Spotify’s Real AI Strategy Isn’t Music - It’s Survival

In an agentic world, consumers may stop choosing content and start choosing outcomes…

Last week I wrote a piece entitled ‘After the Feed: A New Era for Music Gatekeeping has Arrived’. It explored the arrival of AI agents and the implications for content and music in particular. I mentioned an agentic personalised audio app called Huxe that I have used regularly since October. Then, on Thursday last week Spotify held their third investor day presentation and within hours the founders of Huxe announced that they were calling it a day. Perhaps it was a coincidence, but amongst Spotify’s announcements was this:

“…Studio by Spotify Labs, a standalone desktop app that takes Personal Podcasts further and will let users generate private, personalized audio experiences like daily briefings that are saved directly to Your Library in Spotify. It’ll be available soon as a Research Preview for Premium users in more than 20 markets. Studio understands your Spotify taste across music, podcasts, and audiobooks, and can draw on world knowledge to help you find the audio you want faster. You can also choose to let it take action on your behalf: researching topics, using a web browser, organizing information, and helping complete tasks across the tools you use every day.”

Studio by Spotify Labs is pretty much Huxe on steroids. I enjoyed starting my day with Huxe, but if Spotify gets their agent right it’s likely to be significantly more engaging.  

Of course Spotify also announced their entry into the generative AI music market by telling the world that they had completed…

“landmark licensing agreements with Universal Music Group and Universal Music Publishing Group, enabling Spotify to launch a new tool where fans can create covers and remixes from participating artists’ and songwriters’ catalogs, with consent, credit, and compensation built in from the start. The new tool will launch as a paid add-on for Spotify Premium users and create an additional source of income for artists and songwriters, on top of what they already earn on Spotify.” 

I expect SONY Music and Warner Music will follow suit alongside many in the indie sector. Acceptance won’t be instant, it never is with disruptive new technologies, but gradually the twin opportunities of increased revenue and deeper engagement with fans will be too compelling for most artists to deny.

These announcements crystallised something that has become increasingly obvious: standalone AI audio platforms and agents are now competing against platforms with vastly deeper data and distribution advantages. But - as big as Spotify are - they too face potential disruption as AI proliferates and they are positioning themselves for survival in the agentic era. The risk for Spotify is not that AI makes playlists better, or that it enables deep new engagement plays. It’s that AI changes who owns the user relationship. In an agentic world, consumers may stop choosing content and start choosing outcomes. Instead of opening Spotify, they may ask an AI to energise their morning commute, then educate them at lunch and help them unwind at night. Whoever owns that interface captures disproportionate value. Yet it’s also true that media consumption has never been purely functional, many people, and I suspect a lot of music fans especially, still enjoy opening destination apps or websites and going down the rabbit hole rather than delegating decisions. Spotify must hope this will continue to be true, as they are in a race to avoid becoming the service layer of audio consumption with meta-agents becoming the interface. By meta-agent, I mean a personal AI layer that sits above individual services and orchestrates decisions across your digital life. One thing I've found to be consistently true throughout my career is that mainstream consumers are by far the majority, and their expectations and behaviours around music and content are not the same as those of music fans.  

Spotify does have some competitive moats, for instance, they arguably have one of the best data sets on the planet when it comes to understanding audio consumption. Their proprietary “Large Taste Model,” trained on trillions of behavioural signals and years of user interaction across music, podcasts, and audiobooks is unmatched and may prove more defensible than catalogue scale alone. When I was running international editorial and content at Spotify my ambition was for the product to soundtrack your life from dawn till dusk, and they are putting themselves in pole position to achieve this goal. In an AI era where supply becomes abundant, understanding preference becomes the truly scarce asset. Having said that, taste models are powerful but may become less differentiated if AI agents gain permission to observe behaviour across multiple services.

For artists, this raises an uncomfortable possibility. Discovery may shift from persuading editorial teams and algorithms to persuading agents. And it’s likely that agents will replicate at least some aspects of the human approach to curation that I described in my previous articles: ‘The Iceberg Principle of Music Curation’ and ‘After the Feed: A New Era for Music Gatekeeping has Arrived’. Metadata, context, audience signals and fan relationships will become at least as important as the song itself.

Ultimately, Spotify is preparing for survival in an era where the most successful meta-agents are likely to be owned by the tech giants with their dominant operating systems. Apple, Google, OpenAI, Amazon and possibly Anthropic are likely to be at the top of the pile.

Spotify needs to be, at a minimum, the go-to second tier ‘specialist’ agent when a consumer wants audio, and if they can manoeuvre smartly, they may be able to sustain a decent percentage of direct interaction.

The larger story, as they build out their potential moats, is also their investment in the one thing agents can't easily replicate: the relationship between artists and their most engaged fans. Spotify’s ambition isn’t to beat Apple Intelligence or Google’s - it’s to remain indispensable. And if meta-agents reduce direct app interaction, platforms need new forms of lock-in. Hence Spotify’s interest in superfans and new AI enabled experiences. Access, identity and participation become harder to scrape than streams. Their first-look ticket deal with Live Nation is a very strong play in this regard. The logic is clear: an agent can commoditise the stream, but it cannot commoditise the feeling of being first in line for tickets for your favourite artist.

No matter what Spotify does, consumers may reject fragmented specialist agents altogether and default to a single personal AI orchestrator that will treat the digital streaming platform as a utility. Will consumers choose an AI that tracks and facilitates all of their life, or one that knows their music and audio needs best? Of course, the future is unwritten, and new operating systems, new devices, new businesses and reactionary human behaviour may all play a part in the race, but personally I wouldn’t bet against the mainstream consumer’s need for value and simple utility.

The real question may not be whether Spotify survives the agentic era. It may be whether consumers continue using media services directly at all. There’s a chance they will simply choose a meta-agent connected to Spotify as a second-tier platform and then let it orchestrate everything else.