Help Centre
Home AI AI Frontdesk

AI Frontdesk: Letting the assistant learn business knowledge from emails

Valid for: AI Email Assistant (Momentum AI)

Your assistant learns not only your writing style from your emails but also details about your business – such as opening hours, prices, or how long a voucher is valid. It presents these details to you as knowledge suggestions and only uses a suggestion in its replies once you have approved it. This way, it answers recurring questions with the information you have already given your customers yourself.

This feature is available with the AI Email Assistant. The following steps show you how to review and approve suggestions and how to manage the learned knowledge.

Requirement

Where your assistant gets its knowledge from

When replying to emails, your assistant draws on two sources of knowledge:

  • Knowledge learned from your emails: the suggestions you approve as described in this article. This knowledge only applies to email replies – your assistant handles calls and chats without it.
  • Your company knowledge: the details you store for all products in your contract – learned from your website, added by you, or uploaded as a document. The article AI Frontdesk: Storing and managing company knowledge explains how to maintain it.

The learned knowledge does not appear in the Company Knowledge section, only under Approved knowledge in the Learned business knowledge section.

Note

Once you have connected your mailbox, the assistant automatically prepares knowledge suggestions from the emails of the last 30 days.

Open the Email assistant section

  • Log in to your IONOS account.
  • Click on Menu > AI Frontdesk in the title bar.
  • In the contract overview, click on Manage in the tile for the contract you want.
  • In the navigation bar on the left, under Settings, click Email assistant. The Email assistant section opens on the Configure assistant page. You can make all of the following settings there.

Review and approve knowledge suggestions

If there are new suggestions, you can find them in the Learned business knowledge section under Review suggested knowledge.

Please Note

Check every piece of information before you approve it. Once approved, your assistant passes the information on to your customers as fact – whether it is an out-of-date price or opening hours, or up-to-date information.

  • Under Review suggested knowledge, check the individual entries. Below each entry, you can see how many email threads it is based on and the period they cover.
  • Select the checkbox next to each entry you want to keep and clear the checkbox next to all others.
  • Click Approve selected items.

The approved information then appears under Approved knowledge and is used in your assistant’s replies right away.

If you do not want to accept any of the suggestions, click Discard suggestions.

Tip

The daily email summary also points out new knowledge for you to review. The article AI Frontdesk: Setting up the daily email summary explains how to set it up.

Relearn knowledge

In the Learned business knowledge section, click Learn again. The assistant analyses your mailbox again and presents new suggestions for your approval.

The line Latest observed sent email shows the date up to which your sent emails have been analysed.

Delete learned knowledge

Please Note

Deleting removes all entries under Approved knowledge, not just individual ones.

In the Learned business knowledge section, click Clear learned knowledge.

Your assistant then no longer uses the learned knowledge to answer questions on these topics. Until you approve new knowledge, it leaves the relevant emails for you to deal with.

Note on hallucinations

AI systems may occasionally make incorrect or inaccurate statements, known as hallucinations. This means that the AI estimates or interprets answers when no clear information is available.

Please Note:

  • The AI cannot always adhere 100% to your specifications.
  • You should therefore scrutinise important answers carefully.
  • Such deviations are not technical faults, but a typical characteristic of generative AI.

Content