sep oct joom | Page 11

the lack of real-time updates, the actual performance in matching user intentions, and, most significantly, the substantial operational costs. To an untrained eye, the‘ flashy’ quality of explanations— a key benefit of Generative AI— can mask these underlying challenges in a simple example of asking ChatGPT for recommendations,” he adds. BEHAVIOURAL.“ AI has moved well beyond simple‘ More Like This’ functionality when properly implemented,” says NAGRAVISION’ s Pearson.“ The real value lies in behavioural pattern analysis and engagement prediction rather than content similarity matching. Our OpenTV ENTera platform’ s behavioural segmentation creates sophisticated user cohorts that enable nuanced content bundling, directly addressing the two primary drivers of consumer decisions: cost and content.”

“ The‘ Flashy and overwrought’ characterisation absolutely applies to surfacelevel implementations. As we’ ve learned, upgrading UX alone often requires subsequent integration projects, indicating that superficial AI implementations fail to deliver lasting value. The genuine improvement comes from understanding why users engage with content, not just what they consume,” he advises.

“ While traditional recommendation engines rely heavily on viewing history, which can narrow discovery, AI adds a new dimension by forging deeper connections across content using techniques like semantic pairing- identifying thematic, emotional and narrative links between titles- that even an experienced programmer might miss,” advises

Synamedia’ s Bakirtzaki.“ Far from flashy, AI unlocks fresher, more diverse and meaningful suggestions at scale, creating genuine engagement and uncovering value across the full content library.”

“ Many companies claim to be using AI now, but we have been using many different types of algorithms to deliver personalised experiences and increase viewer engagement for many years,” states ThinkAnalytics’ Docherty.“ This has resulted in substantial and measurable impacts across multiple dimensions of content discovery and these are proven through A / B testing on millions of viewers. In addition, understanding the DNA of the content through our agentic AI-powered metadata enrichment helps to deliver increased viewer engagement as part of the ThinkMediaAI platform.” INFRASTRUCTURE.“ I completely agree that it has to be so much more than just a better‘ More Like This’,” says TiVo’ s Ambrozic.“ At Xperi, usage of AI dates back almost a decade, when we realised that traditional collaborative filtering methods had been eclipsed by neural nets in terms of conversion metrics. Then, a few years ago, GenAI exploded into the picture, in terms of backend processing and optimisation techniques. Finally, in the present day, GenAI is ready for its moment for customer facing applications, now that the infrastructure for tool support is present. It is so much more than just better More Like This; it’ s better engagement across the multitude of models and numerous carousels on a screen. More Like This, Because You Watched, Recommended for You, On now for you; the list goes on and on. All use cases can benefit from the right combination of AI & ML modelling, with GenAI LLM interaction.”

“ That being said, there remains expansive room for human editorial and curation creativity. Curators bring a knowledge and understanding of the content that’ s unparalleled. At TiVo, we’ ve seen that business goals— like promoting licensed content or spotlighting tentpole events— often require real-time adjustments that models alone can’ t anticipate. That’ s why our approach at TiVo blends AI with editorial control and continuous optimisation.‘ More Like This’ is a feature. True discovery is a framework— one that adapts to context, content, and commercial intent,” he adds.“ The best discovery systems don’ t just learn— they listen, and now they prompt for continued dialogue.” TRENDS. As to the future, what trends can we expect in terms of personalisation and optimisation?

“ Personalisation will become more precise, conversational, and context aware,” predicts Braun.“ AI will anticipate not just what viewers want to watch, but also when and how, taking mood, schedule, and environment into account. Discovery will move toward natural language and conversational interfaces, freeing audiences from rigid menus and static genre lists.”

“ Recommendation engines will increasingly span multiple services, simplifying choice across a fragmented landscape, while ethical, explainable AI will be essential to maintain viewer trust. The result will be a unified,