Summary
Varun Shenoy, co-founder of Long Lake, presents an unconventional approach to AI deployment in services businesses: rather than selling software, Long Lake acquires the businesses outright, becoming operator-owners with direct accountability when AI fails. Over two and a half years, Long Lake has raised over $3 billion from investors including Elad Gil, General Catalyst, and AlphaWave, acquired 35 businesses across property management, HOA, architecture, and HR services, and recently announced a $6.3 billion take-private of American Express Global Business Travel — the world’s largest corporate travel platform.
The talk draws on this deployment experience to outline three concrete lessons. First, moving agents from co-pilots to co-workers requires progressing through a spectrum of autonomy — from RAG chatbots to synchronous agents to fully asynchronous workers — and the “async agent” model for services knowledge work (analogous to spinning up sandboxed coding jobs) remains an open frontier, since serial industries haven’t yet developed the parallelization instincts that software engineers have. Second, owning real businesses generates proprietary workflow data unavailable to frontier model training, enabling grounding and fine-tuning advantages. Third, the compounding value comes from learning loops: mistakes feed back into improved training data, which improves future agent performance.
Shenoy frames AI diffusion as a 20-year problem, comparing it to electricity adoption (first demonstrated 1882, industrialized only decades later), and argues that operator-owners — not vendors — are best positioned to drive and benefit from that diffusion.
📺 Source: AI Engineer · Published August 28, 2026
🏷️ Format: Workflow Case Study
