Now that you have data on how the Agent will reply with the Knowledge it has, there are 3 main levers to refine the responses:

1. Adding, removing, or editing Knowledge
2. Creating Scenarios
3. Adjusting your Brand Voice

Adding, removing, or editing Knowledge

In cases where the Agent is not replying, or replying incorrectly, the issue is likely the Knowledge it has access to.

• Check to confirm the Knowledge exists by navigating to the Knowledge page and searching. Click in to any article and review the information in it.
• If the Knowledge needs to be edited or updated, you can do this is a number of ways.
  • For text articles, simply click the Edit button on the top right and make any edits you need.
  • For web scraped URLs or APIs, click Refresh if the source page has been updated. If the source data is incorrect, you can follow the instructions for creating a temporary Knowledge article with the correct information.
• If Knowledge needs to be added, follow the steps here to create new sources.
• If Knowledge needs to be removed, follow the steps here to archive it.

Creating Scenarios

If the Knowledge is correct, but as a brand or marketer you have specific ways you want certain types of messages answered, you can create a Scenario to direct the Agent. These are more “exceptions” vs rules, and give you a fine level of control.

• Go to the Scenarios page

• Review the existing default Scenarios

• Create a new Scenario for your need by clicking Create on the top right.

• In the modal, select “exactly matches” or “has similar intent to” from the drop down. In most cases you will want “has similar intent to”, so the Agent can flexibly match on meaning and sentiment.

• In the second drop down, choose how you want the Scenario to be handled.

  • If you want the Agent to reference a specific Knowledge article, select “Answer Using” and search for the article. This is common for handling messages about influencer requests for example.
  • If you want the Agent to start a specific Automation, choose “Start Automation.” This is most commonly used in Comment to DM or Story Reply flows.
  • If you want the Agent to simply ignore the message for a human to handle later, choose “Ignore Message.” This is most commonly used for messages expressing frustration, spam, profanity, or other sensitive topics.
  • If you want to send a very specific message, choose “Send Message” and type the message you want sent. (Note: this will only send this message exactly as written. Consider using automations or Knowledge to take advantage of more robust conversational capabilities. If you’re not sure about this, just ask your Tildei team.)

  

Adjusting Your Brand’s Voice

Your Tildei Agent will naturally take on the tone and voice of your brand from your Knowledge. This is especially true if you have loaded blogs or other editorial assets. However, you have a finer level of control that can also influence how your Agent replies.