Currently tracking 42 active AI roles, up 39% versus the prior 4 weeks. Primary focus: Ship · Engineering. Salary range $110k–$402k (avg $222k).
What leadership said about AI on earnings calls (above the line, stacked by event type) vs how many AI roles the company actually posted (below the line). Each side scales to its own peak — read shape, not absolute height.
Trajectory events (left half, cyan) vs active AI roles posted (right half, slate), bucketed by stage. Darker cell = more activity for this company.
Management confirms development of a proprietary Large Personalization Model (Taste Model) trained on user data.
“I talked last time about the large personalization model, which is a model that we're training from based on open source models. But it's trained on our proprietary data. This is not something that we rent from someone. This is something we're building in-house.”— Gustav Söderström
Management highlights strategy of increasing compute spend per employee to drive productivity rather than increasing headcount.
“We have not increased our headcount, actually we slightly decreased our headcount, but we are spending more compute per employee. And that is because we're seeing tremendous return in terms of productivity.”— Gustav Söderström
Management emphasizes integrating AI across all operations to accelerate product development and efficiency.
“We're integrating AI across every part of Spotify, accelerating how we build and deliver at a pace we haven't seen before.”— Gustav Söderström
Expanded Prompted Playlist feature, enabling users to generate playlists using natural language prompts.
Taste Profile feature entered beta, allowing users to view and edit how Spotify models their listening habits.
Song DNA feature reached 52 million users in the first four weeks post-launch.
AI DJ (IDJ) usage reached 94 million subscribers, driving billions of hours of engagement.
Launched a fitness hub featuring ad-free Peloton content, leveraging user data and personalization strategy.
Spotify's AI-driven mixing tools achieved 50 million mixed playlists, with users executing more than one million daily track transitions.
Spotify is building a proprietary 'language-to-audio' dataset to power its agentic media platform.
“What that means structurally for Spotify is that we are building a dataset that never existed, which is the data set of language to music, language to podcast and language to books.”— Gustav Söderström
Co-CEO Alex Norström explains the value chain of AI-driven personalization leading to enterprise value.
“AI leads to better personalization, better personalization leads to more engagement, more engagement leads to more retention, more retention leads to lifetime value and, boom, more lifetime value leads to more enterprise value.”— Alex Norström
Management acknowledged risks from increased AI-generated 'spammy tracks,' requiring continued investment in content quality controls.
“I also want to mention that one thing that AI can do is to accelerate the amount of spammy tracks. I want to be clear that there has always been people trying to abuse Spotify because it's a big economy using spammy tracks. AI is a tool that could help accelerate that.”— Gustav Söderström
Spotify utilizes an internal agentic coding system called 'Honk' that allows engineers to generate and merge code via natural language prompts.
Spotify launched Prompted Playlists, a tool allowing users to describe and set rules for personalized playlists using natural language.
Spotify's Interactive DJ feature has been used by 90 million subscribers, generating over 4 billion listening hours to date.
Spotify plans to provide AI tools to help artists, podcasters, and authors with the content creation process.
“We think AI tools are also very helpful for podcasters and for authors. So we want to help all creators with these kinds of tools.”— Gustav Söderström
Spotify is advancing its personalization strategy, termed 'Personalization 2.0,' to enable conversational interactions where users can talk to the platform to refine recommendations.
“This is why you can talk to the Spotify DJ in English, and it actually understands what you mean and can give you personalized recommendations. And if you play that out, what you should expect at a higher level is just much more user control, what we internally call Personalization 2.0”— Gustav Söderström
Spotify is transitioning its recommendation engine to generative recommender systems to better understand user intent and content.
“First of all, you should expect, in general, recommendations to just get a lot better as the industry, including us, switches to what is called generative recommender systems.”— Gustav Söderström
Spotify integrated with ChatGPT, allowing users to generate personalized playlists based on natural language prompts.
“Once you've connected your Spotify account with ChatGPT, you can ask ChatGPT to create things like the perfect playlist for the countdown to your 2025 rap experience”— Gustav Söderström
Spotify launched an integration with Meta AI on Meta Ray-Bans, allowing users to connect, stream, and control Spotify via the hardware.
“This integration allows users to connect, stream and fully control Spotify.”— Gustav Söderström
Spotify leveraged AI to translate their iOS application to TVOS, significantly reducing development costs for the new Apple TV app.
“But we were able to drastically reduce this cost by leveraging AI, where we could actually translate our general iOS application to TVOS.”— Gustav Söderström
Management highlights the strategic advantage of LLMs over previous deep learning models due to infinite scaling with data and usage.
“So when you hear AI people talk about the scaling loss, what they mean is that LLMs, unlike the previous deep learning systems, they get better so far infinitely as you add more data and more usage.”— Gustav Söderström
Management describes a transition to 'agentic infrastructure' using Model Context Protocols (MCPs) to enable AI to interact with Spotify APIs.
“Concretely, this means that you wrap all the APIs and something called MCPs, model context protocols, so that you can have an agentic infrastructure on top of all your old school tech infrastructure so that you can basically create a product on the fly by just writing something in English.”— Gustav Söderström
First round of optimization on AI playlist resulted in clear performance jumps, including longer listening times and increased saves.
Spotify updated the DJ feature to accept music requests using AI, now available in over 60 markets.
Spotify expanded the AI playlist feature to over 40 new markets for premium listeners.