AivexaNewsSearch
AI news for builders and product teamsChecked every hour

Spotify’s current AI bet is about being the music service that agents choose

Collected Oct 22, 2026

Spotify reported a strong third quarter on its latest earnings call, reaching 788 million monthly active users and generating close to €5 billion, roughly $5.6 billion. But the part of the call that drew the most attention was co-CEO Gustav Söderström's account of how listening may shift as AI agents become a common way to request things, and why he argues that shift favors Spotify rather than threatens it.

His argument rests on a distinction between chores and entertainment. Agents are good at taking over tasks people would rather not handle themselves, and people are glad to hand those off, he said. Entertainment works the other way: nobody, in his framing, asks an agent to summarize a song for them. The envisioned behavior is more concrete than that. Rather than opening the Spotify app, searching for an artist and assembling a playlist by hand, a listener might tell an assistant to build music for a run, to find a podcast on a topic they have been discussing with an AI, or to produce an audiobook recap that resumes exactly where they stopped.

Spotify's stated goal is to be the media service those agents reach for, so that discovery and interaction happen through conversation as well as through its own app. The company points to integrations as evidence it is moving early: it says it was the first music service to connect with Meta's Muse assistant, and it also integrates with ChatGPT and Claude. Söderström described adoption of Spotify inside these agents as good, and cast the company's posture as leaning into agent-driven usage while others lean out or are still deciding what to do. He framed the aim as turning agent growth into a tailwind rather than a headwind.

The wider market is not uniformly open. Amazon has blocked Meta's Muse from shopping on its platform, a case that shows how large players may restrict outside agents when access, security, privacy or commercial control come into question. That is the backdrop against which Spotify is choosing integration over restriction, and it is worth noting the two positions pull in opposite directions: one treats an outside agent as a channel to be controlled, the other treats it as distribution to be cultivated.

Söderström also pitched the shift as a route to deeper personalization. He described Spotify as an intelligent media service wrapped as an agent, usable directly in its own app or through another agent, and as something users can click on, talk to, question and instruct. His illustration: asking Spotify to walk back through a listening history song by song to 2008, pick out true favorites, beat-stretch them faster or slower, sort them by key and BPM, then mix them into a running playlist at an exact cadence. Whether that specific chain of operations is available to users today is not something the call established; it reads as a demonstration of the kind of compound, data-hungry request the company believes it can serve.

That capability claim leans on a data advantage. Spotify holds years of listening history, including which artists people return to, which tracks they skip and what they discover, and that record can inform recommendations that feel more specific than a generic playlist from an assistant with little sense of a person's taste. This is a familiar argument in the agent era: the model that fields the request may be general-purpose, but the service holding the behavioral history is the one that can answer well. In practice, the value shifts toward whoever owns the preference data and the playback relationship, not necessarily whoever owns the chat interface.

On privacy, Söderström said Spotify does not hand listener data to partners. Under the arrangement he described, the outside agent asks Spotify for something, and Spotify's own systems reason about the request and decide what to play. Architecturally, that suggests a service-side pattern: the agent passes along an intent, Spotify resolves it against its catalog and user history internally, and only the result travels back. That keeps raw listening data inside Spotify's walls, though it also means partners depend on whatever interface Spotify exposes and on the quality of its interpretation of a request. A likely trade-off is that this design protects the data asset while giving Spotify effective control over which agent interactions work well and which do not.

The commercial logic is straightforward. If assistants become a primary front door to media, a service that is reachable from every major agent keeps its role in the chain, while one that is walled off risks being routed around. Amazon's block on Muse shows the alternative strategy is live in the market. For Spotify, the open question is leverage: being the default answer inside someone else's assistant is valuable, but it can also reduce the service to a backend unless the conversational experience pulls people back to the app.

Why it matters: developers building assistants, voice interfaces or media features should expect music and audio services to expose intent-level APIs rather than raw catalogs, with the service resolving requests against its own data. Product teams at media companies face a strategic fork, integrate with outside agents or restrict them, and Spotify's choice suggests integration is being treated as growth. The inference to draw is that access terms, not model quality alone, will shape which services agents can actually reach.

Read at TechCrunch · AI

Based on reporting from the original publisher. Visit the source for full context and later updates.

Publisher excerpt

Spotify co-CEO Gustav Söderström explains why the rise of AI agents could be good news for the music service.