Decagon Reveals 90% of Its Enterprise AI Inference Now Runs on Open-Source Models
kimberlywtan · x · 2026-08-01
a16z published a deep interview with the co-founders of Decagon, an enterprise AI customer service agent platform, discussing the deployment of agents in large enterprises like banks and airlines.\n\nDecagon noted that the current decline in enterprise open-source inference share is not a rejection of open-source, but rather a phase of exploring new use cases. Companies prefer frontier closed-source models to ensure performance when initiating new tasks; once workflows mature, they heavily migrate to open-source models for cost and speed efficiency (currently, 90% of Decagon's inference runs on open-source).\n\nAdditionally, the interview highlighted that the smart-vs-cheap model trade-off is a false dichotomy, fine-tuning at the application layer is highly effective, and cheaper support costs drive enterprises to scale up rather than cut spending.
Related event: a16z x Decagon: Open-Source Migration and Moat in Enterprise AI Agents(6 posts)→
More from coding & agent
- 'LLM code is sloppy' rebutted with real ffmpeg code: identical if/else branches in a top library — kingslyj · 2026-10-16
- Agent actions need checks outside the model: NVIDIA demos lean on guardrails amid wave of real agent incidents — Creamy-And-Crowded · 2026-10-16
- Vitalik builds "paranoid mode" agent wrapper with Tor, zkAPI and local prompt rewriting — DavideCrapis · 2026-10-16
- Agent fine-tunes Qwen3.5-0.8B with synthetic data: 4.7%→76.3% accuracy, 11.9× faster — ben_burtenshaw · 2026-10-16
- Beam-501B agent demo: solo beats 5-agent team on ProgramBench rebuild task — ben_burtenshaw · 2026-10-16
- Beam agent builds BeamSpeak, a local Parakeet dictation app for macOS — ben_burtenshaw · 2026-10-16