Executives from TiVo, Future Today, and Dataxis discuss strategies for managing generative AI costs in CTV discovery by using 'traffic cop' logic to triage queries. The panel highlights the importance of balancing token consumption with latency requirements by reserving LLMs for complex requests while using vector databases and traditional metadata for simpler tasks.
The shift toward hybrid discovery architectures indicates that pure generative AI models are currently too expensive and slow for basic navigation. By implementing a triage layer, platforms can maintain the millisecond response times required for volume and channel controls while offering sophisticated conversational search for complex plot queries. This balance is critical as streaming services face pressure to improve ROI on expensive AI tokens without degrading the user experience. As metadata enrichment becomes more granular, the industry will likely move toward intent-based routing that blends implicit user behavior with real-time semantic search. Watch for whether open-source models become the standard for these intermediate intent-parsing tasks to further reduce reliance on costly proprietary LLMs.
TiVo has long positioned metadata analytics as a core differentiator for content discovery. Chris Ambrozic, vice president and general manager of Discovery at TiVo (a division of Xperi), has stated that churn rates can be reduced between 15-30% at the 15-day mark when personalization strategies are applied to both high- and low-intent consumers, and that operators implementing multi-variate personalization typically see a 35% increase in effective catalog size. Those figures illustrate why TiVo continues to invest in metadata enrichment as the foundation layer that makes AI-driven discovery economically viable at scale.
Future Today, the AVOD platform co-founded by Alok Ranjan, has pursued a complementary strategy by layering contextual intelligence on top of its content library. Future Today integrated with IRIS.TV to assign an IRIS_ID to every video in its library, enabling frame-by-frame AI analysis for contextual and brand-suitability segments, which produced six-figure quarterly revenue increases and a 20% higher eCPM compared to its average. At the IAB Newfronts in April 2026, Future Today announced Audience Advantage, a solution powered by content signals and contextual metadata through partnerships with NOP, Proximic, and IRIS.TV, with VP Tim Ware emphasizing that contextual signals are deterministic and privacy-compliant, requiring no legacy IDs. That deterministic-first philosophy aligns directly with the traffic cop architecture described in the panel, where the majority of queries are handled without invoking expensive generative models.
The broader CTV ecosystem is converging on hybrid architectures that separate deterministic metadata lookups from generative AI inference. Future Today's approach of using AI-powered contextual classification at the video level, combined with traditional metadata for catalog navigation, mirrors the triage logic TiVo employs for search queries. Cinedigm acquired Future Today in a deal valued at $60 million, bringing 5.2 million monthly active users and more than 700 owned-and-operated channels into a single distribution footprint, giving the combined entity the scale needed to amortize AI infrastructure costs across a large content library. As platforms like TiVo and Future Today refine their routing logic, the competitive question becomes which metadata enrichment pipelines can deliver sufficient accuracy at the deterministic layer to minimize the share of queries that require LLM processing.
Add StreamingMeme as a preferred source on Google to see more of our streaming news at the top of your Search results.
Add as preferred source
