Case fileAI Agents & Voice AI
Two voice systems: an AI receptionist that answers calls and books appointments into Google Calendar, and outbound AI calling agents for sales. Both answer from a knowledge base and hand urgent calls they cannot answer to a person.
Two phone workloads depended on people being available. Inbound callers needed answers and appointments, and outbound sales depended on someone working through prospects one call at a time.
Both needed an AI on the phone that answered from what the business actually knows, acted during the call, and recognised when a person had to take over.
What made it non-trivial
Qzentra built an AI receptionist that answers inbound calls, responds from the business’s knowledge base, and books appointments directly into Google Calendar while the caller is on the line.
Alongside it, Qzentra built outbound AI calling agents that place sales calls to prospects, on the same foundation: telephony, real-time speech recognition, an LLM for conversation logic, and speech synthesis.
Both follow one escalation rule. When there is no proper response saved in the knowledge base and the call seems urgent, the call is handed to a person.
What Qzentra was responsible for
The loop a single call travels. Recognition, reasoning, and synthesis have to complete inside a conversational pause; bookings land in the business's own Google Calendar.
The receptionist answers incoming calls; the sales agents place calls to prospects. Both run on a telephony layer.
The caller’s speech is transcribed in real time.
Answers come from the knowledge base saved for the business, not from the model’s general memory.
An LLM decides what to say next: answer the question, book the appointment, or move the sales conversation forward.
The receptionist books the appointment into Google Calendar while the caller is still on the line.
The response is spoken back to the caller.
When the knowledge base has no proper response and the call seems urgent, the call is handed to a person.
The agents respond from what is saved in the knowledge base rather than improvising, so a gap in that knowledge is detectable instead of hidden behind a guess.
The receptionist writes the appointment to Google Calendar while the caller is still on the line, so the booking is settled in the conversation rather than chased afterwards.
A call goes to a person when the knowledge base has no proper response and the matter seems urgent. Routine calls stay with the agent; urgent unknowns do not wait.
The telephony layer carries the call; the conversation logic, knowledge base, and calendar integration sit behind it and can change without changing the number the caller dials.
Named last, with the role each played. The capabilities are what the project demonstrates beyond its primary service.
Capabilities demonstrated
Discuss a project
Tell us the calls you take and make, the calendar or systems the agent must act in, and when a person should take over. That is the specification to start from.