What Remains the Same

Despite the AI revolution, many core principles of defensibility remain unchanged from the best practices of building great software companies:

1. User Experience Still Reigns Supreme

The product surface area where you can build and differentiate remains critical. Just as with traditional software, AI applications that deliver a 10x better user experience create lasting defensibility. This requires a nuanced understanding of user needs that goes beyond simply wrapping an AI model in a basic interface.

Take Granola versus other AI note-taking tools. In a crowded market where everyone has access to similar underlying AI capabilities, Granola has emerged as the preferred choice for many investors I know who use it daily for their calls (myself included). The thoughtful UX details and seamless workflow integration create true delight and utility, which reinforces retention.

The little things matter enormously. While competitors can implement similar AI capabilities, few can nail the entire product experience that makes users stick around. These details aren’t nice-to-haves—they’re the foundation of defensibility.

2. Go-to-Market Strategy Creates Compounding Advantages

A differentiated GTM approach remains as powerful in the AI era as it was before. The right GTM strategy can create two types of advantages:

  • Data advantages: By targeting specific segments or use cases, you can gather specialized data that improves your product outputs even when using the same foundation models as competitors. For example, a legal AI product with access to more case data will likely leapfrog other competitors in terms of result accuracy over time.
  • Personal context advantages: Building stickiness by rapidly accumulating personalized context—whether in B2B (understanding an organization’s business context and unique workflows) or consumer applications (learning individual tastes and preferences)—creates a defensibility moat. When your AI knows more about the user, it produces better results, making switching costs increasingly prohibitive.

What Might Be Different

While the fundamentals haven’t changed dramatically yet, there are emerging differences that could reshape defensibility in the future:

The Potential Disappearance of “User Experience”

If AI models truly approach human capabilities (AGI), the concept of “user experience” as we know it might eventually transform. When we can interact with AI as naturally as we do with human colleagues or friends—through conversation and natural language—traditional UI/UX considerations could become less relevant.

However, I believe we’re still at least a few years away from this reality (although you can never underestimate the speed of technical progress). The most advanced general models today are comparable to “know-a-lot-but-master-of-none” junior colleagues a few years out of college. They’re broadly capable but lack the depth of specialized expertise.

New Defensibility Vectors

In this transitional world, defensibility will increasingly come from:

  1. Domain-specific training efficiency: How effectively can you train AI in your particular domain or style? Companies that develop methods to efficiently specialize general models for specific contexts will maintain advantages.
  2. Human-AI feedback loops: How can you reduce friction in the feedback cycle between humans and AI? Since human expertise will still be required to finish “the last 20% of work” for the foreseeable future, companies that build seamless collaborative workflows and interactions will win.

The Bottom Line

The question “If AI models universally improve, will my product get better or become less differentiated?” remains essential for evaluating defensibility.

While we navigate this transition, the companies that will thrive are those that combine traditional defensibility principles (superior user experience and strategic GTM) with emerging AI-specific advantages (efficient domain specialization and optimized human-AI collaboration).

The winners won’t be those with marginally better prompting techniques, but those who understand that even in the AI era, technology alone is rarely the sustainable differentiator. Instead, it’s how you wrap that technology in experiences, workflows, and go-to-market strategies that perfectly address specific user needs—a challenge that remains as difficult and rewarding as ever.