Key Takeaways

  • Generalist AI labs like OpenAI and Anthropic are often partners, not competitors, for businesses aiming to automate specific services.
  • These labs prioritize generalizable AGI problems, frequently missing the 'lost mile' of complex, specific enterprise workflows and diverse customer contexts.
  • Enterprises need stable, tailored AI solutions, not the rapid development and deprecation cycles common with cutting-edge lab models.
  • True vertical AI solutions demand mastery across three crucial layers: advanced models, sophisticated 'harnesses and orchestration,' and the final product built on top.
  • Melisa Tokmak, Netic's CEO, calls the belief that pure AGI will solve these real-world enterprise problems 'operationally and intellectually a bit lazy thinking.'

Why the AI Titans Aren't Your Competitors

If you're building an AI product for a specific industry, you might look at the rapid progress of labs like OpenAI or Anthropic and wonder if they're about to swallow your market whole. Melisa Tokmak, founder and CEO of Netic—a company building AI to autonomously run services from HVAC to pet care—chuckles at the thought. As she puts it, “10 years ago that same exact question was can Google do this and then now it became can labs do this?”

Tokmak argues that these generalist labs are not your competitors; they are often your future partners. Their mission is to push the frontier of generalizable AI, not to solve the messy, specific problems of enterprise. She points out that the labs’ rapid development cycles, constantly updating and sometimes deprecating models, are completely unsuited for the stability and consistency real businesses demand. While labs might showcase an impressive new coding agent, the actual enterprise market is flooded with “about like 20 products like what is really happening” – a clear sign of a fragmented and often unrefined approach to business needs.

The "Lost Mile" That Generalist AI Ignores

Tokmak's core insight centers on what she calls the "lost mile." This is the critical, complex gap between a powerful general AI model and a truly functional solution for a niche business. Labs, she explains, are focused on solving the "most generalizable way of the problem." Their researchers might ask, "when we get the AGI, we'll ask how to solve it for essential services." Tokmak doesn't pull punches, calling this approach "both operationally and intellectually a bit lazy thinking."

Consider the challenge of automating a service like HVAC repair. It involves interacting with “millions in the country that have completely different worries, different accents.” A generalist AI model, no matter how intelligent, cannot bridge this gap alone. The 'lost mile' requires more than just advanced algorithms; it demands “harnesses and orchestration, the software and the product that you have to build on top.” Vertical AI companies like Netic succeed because they excel at all three layers: the underlying models, the intricate software scaffolding that makes them useful, and the final user-facing product. Without this layered approach, generalist AI remains an impressive engine without a steering wheel or a comfortable seat.

What to Do With This

Stop chasing the latest generalist AI model release as your primary product strategy. Instead, identify the specific 'lost mile' challenges unique to your vertical. Tomorrow, map out three distinct, non-obvious scenarios your customers encounter where a generalist AI would trip up due to diverse contexts or complex workflows. Then, sketch out the custom 'harnesses and orchestration'—the software and product layers—you'd need to build on top of a foundational model to truly solve those problems for your users.