Decagon’s Playbook for Building Enterprise AI Applications

a16z

Summary

Decagon founders discuss their enterprise AI strategy, emphasizing that for complex business workflows, specialized open-source models often outperform generic frontier models. They argue that successful AI application companies must move beyond mere thin wrappers by building robust infrastructure, focusing on 'glass box' product design, and partnering closely with enterprises to integrate AI into existing operations. Ultimately, they view AI agents as the new front door of business, capable of automating complex tasks while expanding the capacity for humans to perform higher-value work.

Chapters

  1. 0:22The persistence of software
  2. 0:31The shift toward open source
  3. 0:35Model Strategy and Performance
  4. 0:42Building Enterprise AI Applications
  5. 4:37Balancing cost, intelligence, and latency
  6. 6:50When to use frontier models
  7. 7:35The importance of evaluators
  8. 10:27Continuous learning and iteration
  9. 18:32Capturing business logic in the application layer
  10. 23:08Avoiding the consulting trap
  11. 26:02Scaling Forward Deployed Engineering
  12. 27:59Converting consulting to product
  13. 31:41Business process vs. customer support
  14. 39:27Glass box versus black box approach
  15. 42:30Enterprise Sales Dynamics
  16. 45:07Founders' role in sales
  17. 45:46Navigating large orgs
  18. 1:11:52The role of X in modern business
  19. 1:17:34The Jevons Paradox in support
  20. 1:17:48The Future of AI and Jobs

Topics covered

  • Open Source vs. Frontier Models
  • AI Agent Architecture

+4 more topics (full content when you follow the channel)

Transcript excerpt

An AI agent should just be the front door of your business. And every interaction, whether it's like reactive or proactive with a customer, should be handled by AI. >> This narrative dominated the first half of 2026, which is that anthropic open AAI. They're the last startups. They're going to take over everything. >> Even once you have AGI, agents are going to need somewhere to store work and pull information from and reason about things. I don't think software as a whole in any meaningful way is going away. Unfortunately, the Frontier Labs, they do have small models, but you can't really control them in the way that you want. So, today 90% of our workflow is on open source >> on the specific task we want them to do. They actually outperform the large smart state-of-the-art model. >> The thing that we built was not an agent that does customer support well, but rather an agent that follo…