Architecting the Agentic Fan Experience
"I redefined the fan experience by transitioning Alexa from a static Q&A bot to a proactive, LLM-powered sports agent. I designed the 'Contextual Threading' logic that allows the AI to anticipate follow-up intent—surfacing deep stats, live updates, and predictive insights to turn a single question into a continuous, immersive discovery loop."
Context: The Evolution of Intelligent Fandom
Challenge
The "Single-Turn" Ceiling
Legacy voice assistants were designed for transactional utility—perfect for setting timers, but failing at complex, open-ended domains like sports. They suffered from the "Single-Turn Ceiling": Alexa could answer "What's the score?" but lacked the context to answer "Why are they losing?" or "Who is injured?" without forcing the user to restart the entire conversation.
Problem Statement
Users treated Alexa as a static search engine rather than an intelligent companion because the interface was a "dead end." The lack of contextual memory and visual persistence meant that 70% of sports sessions ended after just one query, leaving millions of engagement opportunities on the table.
Key Objectives
Unlock "Agentic" Discovery: Transition the system from a reactive "Fetch" model to a proactive "Suggest" model, using LLMs to anticipate the user's next question (e.g., surfacing box scores after a blowout).
Architect Contextual Continuity: Build a "Threaded State" engine that allows users to pivot between games, teams, and players without losing their place in the conversation.
Unify the Ecosystem: Eliminate the "Device Tax" by creating a single, adaptive interaction model that scales from the 5-inch Echo Show to the 65-inch Fire TV.
Roles & Responsibilities
Role: Lead Product Designer (Sports Domain)
Interaction Strategy: I defined the "Conversation-to-UI" logic, determining how verbal intents (like "Show me stats") map to visual states (expanding cards, deep-dive views).
System Architecture: I built the "Live Data Component Library," a set of high-performance UI patterns capable of rendering real-time volatility (scoring plays, clock stoppages) with <200ms latency.
Cross-Functional Collaboration
AI Science & Engineering: Partnered with LLM researchers to define the "Confidence Thresholds" for proactive suggestions—balancing helpfulness with accuracy.
Product Management: Co-authored the PRD (Product Requirements Document) to shift metrics from "Success Rate" (did it answer?) to "Multi-Turn Depth" (did they stay?).
Domain Partnerships: Collaborated with the News, Finance, Local, Entertainment and Music domain teams to ensure my "Live Data" patterns could scale to their use cases (Stock Tickers, Election Results).
Before - Alexa answering a sports related question on Echo Show device.
“If I want to see more stats or details, I have to start a whole new request. Why can’t I just tap and explore?”
“When I switch from my Echo Show to the Alexa app or TV, everything looks different. I have to relearn where things are.”
System Architecture: The Live Data Ecosystem
The old design delivered minimal context with static scores, forcing users to rely on voice for deeper details. I moved beyond static UI kits to engineer a responsive data visualization system. The challenge was to create a component library capable of ingesting high-velocity, real-time data streams (scores, clock ticks, play-by-play) and rendering them with sub-second latency across a fragmented device landscape.
1. Polymorphic Card Logic
I designed "elastic" components that adapt their structural integrity based on game state. A "Live Game" card isn't just a fixed layout; it’s a logic container that dynamically expands to show possession arrows during active play and collapses to a summary view during commercial breaks, ensuring the UI always matches the pulse of the game.
2. Unified Multimodal Tokens
To eliminate design debt, I architected a semantic token system specifically for live sports. By codifying values for "Live State," "Red Zone Alert," and "Final Score" at the platform level, I ensured that a critical game update conveys the same visual urgency on a 5-inch smart clock as it does on a 65-inch Fire TV.
3. Latency-Optimized Visuals
Designing for live sports means designing for speed. I optimized the motion physics and update cycles of the "Live Clock" and "Score Ticker" components to prioritize rendering performance, ensuring that visual updates occur in perfect sync (<200ms) with the audio commentary.
User Journeys: Adapting the Narrative Standard
To ensure a unified platform experience, I extended the "Visual Narrative" framework (which I led for the core Alexa domain) into the high-velocity world of sports. My strategy was to apply the same "temporal logic" to live game data, transforming disjointed Q&A into a seamless narrative flow.
1. From Data Logs to Discovery Paths
I started by analyzing Classic Alexa query logs to identify the most common "Drop-off Points"—moments where users hit a dead end after a single question (e.g., "What's the score?"). I re-architected these high-volume journeys using my "Threaded Context" logic, ensuring that every answer now serves as a bridge to the next logical step (e.g., Score → Box Score → Player Stats).
2. Predictive Discovery
Leveraging the "Active State" patterns I defined for the core system, I implemented predictive logic for sports.
Contextual Anticipation: If a user asks about a blowout game, the system intelligently surfaces League Standings or Playoff Implications rather than just the score, anticipating the user's underlying intent to understand the broader impact of the game.
Seamless Re-Entry: I applied the "Actionable Anchor" pattern to allow users to pivot between a live game and a player deep-dive without breaking the conversational thread, effectively closing the "Intent Loop."
Interaction Strategy: Mapping the Logic of Real-Time Intent
I architected the Interaction Framework for Alexa Sports, moving beyond simple logic flows to design a blueprint for conversational continuity. This map governs how the system maintains context during high-velocity sports journeys—ensuring that as the live game evolves, the interface remains a coherent, responsive, and intuitive partner.
