How 80,000 Companies Build with AI — Asha Sharma (Microsoft)

⬅️ Back to Podcasts

  • Podcast: Lenny’s Podcast
  • Host: Lenny Rachitsky
  • Guest: Asha Sharma — Corporate Vice President of Microsoft’s AI Platform (ex-COO Instacart, ex-VP Product Meta)
  • Duration: ~1 hour
  • Listen: Apple Podcasts | YouTube

Asha Sharma oversees AI infrastructure, foundation models, and agent tool chains at Microsoft. Her vantage point covers 15,000+ customers building AI agents. Here is what she sees.

Product as Organism

Products are no longer static artifacts you ship and maintain. They are living organisms that improve with every interaction. The KPI is the metabolism of the product team — how fast they can ingest data, process rewards, and produce outcomes. This is the new IP of every company.

The Three Stages of AI Adoption

  1. AI Fluency — everyone in the org becomes fluent using copilots and AI tools
  2. Process optimization — apply AI to existing processes (e.g., fraud resolution from 15 days to 10)
  3. Growth inflection — AI drives customer experience and new product concepts

Companies that fail skip straight to stage three without building the foundation.

Agents Will Outnumber Employees

As the marginal cost of good output approaches zero, demand for output goes exponential. The only way to scale is with agents. The org chart becomes the work chart — task allocation replaces reporting hierarchy. You just do not need as many layers.

GUIs Are Dying

Text streams connect better with LLMs than graphical interfaces. The future of products is composability, not the canvas. Product builders need to rewire their thinking away from UI design and toward how products compose and interact.

Post-Training Is the New Pre-Training

Once a model hits 30 billion parameters, the capex for pre-training no longer makes sense compared to fine-tuning existing models. The competitive advantage shifts to proprietary data and reinforcement learning.

The Full Stack Builder

The silos between PM, engineering, and design are dissolving. Success depends on mastering the loop — cost, rewards, system design, UX — not the lane. Full stack builders eliminate handoffs and move faster.

Season-Based Planning

Roadmapping is impossible when foundational technology changes quarterly. Identify the current season (prototyping, models, reasoning, agents). Set loose quarterly OKRs. Build slack for the unexpected.

Optimism Is Renewable

Sharma’s biggest lesson from Satya Nadella: his ability to generate energy and renew everyone’s dedication to the mission is what makes the culture work. In a field this volatile, optimism is a deliberate choice.

Crepi il lupo! 🐺