Most people have had a bad experience with an automated phone system. The robotic voice, the long pauses, the inability to handle anything outside a narrow script. These experiences have set a low bar for what "AI on the phone" means.
Turnless is built to a different standard. Here's what actually goes into a phone call that sounds and feels natural.
Latency: the thing nobody talks about
The single biggest factor in whether an AI phone call feels natural is latency — the time between when the caller finishes speaking and when the AI responds.
Human conversation has a response latency of around 200–300 milliseconds. Go above 500ms and the conversation starts to feel stilted. Go above a second and the caller thinks the call has dropped.
Most AI voice systems have latency in the 1–3 second range. This is why they feel robotic — not because the voice sounds bad, but because the timing is wrong.
Turnless uses Google's Gemini Live API, which is a native audio model — meaning it processes speech directly rather than converting it to text first. This eliminates a full processing step and brings latency down to the range where conversation feels natural.
Interruption handling
Real conversations involve interruptions. A caller might start answering a question before the AI has finished asking it. They might change their mind mid-sentence. They might say "actually, wait" and redirect.
A system that can't handle interruptions forces the caller to wait for the AI to finish speaking before they can respond. This is unnatural and frustrating.
Turnless handles interruptions the way a human receptionist would: it stops speaking when the caller starts, processes what they said, and responds to the new direction. The conversation flows rather than proceeding in rigid turns.
Tone and context
The AI's tone adapts to the business it's representing. A dental practice gets a calm, professional receptionist. A plumbing business gets a practical, no-nonsense dispatcher. A hair salon gets a warm, friendly booking assistant.
This isn't just about the words used — it's about pacing, formality, and the kinds of questions asked. A plumber's AI asks about the address and whether the water needs to be turned off. A dentist's AI asks whether it's a new patient and whether there's a preferred practitioner.
The context comes from the business's steering file — a set of instructions that tells the AI how to behave, what to say, and what to do in specific situations. This is set up during onboarding and can be updated at any time from the dashboard.
What happens when the AI doesn't know the answer
A good AI receptionist knows its limits. When a caller asks something outside the AI's knowledge — a specific price that hasn't been configured, a question about a particular staff member's availability — the AI doesn't guess. It takes a message and lets the caller know someone will follow up.
This is better than a human receptionist who makes something up or gives incorrect information. The AI is consistent, accurate within its knowledge, and honest about what it doesn't know.
The handoff
Not every call should be handled entirely by AI. Some callers want to speak to a human. Some situations require a judgement call that the AI isn't equipped to make.
Turnless handles this with a clean handoff: when a caller asks to speak to someone, or when the AI determines the situation warrants it, it takes the caller's details and notifies the business owner by SMS. The owner can call back with full context about what the caller needed.
This is the right model for most small businesses: AI handles the routine calls, humans handle the exceptions.
Turnless is built on Google's Gemini Live API for natural, low-latency voice conversations. Book a demo to hear it in action.