Step 2 – Configure a Knowledge Skill Agent

General

In the Agent AI Config Page screen, brief your knowledge skill agent with instructions. You will find suggestions for proven defaults to get you started.

Persona

In the Persona screen, you can customize how the agent interacts with callers by editing its prefilled personality and communication style. The Persona settings include:

Role

Define the role in which the agent should generate responses.

Conversation Style

Direct how knowledge skill should formulate responses. You can provide instructions on the tone, length, and format of the responses. For example, you could request concise sentences, an empathetic tone, and an acknowledgment of the user's query at the beginning of the response.

Conversation Strategy

Specify how the agent should use the information it provides.

---

Phone

The Phone Screen provides access to the Voice settings and gives you the option to add elements to compensate for latency in the Sounds tab.

Voice

Voice Tab

You must match the following fields to the configuration that is configured for the Phone environment in your rule-based automation bot:

• Agent Voice Provider

• Agent Voice (TTS)

• Agent Language (TTS)

• Caller Language (STT)

Sounds

Sounds Tab

• End of Speech Messages: Here you can add messages that will be played to the caller. Enter a message, which can be either a word or a sentence. Press Enter to add a new word or sentence to be played. Each new line will be played in sequential order.

• End of Speech – Delay before First Played: Here you can define the delay in milliseconds before the messages defined in the End of Speech Messages field will be played. The default value is 200 milliseconds, both for end-of-speech message and for delay of sound.

• Agent is Busy Sounds: Here are the sounds that will be played to the caller when they are waiting for an agent's response. The list of sounds displayed in the image above are included by default.

Keyboard clicking sounds are included by default. You can use them or remove them.

To add more sounds:

1. Click on the + Sounds button. A new empty URL field will be added to the end of the list.

2. Copy and paste the URL of the new sound in the empty URL field. The new sound file URL must end in .wav.

To delete sounds:

Click the Trash icon next to the sound.

---

Skills

In the Skills screen, you can can configure how to make use of the knowledge sources and branches in rule-based automation.

Knowledge

In the Knowledge screen, you can specify which collections the agent should use as references when generating answers. The configurations include:

Generate answers from

Choose from the available knowledge collections. Multiple collections can be selected.

Similarity threshold

Set the minimum similarity score required for using an answer. A higher threshold ensures greater accuracy in the responses used; a lower threshold broadens the acceptance criteria. The default threshold is 0.7.

Source chunk amount

Determine the maximum number of sources the agent can consult to generate responses. The default limit is 4.

Congratulations! You have successfully configured Knowledge Skill.

---

Branches

By default, answers generated by the knowledge skill are sent to the message_to_user branch in the Rule-Based Dialog Designer. The answer is stored in a variable called user_message, which you can use in rule-based automation.

You can set up additional branches here, which will automatically appear in the Rule-Based Dialog Designer. You can also add variables to these branches, and they will be passed to the dialog designer for use in rule-based automation.

Create a New Branch

To create a new Branch:

1. Click the + New Branch button, shown below:

   

   The following displays:

   
2. Rename the branch in the Name field.

3. Optionally, provide a description for the branch.

4. Optionally, add variables to the branch, which will be passed to the dialog designer in the rule-based automation app.

Branch Examples

No Match

If the knowledge collection does not have an answer to a user's question, you can add a noMatch branch.

talk_to_human

For requests such as “I want to speak to a human”, you can create a talk_human_branch to transfer the call to a human agent.

Continue to the next page to configure the a Knowledge Skill Agent.