The AI receptionist market changes every week. We would rather ship the next useful piece than pretend the product is finished. This page is the public list of work in beta or headed to production — CRM hooks, dispatch texts, finding the right on-call person, outbound campaigns, and intelligence that sits beside a live call.
Production service today remains $24 per month plus $0.35 per minute of AI talk time, billing stops at transfer. Items below are not billed as separate SKUs until they leave beta.
An n8n node so chat or SMS from a workflow hits the same agent that answers the phone — dispatch texts, web chat, or a CRM-triggered message.
If an outbound agent reaches a menu instead of a person, it navigates with voice or DTMF instead of talking over “press 1.”
Library, preview, visual editor with placeholder chips, test-send, history. Bind a template to send-email with From Name and Reply-To.
Restaurant template looks up availability and books through OpenTable. Hours, party size, SMS confirmations, and a small outcomes dashboard.
Retry by outcome with caps and backoff. Consent basis and local-time calling windows. AI disclosure recorded. 24h/7d/30d abandon rate. Org do-not-call list checked before every dial. Optional fixed greeting before the model starts.
One catalogue of nine models (GPT-4.1 default). Gemma 4 31B, Deepgram Flux, Cartesia v3/v3.6, Cartesia STT. Dead Google STT option removed.
Attended warm transfer (SIP REFER + Replaces) so the PBX keeps the original caller ID — verified on NetSapiens. Optional line spoken before a blind transfer can be edited or turned off.
Listen beside a live call: sentiment alerts, intent tags, keyword/compliance pings, post-call action summary, supervisor dashboard. Coaching layer — not a second receptionist.
Requests come from customer calls in Greensboro. If an item is blocking a rollout, say which one. Live price is still on pricing and cost.