Daniel Dines on Why Work Processes Not Models Will Be The Most Valuable Asset in AI

By 20VC

Categories: VC, Startup

Summary

Daniel Dines argues AI models are commoditizing while enterprise work processes become the defensible moat—AI can't learn on the job like humans do because it can't alter its weights through experience, making documented workflows, not model selection, the true competitive advantage for builders.

Key Takeaways

  1. Models are interchangeable commodities; the real value lies in mapping and optimizing enterprise workflows—the processes around work, not the AI doing the work.
  2. AI lacks human-style learning-on-the-job capability—it cannot alter its weights through experience like humans transform through years of domain work, limiting its adaptability to novel enterprise scenarios.
  3. Enterprise intuition and micro-initiatives (hunches, pattern recognition) developed through years of transformation cannot be documented on paper—they're tacit knowledge AI systems cannot replicate without explicit training.
  4. Everything AI queries requires full enterprise documentation and context—unlike humans who absorb organizational knowledge gradually, AI must read entire enterprise context repeatedly, creating a scalability and efficiency problem.
  5. The AI arms race isn't about which model you use—it's about having the most efficient, documented, optimized workflow that any model can execute, making process engineering the defensible differentiator.

Related topics

Transcript Excerpt

In my opinion, even the labs in China that are building AI right now, I would classify them as the good guys. >> Joining me in the hot seat state is one of my dearest friends and one of the greatest founders of the last decade, Daniel Dyn, founder and CEO of UiPath. And today we discuss the biggest questions. What on earth does pacing the frontier mean? Is it even possible? Will we replace humans with AI? >> I've never hidden from my employees that there will be a transformation. Jensen is bound by the success of open source. >> This is a truthtelling, mythbusting conversation, and it's just a fantastic discussion between two old friends on what no one is talking about, but everyone needs to know in the world of AI today. >> Models are interchangeable, but the workflow, the map of work and…

More from 20VC