1. Conversational Threading
I defined the "Re-entry Logic" to solve the problem of fragmented sessions. In a sports context, a user might jump from a live score to a player’s injury history and then back to the league standings. I mapped the Contextual Anchors that allow the AI to "remember" previous turns, ensuring that fans can pivot between deep-dive stats and live updates without ever losing their place or needing to repeat their query.
2. Real-Time Multimodal Choreography
Live sports require millisecond precision. I designed the Synchronization Map for voice cues and visual reveals. By defining the precise hand-off between the AI’s verbal delivery and the UI’s data updates, I ensured that "Big Play" notifications and scoring alerts occur with rhythmic precision—maximizing the excitement of the moment while minimizing cognitive dissonance.
3. Resilience for Non-Linear Intent
AI interactions are inherently unpredictable. I designed the system to gracefully handle "Cross-Domain Jumps" and user interruptions. Whether a fan suddenly asks about the weather in the middle of a game summary or interrupts the AI to ask a follow-up about a different team, my interaction logic ensures the system remains resilient—pausing the narrative thread and surfacing the right information without breaking the overall experience.
Validation: Calibrating Perceptual Bandwidth
I conducted a multimodal experiment to solve the "Redundancy Effect"—the cognitive strain caused when users attempt to read verbatim text while simultaneously listening to audio. My goal was to define the optimal "Sensory Ratio" for different types of sports data.
1. The Hypothesis
I posited that "High-Velocity" data (like live scores) requires Visual Dominance to facilitate rapid scanning, while "High-Density" narratives (like injury analysis) require a Textual Anchor to support working memory.
2. The Prototypes
T1: A "Visual-First" layout that suppresses the transcript, forcing the user's focus onto the data visualization for immediate status checks.
T2: A "Text-Forward" layout that displays the full LLM output, supporting users who need to "read to retain" complex, multi-sentence insights.
T1 – Truncated View: Limits TTS text to a maximum of two lines, keeping the UI clean and ensuring the visual focus remains on key financial data.
T2 – Full View: Displays the complete TTS transcript, providing full context for users who prefer reading along or need additional clarity.
3. The Results
Quantitative testing revealed that T1 was the definitive winner for sports-related intents. Users performed 14% faster on information retrieval tasks when the text transcript was suppressed in favor of high-density visual data.
Glanceable Dominance: In high-stakes moments (like a 4th quarter comeback), users found verbatim text to be a "visual anchor" that slowed them down. By prioritizing the scoreboard and live play-by-play visuals, we reduced the time-to-insight.
The "Eyes-on-the-Game" Rule: Testing confirmed that sports fans treat the screen as a secondary companion. T1’s minimal text approach allowed users to stay connected to the live broadcast while using the Echo Show for quick, high-signal status checks.
Final Heuristic: This finding established the platform's "Data Over Text" rule for all real-time domains, ensuring that for sports, the UI stays out of the way of the action.
Impact & Performance
Transitioning Alexa from a "Fetch Engine" to a "Contextual Agent" drove immediate business value. By solving the "Single-Turn Ceiling," we proved that users are willing to engage deeply with voice AI—provided the interface respects their context and anticipates their needs.
1. Breaking the "One-and-Done" Cycle
The "Threaded Context" engine successfully shifted user behavior. We achieved a 15% increase in multi-turn session depth, validating that when the system proactively surfaces the "Next-Best Action," users naturally transition from passive score-checking to active narrative exploration.
2. The "Sticky" Companion
By removing the cognitive friction of "Cold Starts," we drove a 12% lift in Weekly Active Users (WAU) for the sports domain. Qualitative data showed that users began treating Alexa not just as a utility, but as a live companion during games, trusting the system to keep pace with real-time action.
3. Cross-Domain Scalability
The "Live Data" component system set a new engineering standard. Recognized for its performance and volatility management, the library was forked by the Finance and News teams, reducing the design-to-dev cycle for their real-time ticker products by an estimated 40%.
Final Thoughts: Ubiquitous Sports Intelligence
The final experience represents a fundamental shift from a "Voice Assistant" to a Multimodal Presence. By synchronizing real-time data with ambient soundscapes and cinematic motion, we transformed Alexa from a utility into a Living Companion—an intelligent agent that is aware of the game’s pulse and the user’s environment.
1. Ambient Situational Awareness
We didn't just design a screen; we designed an Atmosphere. On the Echo Show, I introduced "Live Moment" soundscapes—subtle audio cues and visual pulses that reflect the energy of the stadium in real-time. This ambient layer ensures that even when the user isn't actively looking at the screen, they remain connected to the game’s "emotional state," creating a sense of ubiquitous intelligence that lives in the room.
2. Device-Aware Immersion
The system intelligently modulates its "Personality" based on the hardware context:
The Cinematic Fan (Fire TV): I designed high-motion, immersive transitions that turn the living room into a private skybox, prioritizing visual drama and real-time overlays that complement the live broadcast.
The Glanceable Partner (Echo Show/Mobile): In the kitchen or on the go, the UI shifts to a high-utility, "low-friction" mode, surfacing only the most critical updates to respect the user’s primary activity while maintaining a clear path to deeper exploration.
3. The Agentic Standard for Fandom
This isn't a dead-end UI; it is a Continuous Exploration Engine. By establishing the "Threaded Interaction" model, we have set a new benchmark for how fans interact with AI. Alexa now functions as a proactive analyst that anticipates the "Next Big Play," ensuring that as the world of sports becomes more data-rich, the fan experience remains effortless, intuitive, and deeply human-centric.