Keep pulling the thread on Jesse Zhang.
The primary competitive moat for AI agent companies is the continuous improvement loop, where an agent gets progressively better by learning from a specific client's data, making it difficult for a new competitor to match its performance.
The most critical internal asset a customer can have to ensure a successful AI agent deployment is a comprehensive set of APIs for looking up data and taking actions.
The speaker argues that AI application companies should prioritize gaining market share over achieving high gross margins in the near term, as model costs are expected to decline exponentially.
The speaker predicts that foundation model providers like OpenAI will increasingly build their own applications because the API layer is becoming commoditized with high competition and low switching costs.
The speaker predicts that AI agents will eventually become the primary user interface for all interactions between users and brands.
After implementing Decagon's AI system, Oura Ring saw the rate of customers immediately trying to bypass the automated system to reach a human drop from 1-in-3 to 1-in-20.
The speaker estimates that current voice-to-voice models have a hallucination rate approximately 8 times higher than text-based models, making them unsuitable for many enterprise applications.
AI adoption initiatives within large enterprises are currently being driven top-down from the C-suite and board level.
The ROI for AI coding agents is typically justified through polls of engineers who self-report significant productivity gains, such as 50%.
The speaker believes the current venture capital environment for AI startups is characterized by "too much excitement," making it "way too easy" to raise money.
Decagon has received preemptive offers for new funding rounds almost immediately after closing previous ones.
The speaker advises founders to test the helpfulness of potential VCs by asking for assistance before an investment, as their willingness to help during this period is a strong proxy for their post-investment support.