Decagon’s Playbook for Building Enterprise AI Applications

Enterprise AI is shifting from thin UI wrappers toward deep, agentic systems that automate complex business processes. Decagon co-founders Jesse Zhang and Ashwin Sreenivas explain their transition to an open-source model stack, which provides superior latency, performance, and control compared to closed-source frontier models. Rather than relying on foundation models alone, the true value lies in the application layer, where agents encode specific business logic and procedures. While forward-deployed engineering currently helps navigate the complexities of early-stage AI adoption, the goal remains to productize these workflows into scalable software. Ultimately, AI agents serve as the "front door" of the enterprise, managing end-to-end customer interactions and operational tasks. This evolution demonstrates that while AI may displace mundane, repetitive jobs, it simultaneously creates opportunities for higher-value, revenue-generating work, effectively expanding the scope of human-led enterprise operations.