TiVo has introduced Agent TiVo, a conversational discovery experience for pay-TV and streaming platforms. Viewers can use voice or text, refine results through follow-up questions and move from a recommendation directly to the available programme or service.
The system is designed around entertainment-specific queries including mood, genre, actor, quotation, episode, sports team, source and availability. TiVo says it combines large language models with its existing voice-recognition and discovery technology in a hybrid architecture that routes each request to an appropriate model.
That routing choice is technically important for operators: deterministic catalogue and entitlement functions do not need to be delegated to a general model, while natural-language interpretation can still make fragmented libraries easier to navigate. The platform is white-labelled and intended to extend existing TiVo deployments.
TiVo has not disclosed the models, availability timetable, supported languages or measured error rates. Operators will need to test catalogue grounding, household privacy, entitlement accuracy, unsafe prompts and the handoff from recommendation to playback before treating conversational discovery as a primary navigation layer.
