Fetch.ai and RedSquid TV announced on 21 July what they describe as the first operator-grade agentic AI television platform, embedding autonomous software agents into RedSquid's white-label TV operating system1. A white-label TV operating system is the software layer that telecom and pay-TV operators run on subscribers' set-top boxes under their own branding. Humayun Sheikh, Fetch.ai's founder and chief executive, has invested personally in RedSquid and joined its board.
The agents, as described by the two companies, discover content, compare services, coordinate connected devices, manage subscriptions and complete purchases from natural-language commands, with Edge-AI hooks threaded from chipset to user interface so inference runs on the device rather than in a cloud data centre. The claim rests on a joint announcement relayed by a single trade outlet, with no operator customer named and no deployment disclosed. Treat it as a product launch, not as evidence of adoption.
RedSquid's layer of the stack, rather than its agents, is what makes the announcement worth logging. Agentic AI in this beat has lived in production and in ad buying: Warner Bros. Discovery rebuilt its US ad stack on Amazon Bedrock AgentCore, and Horizon Media and WPP pointed roughly $17bn of annual spend at the Model Context Protocol, an open standard connecting agents to tools and data. Both sit upstream, inside the businesses that make and sell the advertising.
A set-top agent sits downstream, in the household, on the operator's own hardware. Retention economics change when the interface that presents a cancellation flow can also execute one on request. Any operator evaluating this should ask which actions the agent is permitted to take unprompted, and whether that permission list is theirs to set or the platform vendor's.
