Skip to main content
All CollectionsAI Assistant
Training the AI Assistants
Training the AI Assistants
Jascha avatar
Written by Jascha
Updated this week

In order for our AI Assistants to answer your candidates' and team members' questions in real-time, you'll need to provide training materials from which the Assistants can draw their knowledge.

You can do so directly in the Mercu portal, either from scratch or using files uploaded to the Mercu Library.

To create a new training file from scratch, first head into the Assistants screen.

If you only use Mercu Hire, you will only see the Candidate Assistant. If you only use Mercu Ops, you will only see the Team Member Assistant. If you use both Mercu Hire and Mercu Ops, you will see both Assistants.

Each Assistant can be trained on its own set of data.


Creating a Training File from Scratch

Step 1: Go to the relevant Assistant, select the "Content" tab and click on 'Add Content' > 'Create New'

Step 2: Add your written content to the editor.

  • The editor supports markdown language.

  • We recommend using headers and bullet points where possible to help the AI model in the backend better digest the information provided.

Step 3: When you're ready, hit "Publish" to have the training file indexed to the Assistant repository, or click "Save & Exit" to save your changes as a draft.

Once the files are published, you can test them by querying the AI Assistant directly in your WhatsApp channel!


Train the Assistant from an Existing Document

To train the Assistant from an existing document:

Step 1: Upload the source file to Mercu Library.

Step 2: Head into the Assistants screen.

Step 3: Click 'Add Content' > 'Add from Library' > Select files that you wish to upload to the AI Assistant.

Step 4: Once the file's status changes from 'Extracting' to 'Draft,' click on the file to view the extracted content.

Step 5: Review and edit the extracted content as required.

  • The editor supports markdown language.

  • We recommend using headers and bullet points where possible to help the AI model in the backend better digest the information provided.

Step 6: When you're ready, hit "Publish" to have the training file indexed to the Assistant repository, or click "Save & Exit" to save your changes as a draft.

Once the files are published, you can test them by querying the AI Assistant directly in your WhatsApp channel!

Did this answer your question?