Legal AI startup Sandstone's approach shows that the model is a commodity. Real defensibility comes from creating a "context layer" that integrates data from CRM, CLM, and communications, giving the AI the business context required to be truly useful for in-house teams.

The competitive advantage for vertical AI isn't just data, but creating increasingly difficult, proprietary evaluation benchmarks. By creating and continuously improving performance against a moving target for specific tasks, vertical AI companies build a durable product advantage that general models cannot easily replicate.

While foundational AI models threaten broad applications like writing aids, startups can thrive by focusing on vertical-specific needs. Building for niche workflows, compliance, and deep integrations creates a moat that large, generalist AI companies are unlikely to cross.

Harvey open-sources its legal benchmark because enterprise clients like law firms can't risk vendor lock-in or conflicts with a single AI lab. For example, a firm representing OpenAI cannot send sensitive data to Anthropic's models. Open sourcing provides a necessary neutral layer.

For entrepreneurs building on top of large language models, the key differentiator is not creating general platforms but achieving deep domain specialization. The call to arms is to know a vertical better than anyone and imbue that unique knowledge into AI agents, creating a defensible moat against more generalized tools.

As AI models become commoditized, the ultimate defensibility comes from exclusive access to a unique dataset. A startup with a slightly inferior model but a comprehensive, proprietary dataset (e.g., all legal records) will beat a superior, general-purpose model for specialized tasks, creating a powerful long-term advantage.

Companies create defensibility by generating unique, non-public data through their operations (e.g., legal case outcomes). This proprietary data improves their own models, creating a feedback loop and a compounding advantage that large, generalist labs like OpenAI cannot replicate.

YC Partner Harsh Taggar suggests a durable competitive moat for startups exists in niche, B2B verticals like auditing or insurance. The top engineering talent at large labs like OpenAI or Anthropic are unlikely to be passionate about building these specific applications, leaving the market open for focused startups.

As base model capabilities converge, the key differentiator is shifting to the "agent harness"—the infrastructure, tools, and skills built around the model. For vertical AI, this is where domain expertise is injected, creating specialized agents with custom tools that outperform generalist models.