Building products with artificial intelligence calls for fundamentally reimagining conventional development strategies. Where traditional methodologies progress linearly through planning, prototyping and launch, AI demands a more iterative, adaptive approach.

Consider Spotify’s Discover Weekly playlist feature, which leverages AI to provide users personalised song recommendations. The key wasn't just planning recommended songs, but exploring different algorithms and prototypes to strike the right balance between familiarity and discovery in selections.

What makes creating AI products like Discover Weekly so different? A few critical contrasts:

• Data Ascends: Robust, unbiased datasets are the lifeblood for training performant AI models. For Spotify, this meant aggregating signals from millions of listeners.
  
  

• Prototypes Explore: AI prototypes are hypothesis tests for model viability, not UI demos. Spotify diligently experimented to tune recommendations.
  
  

• Launches are Beginnings: AI products treat launch as the starting line, with ongoing retraining essential to optimisation. Spotify continuously monitors and improves song suggestions.

While these differences complicate development, they unlock immense opportunities. AI holds the potential to generate personalised, responsive and intelligent experiences. But harnessing this potential requires recalibrating processes, capabilities and expectations.

By embracing exploratory, science-infused development we can realise AI’s possibilities. With creativity channelled through rigorous engineering and analytics, products transcend static software to become truly adaptive.

AI will only grow more central to products. To harness its powers, we must evolve our mindsets and methodologies. Though the path remains winding, it’s one well worth pursuing.