Built for the people who run the call.
Voice Copilot is a real-time AI co-pilot for customer-service voice workers. As an agent speaks, the customer hears a clean, professional voice persona of them — while a live coach, sentiment tracker, and per-call scoring rubric run alongside the conversation, and a memory layer carries context across calls.
The product
Frontline voice work is hard, high-volume, and judged on every call. Voice Copilot was built to put a coach in the agent's ear in real time — neutralizing accent friction, surfacing the next-best thing to say, tracking how the customer actually feels, and turning every call into structured, reviewable feedback instead of a black box.
It runs on OpenAI's Realtime API for speech-to-speech, with Structured Outputs powering the coach, function-calling driving sentiment, and a vector-store memory layer for cross-call context. A manager dashboard aggregates the per-call rubric into appraisal-ready scoring.
Voice Copilot is a working, end-to-end prototype rather than a polished commercial product — first built in an intense one-week sprint and hardened steadily since. The compliance and enterprise-readiness items shown across the site are framed honestly as roadmap, not shipped guarantees.
About the team
Voice Copilot is built by a three-person team. Suraj Srivastava (InMobi, Bengaluru; IIT Kharagpur 2012–2016) built the original prototype and leads product and engineering. Akshey Walia (product leader — previously group product manager at InMobi/Glance; ISB) drives the evaluation pipeline and strategy. Jai Goyal leads the live guardrails workstream.
Voice Copilot started from a simple observation: BPO agents are coached after the fact, if at all. Suraj built the first prototype in a week to see how much of that coaching could happen live — and the team has been hardening it since.
Connect on LinkedIn →