How to build a phone answering AI assistant
A phone assistant is four pieces: something that answers the line, ears that turn speech into text, a brain that decides what to say, and a mouth that says it. In AIK all four are already wired together; what follows is what you choose and what you have to bring yourself.
The line
AIK answers calls through a telephony provider — you buy a number there, and point its "a call comes in" setting at your AIK address. That is the one part that is not free and not ours: a number and per-minute charges belong to whoever sells you the line. The same number can also take SMS and WhatsApp messages, answered by the same assistant.
Ears and mouth: free, or paid, and they are chosen separately
with AIK, both running locally, nothing per minute. Bulgarian and English.
with.
account with a specialist vendor, if the voice matters more than the cost.
- Free, on your own machine. Whisper listens and Piper speaks — both installed
- The phone network's own. The provider's speech services, simplest to start
- Paid cloud voices. Better-sounding speech and transcription on your own
Ears and mouth are separate settings on purpose: local listening with a paid voice, or the other way round, is a normal choice.
The brain
The same model your other agents use. The assistant has a persona you write in plain words — who it is, what it may promise, when to take a message — and a set of tools: it can look something up, write to a spreadsheet, or book a slot in a calendar while the caller is still on the line.
What it says before it says hello
Every incoming call opens with a sentence saying the caller is talking to an AI assistant, and every outbound call the same — that one is in code and cannot be switched off or written out of the persona, because the EU AI Act asks for it and a persona is editable. Every turn of every call is written to a tamper-evident log, so "what did it tell that customer?" has an answer.
Setting it up
Assistants → pick or duplicate one → write the persona → choose the engines → paste the phone number. There is a browser call button on the page, so you can hear it before any real caller does. The voice settings are one shared installation-wide set; each assistant can override the persona and language.
Trying it
You can try AIK for free at aikautomation.com — the browser call works without a phone number at all, which is the quickest way to hear it. To have it on your own computer or server, with the free speech engines installed and a number connected, you buy a licence and we install it for you. Get in touch.
More from AIK
- How to extract invoice data from PDF to Excel automatically (no code)
- A self-hosted alternative to cloud automation tools for AI workflows
- How to run AI agents on your own server with Ollama and AIK
- EU AI Act Article 50: how to mark AI-generated documents and calls
- Invoice fraud: how a swapped IBAN gets caught before payment
- What is a "process audit" for automation, and what does it tell you
- How much does an AI agent cost to run per month
- RPA vs AI agents: which processes belong to which
- How to give Claude access to your own automation platform (MCP)