AI transparency

Do you have to tell a customer they are talking to an AI?

Yes, you should tell a customer they are talking to an AI, and a Muzau agent takes care of it. Every email carries, by default, under the signature, “This message was prepared by the AI assistant of” your company, the chat shows “Artificial intelligence” from the moment it opens, and the first SMS starts with “Hello, this is the AI assistant of” your company. By email and SMS, your team chooses what goes out without it and can take over.

Why tell the customer, even when the answer is right

A customer who writes to a company wants to know who they are talking to. If they find out at the third exchange that an AI is replying, they reread everything with suspicion, even if every answer was accurate. Said from the first message, it takes nothing away from a good answer: the customer gets their information, knows where it comes from, and knows a person can step in.

It is also what the European AI regulation asks: a person talking to an AI system must know it (see what the AI Act is). What remains is where to say it, and how, without weighing down every message.

How the agent goes about it

Input

Every message an agent writes to a customer, by email, in the site chat or by SMS

It does

Adds the AI notice that fits each channel, in French or in English

It delivers

A customer who knows who is writing to them, and a team that can take over

By email: one line under the signature, which you can reword

Under the signature of every email an agent sends to a customer, reply, follow-up or meeting recap, the platform adds, small and in grey, “This message was prepared by the AI assistant of” followed by your company name. It also goes out from a connected Gmail or Outlook inbox, in French or in English depending on the language of the message. In the email settings, you reword it or remove it: telling the customer they are talking to an AI is then your responsibility.

In the site chat: shown before the first message

As soon as the visitor opens the conversation, “Artificial intelligence” appears under the agent's name, at the top of the panel, even before the first message (“Intelligence artificielle” on a French page). The notice stays visible throughout the conversation and cannot be changed. The chat replies on its own, without approval: the visitor knows it before reading the first reply. The notice is in the open panel, not on the closed bubble. For everything the chat does, see answering site visitors at any hour.

By SMS: the first sentence of the first message

In each conversation, the agent's first SMS starts with “Hello, this is the AI assistant of” followed by your company name. A customer who texted your 09 number knows right away who is replying. The platform adds this sentence, not the AI, and the reply is written to fit in one SMS, notice included.

Who signs, and what the customer reads when the team takes over

The notice says who prepared the message, the signature says who stands behind it. A meeting recap is written in the name of the person who took the notes and signed with their name when it is known, unless you have set a shared signature, with the AI notice below. When the agent hands over by SMS, it writes to the customer “We will reply to you here by SMS.” Your team then replies from the same number, without the notice, since a person is writing. In the chat, a reply held back by the automatic check lets the visitor read “I would rather have the team check my reply before answering you.” See also how to take over a conversation led by an AI.

How to tell your customers it's an AI without weighing down messages

• Say it at first contact, then always in the same place.
• Name the company: “the AI assistant of” your company says both what is writing and for whom.
• Write “prepared by” rather than “written by”: the word stays true whether the message goes out alone or after someone on the team has read it.
• Give the agent a name, not a false identity: no first name with a colleague's photo, no “I'll call you back tomorrow” that no person promised.
• Say how to reach a person: a number, an address, or a sentence saying the team takes over.
• Put the notice in the message itself, not behind a link or in your terms and conditions.
• One line is enough, with no warning or apology: the notice informs, the message answers.

If you reword the notice in your emails, keep three elements: the word “AI”, your company name, and what the AI does. For example: “This message was written with the AI assistant of [your company] and reviewed by our team.” Only write “reviewed” if every email really is before it goes out: a false notice does more harm than none.

What it does not do

• The email notice can be removed in the settings: if you remove it, telling your customers they are talking to an AI becomes your responsibility.
• By SMS, the notice is in the agent's first message of each conversation, not in every SMS, and SMS written by your team do not carry it.
• The default email notice is only written in French or English: a customer who writes in another language reads it in English.

Frequently asked questions

Does a chatbot have to say it is an AI?

Yes, the European AI regulation asks that a person knows they are talking to an AI. The chat of a Muzau agent shows “Artificial intelligence” under its name as soon as the conversation opens, before the first message, and the notice stays visible to the end.

Should you flag that an email was written by an AI?

With Muzau, the notice is there by default: every email from an agent carries, under the signature, small and in grey, “This message was prepared by the AI assistant of” your company. You can reword it or remove it in the email settings; informing the customer is then up to you.

Should you warn a customer that an AI is replying by SMS?

Yes, from the first message, so they know right away who they are writing to. With Muzau, the agent's first SMS in each conversation starts with “Hello, this is the AI assistant of” your company: the platform adds this sentence, not the AI. The SMS your team writes afterwards do not carry it.

Can an AI sign an email in a salesperson's name?

Yes, with Muzau: the meeting recap prepared by an agent is written in the name of the person who took the notes and signed with their name when it is known, unless you have set a shared signature. The notice below says it was prepared by the company's AI assistant: the signature says who stands behind the message, the notice says how it was written.

What does the customer read when the AI hands over to the team?

By SMS, the agent writes “We will reply to you here by SMS.”, or “we will reply to” their email address if they gave one, and your team can take over from the same number. In the site chat, when a reply is held back by the check, the visitor reads that the reply is being verified by the team and that they will come back very soon.