How do you automate business email with AI?

What does the agent do when an email comes in?

It does the first pass and gets things ready for whoever will reply:

• It sorts the message: quote request, complaint, invoice, customer question, or something that can wait.
• It routes it to the right person or team, and flags what is urgent.
• It sums up a forty-message thread in a few lines: what was asked, what was promised, what is still open.
• It drafts replies to simple requests, like sending a document or passing on information that's already written down somewhere.
• For a sensitive email, it pulls together past exchanges and the relevant documents, and suggests a draft.

What about shared inboxes like info@ or sales@?

That's usually where things overflow. Several people read the same inbox, nobody knows who answered, and a message sits unanswered for three days.

Picture a company that gets hundreds of emails a day at one shared address: part of the morning goes to sorting before anyone handles anything. The agent sorts as messages arrive, assigns each one and shows what is still waiting.

How do you know the reply is right?

The agent works from your documents, read from SharePoint, OneDrive or Dropbox, and from your past exchanges.

Every piece of information it puts forward links back to its source. Whoever reviews the draft can see where the deadline, the price or the attachment came from, and fix it if needed.

Who approves before it is sent?

Always someone on your team. The agent drafts, they review and send. Nothing goes out without a person signing off, even a simple reply.

You decide with our engineers which kinds of email the agent drafts and which ones it leaves entirely to your team.

Do we have to switch email providers?

No. The agent gets a custom connection to the email you already use. Access is granted in two clicks, with no password to share with us.

Your team finds the agent in the Saména app, and can also reach it from Claude, ChatGPT or Copilot if they already use them.

Who builds it, and what happens next?

Saména's engineers, Télécom Paris graduates who specialize in artificial intelligence. They start by looking at how email moves through your company: which inboxes, which requests keep coming back, who answers what. They find where an agent saves time and where a person should stay in charge.

Two weeks after the first call, a first agent is running. A dedicated contact then keeps it current: a new type of request, another inbox, another team.

What happens to your email?

What the agent builds from it is encrypted with a key unique to your company, on servers in France, and can be deleted on request.

You choose the model that reads it: Claude, models from OpenAI, or Mistral if you prefer a French model. It runs in Europe and keeps none of your data. Saména can also run on your own servers or on a dedicated instance.

Frequently asked

Can the AI answer emails on its own?

No. The agent drafts the reply, and a person reviews it and sends it. Nothing goes out without a human signing off.

Do we need a specific email provider?

No. Your email gets a custom connection. We look at your setup on the discovery call.

Can we start with a single inbox?

Yes. You can start with one department and a first agent, say on your main contact inbox, then expand whenever you want.

Will our customers' emails be used to train an AI?

No. The model you choose (Claude, models from OpenAI or Mistral) runs in Europe and retains none of your data. Servers are in France.

Does the price depend on the number of users?

No. The monthly subscription covers the whole company, with unlimited users, and is priced on the agents we build. We go over pricing on the discovery call, which is free and comes with no commitment.