Building a 300ms speech pipeline that survives the real internet.
Streaming ASR, predictive buffering, and why we gave up on waiting for end-of-speech detection to finish.
Research, benchmarks, and deployment stories from teams running millions of calls through AI voice agents.
Humans interrupt conversations 3× more often than most voice AI benchmarks account for. Here's how we rebuilt our turn-taking model — and what we learned about why callers hang up on voice AI that “waits its turn.”
Streaming ASR, predictive buffering, and why we gave up on waiting for end-of-speech detection to finish.
A read on the BPO market through 2026. And what operations leaders should ask before renewing their contract.
Custom names, tone dials, and vocabulary packs. Ship a Vox that sounds like your brand in under 20 minutes.
Most AI voice products claim '40+ languages.' We audited what that actually means at intent accuracy below 95%.
Regulatory landscape, customer experience data, and our internal guidance for when the AI should out itself.
High-volume scheduling, stable intents, and frustrated front desks. Why dentistry is Vox's biggest category.
How we moved to per-tenant warm pools and cut cold-start from 1.8s to 90ms.
Bi-directional, field-level mapping, and audit trails by default. Live for all Business and Scale customers.
We A/B tested backchanneling — the 'mm-hmm' and 'right' sounds — and found something we didn't expect.