How to Create Agents

An Introduction to Creating your Agents

Adding an Agent

In order to add a new agent in your application, you will first need to be in the administration portal. Locate the 'Agents' tab in your left-hand navigation bar under 'Manage'.

When agents have been created, you'll find a comprehensive list of your previously configured ones along with their corresponding handles displayed here.

For instance, within this application, the agents named Michelle and Angie have been set up. Their respective handles, @Michelle and @Angie, are visible in the list under the 'Agents' tab.

To initiate the creation of a new agent, select the 'add an agent' button.

Each agent represents a specific use case and purpose. Often, an agent can be built for a task or role. They can provide answers to any queries, if they have been connected to the right data sources, and execute tasks that align with their role.

Setting up your agent

After clicking the 'Add an agent' button, you'll gain access to configuring the agent according to your preferences.

Settings

Next, under the 'Settings' tab, you can add in the agent name, agent handle, and a brief agent description. Under this tab, you can also control the agent visibility. Toggling this button on will enable your end-users to see and interact with the agent created, in the front-end of the application. We recommend that all the agent configuration and testing be completed prior to enabling its access to end users.

Behavior

Following this, you can begin setting up the agent's behavior within the 'Behavior' tab. Here, you're able to input behavior instructions including the agent's purpose, personality, policy rules, output formatting, and any additional information that may assist the agent in answering questions effectively and efficiently.

You can also add in the interaction processes for your agent. Including icebreaker message instructions, and a completion step.

Please note: You can find more information about each of these by either clicking the small question mark icon next to the headings or alternatively clicking here.

Capabilities

Next, you can set up your agent capabilities under the 'capabilities' heading. An agent capability is a specific set of steps that this agent is able to perform to the user. In order to find out more about capability configuration for your agents, please read our 'capabilities' article here.

Tools

In the next step of agent configuration, you'll choose the tools you want the agent to utilize when answering end-users questions, within the 'Tools' tab.

Here, you have the option to decide if you want the agent to utilize the clarification tool and prompt library.

Further details about these functionalities can be found here.

Lastly, you can choose additional response details that the agent will incorporate when answering user inquiries. This includes including whether to display segments, topics, and speakers to the user.

Please note: You do not have to select all 3 of these features together for the same agent.

Design

Under the Design tab, the agent's primary color and text color can be configured.

To change the color simply click the 'select color' button and choose your color from the color wheel or from your designated color library.

Knowledge

Under the 'Knowledge' tab, you can select the knowledge banks that the agent can have access to.

Please note: The agent can have access to the information from multiple knowledge banks. More than one knowledge bank can be selected here.

Once the selected knowledge banks have been toggled on and saved, the agent will then be able to utilize all information from those associated banks when answering end-user questions.

Bias

Under the Bias tab select Bias Assessment to see a summary of the Agent’s available Banks of Data. This provides information on the key themes and concepts of the bank's content.

The Weighted Concepts tab displays the concepts and the related segments, with the highest weighted concepts at the top and the lowest at the bottom.

Testing

Under the Testing tab, you can run tests and compare agent responses to understand end-users experiences. Further further details on the agent testing framework can be found here.

Preview

Under the 'Preview' tab, you can practice trialing out your agent to see how it responds to questions asked. We recommend you take your time testing your agent here prior to sharing it with your end-users, to ensure it is performing correctly.

Access

The Access tab allows you to choose which of your accounts have access to the selected agent.

The agent can either be configured with open access - where all application members are able to view and search published content, including users who are members of any account; or restricted access, where only members of selected accounts are able to view and search published content in the knowledge bank.

To restrict access to your agent to specific accounts, select the 'restricted access' button and click save.

Next, you can select from your list of accounts below, the ones you wish to have access to the selected agent. To do this, toggle on the button next to the account you wish to grant access to the agent, and click save.

You can additionally grant access to all of you accounts in one go by selecting the 'grant access to all' button, or revoke access for all accounts in one go by selecting the 'revoke access for all' button.

In order to find out more about access management, please refer to our article titled 'Agent entitlement and Knowledge banks'.

Last updated