Instagram's AI pivot: Taste beats automation in product strategy

Jul 9, 2026 · Lenny's Podcast
🎧 PodShort 42 min squeezed to 2 AI SprinklerAS AI / ML
Episode artwork
Adam Mosseri
Head of Instagram at Instagram
Lenny's Podcast
42 min squeezed to 2
Full episode from Lenny's Podcast
Quotable Moments

I think people are going to seek out creativity and authenticity and people.

Adam MosseriHead of Instagram at Instagram
35:59

I think people are fundamentally more afraid of things that they don't understand and about things where people are more secretive and less accessible.

Adam MosseriHead of Instagram at Instagram
40:01

You can't launch something to 3 billion people and not test it first, but you can't test something at our scale and not expect people to cover it and not and be so you have to be ready to talk about it just before you even know you want to launch it.

Adam MosseriHead of Instagram at Instagram
40:37
Key Insights
  • Instagram is shifting to smaller, generalist product teams called 'pods' of 4-6 engineers and a 'product staff' (a generalist PM capable across design, data, research) to improve speed and decision-making.
  • Taste is a critical skill in the age of AI because as it becomes easier to build things, determining 'what to build' becomes more important.
  • Designers, despite some anxiety in the industry, will remain valuable because their inherent 'taste' is a quality that is difficult to automate.
  • The traditional data scientist role is evolving; basic data analysis tasks (like waterfall analysis) are being automated by internal tools, allowing generalist product staff to perform them.
  • Hiring priorities are shifting towards individuals who are curious and willing to 'put themselves out there' (try things, make mistakes) because the future demands adaptability.
  • AI is resetting people's impact and success; some individuals who were previously 'low performers' can now thrive by leveraging AI to do things they were once bad at.
  • Human brains will be most valuable in defining product vision and strategy, as these roles require judgment, deep contextual understanding, and the ability to articulate opinionated, even controversial, paths.
  • Historically, Instagram's algorithm didn't semantically 'know' user interests as deeply as people assumed; progress came from embedding models. Now, LLMs are enabling a description of these complex patterns in understandable terms.
Metrics Mentioned
  • 3 billion people (use Instagram monthly)
  • 1 in every 3 people alive (use Instagram)
  • 40-60% (of engineering work used to be writing code)

RevBots.ai View:

  • AI Sprinkler teams bolt on AI without rethinking workflows: Instagram's pod structure shows early signs of transformation.
  • ARM orgs would flip Mosseri's model: AI handles execution while humans focus on high-judgment strategy.
  • The 'taste' argument mirrors ARM's core thesis: commodity tasks get automated, leaving differentiated human skills.
🎧Full Episode:Lenny's Podcast →