Spotify AI DJ Reimagined.

Overview

The Challenge - To address the growing desire for personalized music discovery, a design sprint was initiated by Alex Freeman, user researcher at Spotify: How might Spotify better serve music discovery behaviors using AI on new or existing surfaces based on both implicit and explicit signals? The team was given 2 weeks to complete the challenge.

The Solution - After deep diving into existing surfaces, the team decided to focus on enhancing AI DJ's user experience. A key pain point emerged: lack of explanation for song choices, leading to user frustration. However, over-explaining could stifle the surprise element, which is crucial for AI interaction. Our solution aimed to balance user delight with strategic interactions to combat confusion.

Approach

Analyzing Spotify & Focus Group Interview

To identify potential opportunities, we started with an audit of Spotify's existing features and surfaces, focusing on relevant user signals. During this exploration, intrigued by the unexplored potential of user signals beyond listening history (e.g., location, prompts), the team chose to focus on the recently launched AI DJ feature for exploration.

Then to gain user insights, we facilitated a quick focus group session with four participants, combining observation with in-depth interviews as they interacted with the AI DJ. Throughout the session, we discovered both confusion over the AI DJ's selections and boredom due to too predictable choices, hindering continuous user engagement and differentiation from existing features like "Radio."

How Might We..... (HMWs)

Having identified the target feature and pain points, the team brainstormed focused HMW questions to guide our design. Following this exploration, we collaboratively evaluated the generated questions and identified the most promising areas for improvement. These key areas centered on achieving a balance between surprising and confusing the user with AI selections, ultimately aiming to deliver an optimal AI experience.

Sketching and Merging Ideas

Fueled by HMW insights, a flurry of concise sketches emerged from each team member, sparking a lively exchange of ideas that refined and elevated concepts. By the end, we had identified the best aspects of each, culminating in a well-defined user journey and the ideal surfaces to bring the feature to life.

The Outcome

#1 DJ on Home

Here, we surfaced DJ on home as a chip. Within the chip, we curated a feed of past DJs users interacted with, along with inspirations for playing with the DJ, such as trending prompts and previewing friends' recent DJs.

#2 DJ Customization

First, the DJ starts as usual to keep the consistency of the legacy experience. The only noticeable difference is the name of the DJ, implying that the name is editable later on.

This is one of the main parts of DJ’s evolution. The DJ can be spiced up by a text prompt provided by a user, adding the taste of artists and friends. This user-driven approach empowers users by allowing a joy of autonomy, while simultaneously training Spotify's AI to deliver unexpected delights, not just expected experiences.

Driven by the pursuit of various signal gathering, we integrated DJ an intelligent question-asking mechanism. This enables to collect a wealth of actionable insights, fostering iterative improvements in personalization even in detail. Tell the DJ and the rest will follow.

#3 Saving DJ

To elevate DJ from a one-time experience to a personalized curator, we proposed saving its selections like a playlist. This empowers users to revisit favorites and, for those truly captivated, embark on an AI-driven musical odyssey. DJ would continuously learn and evolve user preferences, unveiling a world of music and showcasing the transformative power of AI in shaping personal experiences.

Final Prototype