Voice-Enabled AI for OTT Content Discovery
Enhanced OTT content discovery and user engagement through a voice-enabled generative AI experience that combined LLMs with Quickplay's CMS to enable conversational, intent-driven content discovery on TV.
Role
Lead Designer, Individual Contributor
Skills
Research, Strategy, Interaction Design, Prototyping, Motion
Timeline
6 weeks
Aug - Sept 2023
Collaborators
Designers, Engineers, Client Stakeholders, Product Manager
Platform
Television
AWARDS AND RECOGNITION
International Broadcasting Convention 2023
This project formed part of a collaboration between Quickplay Media and Google Cloud, exploring how generative AI could unlock new opportunities for user engagement and content discovery within the media and entertainment industry.
The work was showcased at the International Broadcasting Convention (IBC) 2023 in Amsterdam. It contributed to Quickplay being named Google Cloud Industry Solution Technology Partner of the Year (2023).
PROBLEM
Content overload is increasing discovery time and decision fatigue on OTT platforms
USER CHALLENGE
As OTT libraries expand, finding the right content becomes increasingly difficult. On average, viewers spend 16 minutes browsing instead of watching. Static rows and keyword search struggle to surface relevant content quickly, leading to decision fatigue, drop-offs and reduced engagement.
BUSINESS CHALLENGE
For OTT platforms, discovery directly impacts engagement, retention, and content ROI. As content volume grows, traditional discovery models fail to scale, limiting monetisation potential and increasing churn risk.
CONSTRAINTS
Television is a lean-back environment, optimised for remote navigation. Input is limited, attention is fragmented, and users expect immediate results. Designing a voice-based AI experience for this ecosystem required navigating CMS data limitations, LLM boundaries, and a six-week delivery timeline ahead of IBC 2023.
BUSINESS GOAL
Drive measurable improvements in content discovery and engagement
🔍
Content Discovery
Increase first-time title plays and reduce time-to-content by enabling intent-led exploration.
📈
User Engagement
Increase average time spent and shared session completion through gamified, collaborative experiences.
MY PROCESS
The UX thinking behind the solution
View my process
SOLUTION
A voice-first AI built for television
Distinct visual states were designed to communicate system status, reduce ambiguity in voice interactions, and build trust.
YOUR OTT AI ASSISTANT
DISCOVERY & ENGAGEMENT
Reducing discovery friction through conversational AI
Voice-first conversational discovery reduces browsing fatigue by guiding users through natural dialogue.
DISCOVERY & ENGAGEMENT
Eliminating decision fatigue for shared viewing
AI blends both viewers’ preferences into shared recommendations, backed by a compatibility score to reinforce trust and promote delight.
DISCOVERY
Context-driven content discovery
AI leverages topical cues like weather, seasons, festivals, current events, and location to surface targeted content.
ENGAGEMENT
Gamified personality quiz to drive engagement
An interactive quiz translates user personality traits into tailored recommendations, making discovery more engaging, personal and memorable.
DISCOVERY
AI-driven personalised recommendations
AI curates real-time recommendations based on user intent and history, using conversational cues to improve discovery.
OUTCOMES
Measured impact of conversational AI-led discovery
Following launch, an 8-week pilot was run across 10% of active users to evaluate the impact of the AI-assisted features. Behavioural engagement metrics were tracked to assess shifts in exploration, session depth, shared viewing outcomes, and decision efficiency.
NEW CONTENT DISCOVERY
+30%
Users interacting with the AI assistant increased their first-time title plays from 8 to 10.4 new titles per user.
DISCOVERY TO CONTENT
3mins
The Topical Sync feature reduced time-to-content from 4.5 mins to 3.0 mins, indicating a 33% decrease in decision time.
ENHANCED USER ENGAGEMENT
9mins/session
Average Time Spent increased by 50%, rising from 6 to 9 minutes among users interacting with the AI assistant.
SHARED SESSIONS COMPLETION
+14%
Shared sessions using the Couple Decider feature saw a 14% uplift in completion rate compared to standard browsing.
LEARNINGS
What I learnt about AI-Product Design
Designing an AI-enabled television experience under real platform and time constraints required systems thinking, feasibility alignment, and intentional trade-offs. The project required balancing innovation with technical feasibility, knowing when to refine, simplify or pivot.
Design within constraints
Conversational interaction mental models
Syncing design with development early
CMS data limitations and LLM behavioural boundaries required me to think beyond ideal AI capabilities. Constraints led to clearer system definitions, and more intentional interaction design.
Moving from static, remote-first browsing to a conversational voice interface required new mental models, redefining focus states, feedback loops, and control within a television environment.
Early engineering alignment ensured technical feasibility, prevented scope creep, and enabled rapid iteration within the six-week timeline.
This project strengthened my ability to navigate trade-offs, iterate quickly with cross-functional teams, and ship reliable AI-enabled experiences.