While AI threatens many software companies, those built on strong network effects (like Slack) could become even more vital. AI agents will need to use these platforms as tools to perform tasks, solidifying their position as the central hub of work.

Instead of competing with labs on model training, the defensible strategy is to build the ideal environment or 'habitat' for an LLM in a specific vertical. Replit did this for programming by adapting its editor, cloud infrastructure, and deployment tools to serve the AI, not just the human.

Even if AI makes it easier to build competing software, incumbent SaaS giants retain customers due to immense switching costs. The operational disruption, retraining, and integration challenges of migrating a large organization create a powerful moat against new entrants.

User stickiness for AI models is increasingly driven by the 'harness'—the custom prompts, workflows, and integrations built around a specific model. This ecosystem creates high switching costs, even when a competing model offers incrementally better performance.

While most current AI agents are just replicable instructions, a potential moat exists for tools that build truly autonomous, self-improving agents. The history and learnings of such an agent would create high switching costs, as moving to a new platform would be like training a new employee from scratch.

An enterprise CIO confirms that once a company invests time training a generative AI solution, the cost to switch vendors becomes prohibitive. This means early-stage AI startups can build a powerful moat simply by being the first vendor to get implemented and trained.

The threat of AI models replicating SaaS features is real. Superhuman's defense isn't a superior core technology but a platform strategy. The bet is that users won't build their own tools if the platform offers a powerful network effect of pre-built, integrated agents that work everywhere, creating a defensible ecosystem.

A key competitive advantage for AI labs is using their own advanced coding agents internally to build next-generation models. This creates a self-reinforcing loop where the best models help build even better models faster, a realization that has sparked a "crisis" in other labs now playing catch-up.

In enterprise AI, competitive advantage comes less from the underlying model and more from the surrounding software. Features like versioning, analytics, integrations, and orchestration systems are critical for enterprise adoption and create stickiness that models alone cannot.