How 80,000 companies build with AI: products as organisms, the death of org charts, and why agents will outnumber employees by 2026

Asha Sharma, CVP of AI Platform at Microsoft, discusses how AI products are shifting from static artifacts to living systems that learn from usage, how work and org structures may change as agents scale, and the growing importance of post-training. The conversation also covers planning under fast-moving AI change, common traits of successful AI builders, and leadership lessons.

01Key learnings

• Treat AI products as living systems: build the loop of collecting usage signals, defining rewards, A/B testing, and fine-tuning, since that feedback loop is becoming the core IP.
• Invest in post-training and fine-tuning on your own data, not just off-the-shelf models; once models are large, adapting them to outcomes like price, performance, or quality gives more leverage.
• Watch for a shift from GUIs toward composable, code-native interfaces, since text streams and composability connect better with LLMs and agents than canvases do.
• Keep AI projects tied to real processes with clear measurement, observability, and evals; avoid running many AI experiments for their own sake without a blueprint.
• Build for the slope, not the snapshot: favor a platform or abstraction layer that lets you swap models and tools, because the technology landscape changes constantly.
• Plan in seasons: align the team on secular trends and a north-star metric, set loose quarterly OKRs, and deliberately leave slack in the plan for shifts in the market.
• Platform success often comes from invisible fundamentals like reliability, performance, privacy, and data residency rather than feature count.
• Expect the org chart to flatten and tasks and throughput to matter more as agents take on work; reviewing agent output via observability and evals becomes critical.

02Frameworks mentioned

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