Jesse Zhang

Sarah Wang × Kimberly Tan × Jesse Zhang × Ashwin Sreenivas

Decagon runs 90% of production workflows on open-source models, where task-specific fine-tuning delivers higher accuracy, lower latency, and lower cost than frontier systems on bounded jobs.Frontier models remain the discovery engine, while Decagon’s continuously rebuilt model factory and enterprise infrastructure turn deployment pain into reusable product; migration will remain constrained by proprietary data, security, and governance.

Patrick O'Shaughnessy × Jesse Zhang

Decagon’s enterprise AI wedge is customer service, where existing resolution rates of 15–20% create a quantifiable path toward 50–80% while escalation to humans de-risks deployment.The moat is continuous improvement: models can review every one of a million monthly conversations, identify the 2% failing, and draft fixes that compound into a brand-level digital concierge.Voice remains the largest unresolved frontier, with voice-to-voice hallucinations probably “like 8x higher” than text, while falling model costs intensify the race for market share and mindshare.

Harry Stebbings × Jesse Zhang

Opendoor is being refounded as a software-and-services platform that uses AI to price homes fairly, then monetizes a longer customer relationship rather than relying on the purchase spread.The broader panel highlights a sharper AI market: applications are expanding TAM rapidly, but weak switching costs and two-week competition make liquidity, strategic acquisitions, and IPO windows important catalysts against paper valuations.

Elad Gil × Jesse Zhang

Decagon’s customer-support wedge offers unusually legible agent economics, with Bilt Rewards stopping team expansion within roughly one month and later reporting around 65 agents of headcount saved.Its differentiation lies in eval-driven orchestration, business-logic customization, observability, and control rather than exclusive model access, while voice and computer-use expansion depend on latency, accuracy, and measurable ROI.

Jesse Zhang × Elad Gil

Customer support is emerging as the near-term “golden use case” for AI agents because deployment can start small while savings, satisfaction, NPS, and accuracy remain measurable.At Bilt Rewards, Decagon stopped scaling support within roughly one month and later quantified around 65 agents’ worth of headcount saved, while orchestration above shared models remains the differentiation.Voice expands the workflow opportunity but introduces a persistent latency-versus-computation trade-off worth monitoring.