In the Future of Voice AI series of interviews, I ask three questions to my guests:
- What problems do you currently see in Enterprise Voice AI?
- How does your company solve these problems?
- What solutions do you envision in the next 5 years?This episode’s guest is Jordan Dearsley, CEO and Founder at Vapi.
Jordan Dearsley is the cofounder of Vapi (YC W21), a developer platform for building Voice AIs that talk like people. Previously founded Superpowered, an A.I. notetaker for meetings.
Vapi.ai is a leading platform for deploying AI voice agents. With a developer-first approach, Vapi lets engineering teams deploy voice agents in their products and services.
Recap Video
Takeaways
Vapi officially launched on Product Hunt in March 2024, achieving in one month what took their last company three years.
Vapi handles everything between AI models and deploying voice agents to production, including model orchestration, infrastructure, conversation flows, and integrations.
Vapi strategically shifted from serving developers to enterprises because enterprises move slower and offer more stability than developer-driven platforms.
AI developer platforms are a race to the bottom, competing with open-source and price-driven rivals.
Enterprises need guardrails; developers need configurability.
AI models that hallucinate or skip key steps can completely derail enterprise adoption, making determinism critical.
Many workflows in enterprises lack APIs, making automation difficult; future AI developments need to bridge this gap.
IVRs tank customer satisfaction, dropping CSAT scores just by being part of the call.
Enterprises want control and observability over call flows because their revenue depends on phone interactions.
Legacy IVR systems limit flexibility, making CX adjustments difficult.
Generative AI solutions need to balance being as flexible as humans and determinism.
IT teams have more buying power—now their signoff is needed in addition to end champions.
The industry still lacks real-time tools to experiment and track AI performance.
Latency is solved; the next challenge is making AI agents reliable and predictable in real-world environments.
Share this post