Challenge
Intercom wanted to extend its customer support AI agent from chat and email into phone calls, where latency and interruption handling are much less forgiving than text support.
OpenAI's 2025 enterprise report and Intercom's 2025 architecture story show how Fin Voice moved beyond chat into phone support with lower latency and higher automation coverage.
lower answer-generation latency
calls resolved end-to-end on average
faster resolution for calls that still reach humans
Each case study page keeps the narrative tight: the constraint, the intervention, and the measurable effect.
Challenge
Intercom wanted to extend its customer support AI agent from chat and email into phone calls, where latency and interruption handling are much less forgiving than text support.
Solution
Intercom built Fin Voice on OpenAI's Realtime API and combined structured evaluations, A/B testing, and modular AI architecture to support complex phone workflows such as refunds, troubleshooting, and multi-step request handling.
Public source
This case is based on a public disclosure rather than a confidential client delivery.
We can translate the same structure into your inquiry flow, knowledge system, or automation roadmap.