Everything on voice and chat AI agents — real-time speech processing, telephony and phone-system automation, conversational AI, voice UX design, and production deployment.
From the blog
Sentiment detection plus a clarifying-question-first protocol cut hostile abandonment by 47%. We unpack PolyAI/Cognigy escalation triggers, real transcript blueprints, and the CallSphere empathy ladder.
PolyAI raised Vodafone NPS from 14 to 64. Web surveys hit 2-5% response, AI voice surveys hit 25-40%. Here is the outbound CSAT/NPS build that actually finishes the conversation.
<Gather> now picks Google V2, Deepgram Nova-2, or Twilio's own model per language. We benchmark the four modes, show hint tuning, and explain when to ditch <Gather> for <Stream>.
Vocode-core wires phone, browser, and Zoom voice into one Python agent class. Build a Twilio inbound bot with GPT-4o and Azure TTS — code + pitfalls.
PushKit is the only API that wakes a terminated iOS app for an incoming AI call. Here is what changed in 2026, why CallKit reporting is enforced, and how to ship it.
Local presence dialing - matching outbound caller ID to the recipient's area code - lifts answer rates up to 4x in 2026 if you avoid carrier filtering. Here is the rotation strategy, DID hygiene, and reputation management that keeps numbers warm.
End-to-end speech-to-speech now clears 700ms in 36 languages. Cascaded ASR-MT-TTS still rules enterprise at 800ms-2s but covers 100+. Here is when to pick each, and how to ship them on WebRTC.
FCC's 2024 TCPA ruling plus 2026 NPRM proposals require clear AI disclosure within the opening of every call. We compare 7 disclosure phrasings, cover the 2-second opt-out rule, and ship CallSphere's compliant template.
Reg F caps debt-collection contact at 7-in-7. AI voice tracks frequency in real time, self-discloses, and resolves 45-50% of accounts without a human. Here is the FDCPA-safe outbound build.
