Will your customers accept talking to an AI?

The question is often framed as if the choice were between an AI and an attentive, available, smiling person. For a small, overstretched business, that is not the real alternative. When the tradesperson is on a roof or reception is already on another call, the caller is not choosing between a machine and a human: they are choosing between a machine and nobody.

Put that way, the question changes. A customer who gets an answer to their question about opening hours, leaves their details with someone who understands their request and receives a well-prepared callback is probably better served than one who listened to eight rings before reaching a full voicemail box.

Many bad memories do not come from AI as such. They come from keypad IVR systems that impose a menu of nine options, loops that take you back to the start, messages that repeat “your call is important to us” without ever putting the call through to anyone.

They also come from the feeling of having been manipulated: discovering mid-conversation that you were talking to a machine posing as a person. This sense of deception can damage trust more lastingly than an imperfect answer.

Since 2 August 2026, Article 50 of the EU regulation on artificial intelligence (the AI Act) has required providers to design and develop AI systems intended to interact directly with people in such a way that those people are informed that they are interacting with an AI system, unless this is obvious. According to the European Commission, this information must be given at the start of the first interaction. For a voice agent on the phone, this means a clear announcement within the first few seconds.

Recording follows a similar logic. In a guidance sheet published in 2009, which its website now flags as being updated, the CNIL, France’s data protection authority, calls for a spoken notice at the start of the conversation about the existence of the recording system, its purpose and the possibility of objecting to it. Well worded, this double announcement sets an honest framework: the caller knows where they stand.

An acceptable voice agent is not a wall. At any moment, the caller should be able to say “I want to speak to someone” and get either a transfer to a set number or have their details taken so that a member of the team can call them back. This way out is what makes the automated conversation bearable for those who do not want it.

Some situations make it essential: a caller in distress, an elderly person struggling to make themselves understood, an emotionally charged complaint. A good configuration anticipates these cases and quickly directs them to a human rather than prolonging the exchange.

Every customer base is different, and the best way to find out whether yours accepts a voice agent is to look at what happens. The transcripts show where callers hesitate, which questions go unanswered and at what point they ask for a human.

These observations are used to adjust the information provided, the transfer rules and the tone of the greeting. After a few weeks, you have concrete evidence about your own customers, far more reliable than any generalisation.

Alyas is built around these principles: from the start of the call, the voice agent announces that it is an AI and that the conversation is recorded, answers from your business’s information, transfers to the number you have chosen and provides a transcript of every call so you can check how your customers react. The ten-day free trial is enough to judge it on your own calls.