Gustav Söderström

Gustav Söderström is co-president and Chief Product and Technology Officer at Spotify, having joined in 2008 and held many roles as the company scaled. The conversation covers his view that the internet has moved from curation to recommendation to generation, how Spotify approaches AI and product design, its organizational model, and the lessons from its recent homepage redesign.

01Key learnings

• Expect the internet's next shift from recommendation to generation to require rethinking products and business models, much as the move from curation to recommendation did.
• Design UIs to match your model's real accuracy: if predictions are right only one in four or five times, show several options at once with an easy escape hatch.
• Treat generative AI as a distinct paradigm rather than more of the same machine learning, and ask what experience couldn't exist without it, like the AI DJ.
• Keep AI features minimal and get out of the way; users came for the music, so avoid letting the technology show off at their expense.
• Place autonomy at the VP level rather than the bottom or top of the org, balancing freedom of thought against incoherence and bottlenecks.
• Discovery features need a low hit rate and fast swiping, and must not replace recall; a homepage tuned for discovery can break users' ability to find known playlists.
• Separate feedback about a redesign that was right but disruptive from feedback about a design that was actually wrong, using cohort comparisons and user research.
• Limit planning to about 10% of the cycle, and push yourself to explain decisions clearly, since being unable to explain something usually means you don't yet understand it.

02Frameworks mentioned

Summary and takeaways written by PM Atlas; quotes are short excerpts. © the original author.