Conference Presentation, Keynote
The New Way To Build A Startup
- Shift in Internal AI Adoption: Developers at Anthropic, including those building foundational architecture, currently manage 3 to 8 instances of "Claude" simultaneously to implement features, fix bugs, and research solutions, signaling a transition toward AGI-level internal utility.
- The "20X Company" Definition: A "20X company" is defined not by automating isolated functions (e.g., code or support) but by automating all internal workflows, including marketing, sales, hiring, QA, and operations, thereby rendering each employee orders of magnitude more productive.
- Compound Startup Evolution: This concept evolves Parker Conrad's "compound startup" theory, which posits that building multiple parallel products creates a harder-to-displace product-market fit; the 20X model applies this parallelism to internal automation rather than just external product lines.
- GigaML Case Study (Atlas):
- GigaML, a startup with 4–5 engineers, successfully secured DoorDash and 10+ Fortune 500 clients despite competitors being 100x larger in headcount.
- They utilized an internal AI agent named "Atlas" capable of full product manipulation, including browsing, policy editing, and code writing.
- Productivity Impact: Atlas doubled or tripled individual engineer scope by eliminating boilerplate integration work.
- Operational Scale: A single human Full-Time Employee (FTE) now services dozens of accounts, including clients processing 500,000 to 1 million calls daily, by handling the full range of customer requests and feature translation.
- Legion Health Case Study (Unified Source of Truth):
- The AI-native psychiatry network built a custom internal interface acting as a single source of truth for care operations, aggregating patient history, scheduling, insurance codes, and communication logs.
- Growth Metric: The company grew revenue 4X over the past year while maintaining a flat operations headcount.
- Workforce Structure: Operations remain minimal with one clinical lead, one patient support person, and one billing person, despite serving thousands of patients monthly and dozens of providers.
- PhaseShift Case Study (Custom Agent Creation):
- A 12-person team competing against incumbents with hundreds of employees since 2006.
- Implementation Method: The company mandates employees document manual daily tasks, which engineers then immediately convert into custom AI agents.
- Cost Avoidance: This "relentless automation" culture allowed the company to avoid hiring a dedicated design staff entirely by using engineering teams to build front-end designs via "magic patterns."
- Strategic Outcome: Startups adopting these integrated AI approaches are maintaining low payroll, preserving company culture, delaying hiring for entire functions, and achieving record-high growth rates while remaining smaller than their traditional competitors.