selection.
He earned his PhD in Computer Science from Stanford University, where he worked closely with Matei Zaharia and Fei-Fei Li, and received his MEng and BS in Electrical Engineering and Computer Science from MIT. His work has been supported by the National Science Foundation Graduate Research Fellowship Program, the Stanford DAWN Project, the Open Philanthropy AI Fellowship, and a Facebook Hardware & Software Systems Research Award.
Cody's personal story is featured in Grit by Angela Duckworth (pp. 220-222), highlighting his approach to sustained effort, discipline, and continuous improvement.
Coactive AI is building the contextual intelligence layer for modern media. Its multimodal AI platform scans video and image content once, enabling users to generate intent-based metadata that powers any downstream use case: search and discovery, creative intelligence, content recommendations, contextual advertising, brand suitability, and measurement. The platform works across any content type, including user-generated content, ads, CTV, digital, archive, and live, making it the intelligence foundation that scales with how media businesses operate. Applications include Context Studio for contextual advertising packaging, Dynamic Tags for industry and custom metadata at scale, and the Search and Discovery API for visual content discovery. NBCU is the anchor customer for Context Studio, Fandom the public reference for Dynamic Tags, and Reuters the public reference for Search and Discovery. For more information, visit coactive.ai.
Follow & Learn More:
Dr. Cody Coleman
Coactive AI