AIVA — AI Voice Assistant

An AI operator that answers the phone and resolves the call.

Built into Metron and wired into your telephone exchange: the assistant speaks with customers, handles the everyday requests on its own and only escalates what truly needs a person — running on a local AI so answers come back in real time.

Outage checks, non-payer handling, ticketing and payment verification — all on your own hardware, nothing leaves the building.

~80 % of calls resolved local AI · real-time RTX 5080 for the model RTX 5050 for STT + TTS on-premises · GDPR

What it does

Voice AI assistant

AI · Voice

An AI operator that picks up the phone — and actually resolves the call

Wired into the operator's exchange (PBX): it talks to customers in natural speech and resolves about 80 % of calls end to end — no human needed. It runs on a local AI, so the caller gets an answer in real time.

Answers incoming calls, understands the request and replies by voice — STT (speech-to-text) on the way in, TTS (text-to-speech) on the way out
Resolves ~80 % of calls fully; the rest is handed to staff with the full context attached
Line / circuit outages: verifies availability, confirms the fault and e-mails a notification
Non-payers: 8 in 10 simply never received the invoice — it re-sends it by e-mail and checks the payment
Handles the everyday “have you tried turning it off and on”, and confirms the service is reachable
Opens and updates tickets and logs the whole call on the customer card

AI operatorPBXSTTTTS80% resolved

On-prem AI

Local AI — because on a live call, speed is everything

The model runs on our own GPUs in-house, not in a distant cloud, so there is no awkward pause before the assistant answers and no customer data ever leaves the building.

1× NVIDIA RTX 5080 for the AI itself — the language model that understands the request and generates the reply
1× NVIDIA RTX 5050 for STT and TTS — speech recognition (speech-to-text) and speech synthesis (text-to-speech)
Latency low enough for a natural conversation; fully on-premises and GDPR-friendly

local AIRTX 5080RTX 5050low latency

CRM

It doesn’t just talk — it acts in the system

The assistant reads and writes the same data as your staff, so a phone call turns into a real action in the CRM instead of a note someone has to process later.

Reads the customer card, services and payment status straight from the CRM
Creates tickets and records the call and its outcome in the customer history
Sends e-mails (invoice, outage notice) and flags anything that needs a person
Checks payments and restores the service once the customer has paid

CRMticketscustomer cardpayments

How it works

Call flow

Customer calls voice in

RTX 5050 STT · TTS speech ↔ text

RTX 5080 AI model understands & decides

Metron CRM actions · tickets · payments

↩ The reply travels back along the same path: CRM → AI → TTS → caller.

Want it on your exchange?

Let the AI answer the phone — you keep the exceptions.

The voice assistant is part of Metron: it already knows your customers, services and payments, so a call turns straight into a resolved request or a ticket.

Talk to us → See Metron →