Every other system loses control after warm transfer. GoFonIA stays in the room — as a silent supervisor: live monitoring, AI coaching via whisper, barge-in when needed, and real-time agent rating. Inbound and outbound. The human sells. The AI guides.
In classic transfer implementations — Vapi, Retell, even native LiveKit transfer — here's what happens: The AI agent leaves the room or hands off the SIP session — and thereby atomically loses all control. The subscription matrix dissolves. No more tracks. No audio access. Control loss is architecture-immanent, not a bug.
It's like a surgeon leaving the OR after the first incision. No serious system would accept that — but in AI telephony, it's unfortunately standard. "Move participant" is a workaround for systems that don't know a persistent supervisor role.
The LiveKit room holds Caller, Agent, and GoFonIA. The subscription matrix determines who hears whom. GoFonIA is muted but fully active: It listens to both sides, analyzes in real time, and can whisper-coach or barge-in at any moment. No human supervisor needed — the AI takes on this role.
The key insight: The LiveKit room is not a transport medium — it is the Call Control Layer itself. As long as GoFonIA stays in the room, it retains:
Full access to all audio tracks (Caller + Agent) via subscriptions — anytime, without re-establishment.
Selective publishing into tracks: Whisper only to Agent, Barge-in to everyone.
Complete session state in Redis without re-initialization — the room lives, the state lives.
Control over the call lifecycle until the final BYE — no control loss, no blind flight.
"Move participant" is a workaround for systems that don't know a persistent supervisor role. No other approach can simultaneously deliver these four properties without keeping the room open.
A customer calls. GoFonIA qualifies, transfers to the specialist agent — and stays as a silent observer in the conversation.
ViciDial routes the inbound call to GoFonIA as a Remote Agent. GoFonIA answers, Leg 1 (Caller) is in the LiveKit room. AI conducts the initial conversation, captures the request, checks customer history.
GoFonIA calls the right agent from the inbound group via outbound SIP. Leg 2 (Agent) joins the room. GoFonIA introduces: "Mr. Müller has a question about the invoice from May 15." — then mutes.
GoFonIA tracks sentiment, script adherence, compliance flags — and acts as AI supervisor: identifies bottlenecks, whispers instructions ("Ask for the customer number", "Offer solution B"). The agent only hears the Caller and GoFonIA's voice. No human supervisor needed.
Conversation ends. GoFonIA creates an agent score: friendliness, resolution rate, call duration, compliance. Score goes to ViciDial disposition. Agent is immediately free again in the inbound group. Caller leg closed.
ViciDial dials leads, GoFonIA pre-qualifies, warm transfers to the closer — with live sales support.
ViciDial outbound campaign hands off to GoFonIA as Remote Agent. GoFonIA calls the lead. AI qualifies: need, budget, timeline, decision-maker. Positive qualification → Step 2.
GoFonIA briefs the closer before joining: "Mr. Schmidt, 5k budget, wants to decide by end of Q2, objection was price." Closer joins the room, Leg 2 open. GoFonIA mutes.
GoFonIA's objection detection reports "price" for the third time. GoFonIA whispers to the closer: "Offer discount tier 2 now." The closer responds immediately. The lead notices nothing. The AI is the supervisor.
Call ends. GoFonIA rates the closer: closing rate trend, objection handling, conversation flow. Score to ViciDial for incentive calculation. Closer returns to closing group.
The decision whether a call is handled entirely by AI or a human takes over is made by GoFonIA in real time based on the situation — not based on a static routing table.
GoFonIA's tool registry can integrate up to 10 APIs per tenant. We distinguish two categories — depending on whether GoFonIA only reads or actively takes action.
Vapi, Retell, and Synthflow think in calls, not rooms. Their architectural paradigm is: accept call → process → transfer → done. GoFonIA's paradigm is fundamentally different:
This is not a feature difference — this is a fundamental paradigm shift. The competition would have to rebuild their entire architecture to replicate this. Systems that think in calls cannot have persistent rooms.
ViciDial's classic Remote Agent bridges Caller and Agent via SIP B2BUA — after that, the connection is blind. No control, no monitoring, no coaching. GoFonIA doesn't just replace the Remote Agent, it augments it: The Call Control Layer stays continuously in the signaling chain. Both legs fully instrumented. And GoFonIA itself is the supervisor — no human needs to listen in. The AI analyzes, whispers, intervenes. LiveKit controls the tracks — and GoFonIA controls the entire call lifecycle.
Not Vapi. Not Retell. Not Synthflow. Not a classic ViciDial Remote Agent. Only GoFonIA.
Try it now →