Television often starts with uncertainty. Someone may know the mood they want, remember one actor, or need something that fits the time before a game. I explored a voice-forward starting point that did not require the on-screen keyboard. Agentic Search could become useful before a query was complete, combining personalized titles and suggested prompts based on viewing history, household interests, available time, and what was happening now.

The key decision was to make content and conversation work together. Instead of waiting passively for a precise query, the experience could lead with promising titles and short ways to refine them. A recommendation could answer the immediate need while a nearby suggestion invited the viewer to adjust the mood, compare options, or ask for more like it. Each action advanced the search without demanding a perfectly phrased request.

I adapted multimodal patterns established by the broader Alexa+ team for Echo Show and the Alexa app to a television viewed from across the room. Responses needed a constrained reading width, short turns, and unmistakable focus. Suggestion pills had to remain legible while moving with a directional remote. I also explored access to earlier turns so viewers could revisit the path that produced a recommendation.

Input intent shaped the path. Someone who started with voice could receive voice-forward results or continue into a turn-by-turn conversation. Someone typing could remain with the keyboard, progressive suggestions, and remote browsing. The screens did not need to be identical, but the intent needed continuity. Spoken requests, typed refinements, selected titles, and suggestion pills could all inform what happened next.

My contribution focused on early television experience definition, exploration, and prototyping. I presented the direction to VP-level leadership and helped establish the foundational interaction model for conversational Search on Fire TV. The broader Alexa+ team owned the final conversational product. My role was to translate its multimodal principles into a credible television model.

The launch timeline required a smaller first step. With a compressed year-end engineering window ahead of a public milestone, I translated the broader direction into a focused Alexa+ layer that could be added without destabilizing the Search redesign. The handoff specified updated voice education, an animated Alexa+ cue, conversational prompting, and LLM-generated content rows while preserving existing keyboard suggestions and remote navigation. It also identified exactly which layout and motion changes were feasible.

That launch layer sat on top of a durable Search system I owned end to end as lead and sole designer. The production experience delivered progressive results with each keystroke, remote-friendly text entry, localized keyboard patterns, support across markets, and Alexa+ capabilities. It was designed as a long-term product experience, not a temporary bridge, and continues to serve customers with Alexa+ and customers without it.

Text accounted for about 35% of Fire TV Search traffic. In the earlier experience, surveyed ratings were 46% for finding a specific video, 39% for cross-service Search, and 32% for comparing videos across services. In a 15-participant prototype study, 67% completed the specific-content task without friction and rated the proposed experience very easy, compared with 43% for the earlier experience. The proposed prototype recorded no failures for that task, compared with 21% earlier. These findings were not post-launch metrics.

Together, the work established a practical model for agentic Search on television: begin with an informed point of view, let the viewer’s chosen input shape the interaction, and preserve continuity as intent becomes clearer. The production system proved the mechanics at scale. The AI and LLM explorations showed how that foundation could support more personal, conversational discovery without giving up clarity or control.