Customize your POC: This guide walks through building a working example of the Conversation Assist capabilities within 30 minutes. The provided examples and starter resources are intentionally simple to move through the process quickly. Keep an eye out for "Customize your POC" sections like this one for indications as to where you can further develop and customize your implementation of a Conversation Assist proof-of-concept to fit your specific needs.

This section walks through the steps that are necessary to complete before configuring Conversation Assist. These steps include the creation of the following:

• Airline domain
• Airline-specific knowledge base to serve FAQ articles tied to intents in the NLU domain
• User profiles and associated skills to be tied to both human and bot agents
• Several bots in Conversation Builder with dialogs triggered by intents from the starter pack domain
• Web messaging engagement directed to the signed-in “human” skill

Before you begin

Before you get started with this tutorial, download this ZIP file, which contains a few, necessary resources.

Step 1: Activate Generative AI features

You need administrator access on the account to activate our Generative AI features. Contact your admin if you need assistance.

1. Log into Conversational Cloud.

2. Open the menu on the left side of the page, and select Manage > Management Console.

   

3. In the Management Console, search for “Generative AI” to display the Generative AI Enablement page.

   

4. Click Sign to activate.

   

5. In the pop-up window, specify the following:
   • Brand name: Acme Airline
   • Industry: Transportation

   During queries to the Large Language Model (LLM) service, your brand name and industry are included in the prompt that gets sent to the service. It helps the responses to stay in bounds, i.e., specific to your brand, with fewer hallucinations.

6. Review the terms and conditions, and check the box to indicate that you’ve done so. Click Agree.

   

7. Review the info in the confirmation window: You’ll need to log out and back in. And it might take 15 minutes or so before the features are active. Click Ok, got it.
8. Wait 15 minutes or so before proceeding to the next step.
9. Log out of Conversational Cloud and then log back in.

Step 2: Create the Airline NLU domain

LivePerson offers a variety of domain-specific starter packs (prebuilt domains) to help you get up and running with Intent Manager quickly. After signing into the LivePerson Conversational Cloud, navigate to the Conversational AI portal and complete the following steps to build out the NLU domain that will be used in the Conversation Assist solution.

Customize your POC: For demonstration purposes, this guide assumes a build-out suited for an Airline client. Add in a vertical-specific, pre-built domain to customize for your use case, or optionally add additional domains to supplement the primary one you are using. Conversation Assist can pull from multiple domain intents to offer recommendations to agents.

1. Open the menu on the left side of the page, and select Automate > Intent Manager.

   

2. Click Build on the menu bar, and then click Add Domain in the upper-right corner.
3. In the window that's displayed, select Prebuilt domains along the top.

4. Move your mouse over the Airlines domain, and click Add.

   

   Once completed, the LP_Airlines domain is created, fully trained and containing intents for use with any airline.

Step 3: Create the Airline FAQ knowledge base

With the NLU domain created, the next step is to create a knowledge base with articles to be tied to domain intents. For the purpose of this demonstration, a CSV file has been provided to simplify the process. You downloaded this file earlier.

Customize your POC: This example uses a single knowledge base in order to quickly showcase the capabilities of recommending articles using Conversation Assist. Adding additional knowledge bases to showcase that Conversation Assist can pull from multiple data sources can help to create a more substantial and impressive demonstration with your proof-of-concept. Each additional knowledge base can pull from any domain you previously created to showcase their functionality.

1. Open the menu on the left side of the page, and select Automate > KnowledgeAI to open KnowledgeAI.

   

2. Click Add knowledge base in the upper-right corner.

   Don't see the Add knowledge base button? If you're working with a brand new account, you might need to click Let's go on the landing page instead. Then proceed to the next step.

3. On the AI Enabled tab of the window that appears, specify the following:

   • Knowledge Base Name: Airline FAQ
   • Domain: LP_Airlines
   • Language: English
   • Content source: CSV
     • Use the CSV file that is in the ZIP file that you downloaded earlier.

4. Click Save to upload your knowledge base articles, with intents from the LP_Airlines domain automatically associated with each article.

5. Click Articles in the menu bar to review the articles.

   

   Note: The intents will be automatically associated as long as you created the NLU model as per the previous section. If you didn't do this, you can make your intent associations afterward by aligning the intents with the correct article:

   Article Title	Intent
   Seating assignment	ask about seat assignment
   Refund policy	request refund or credit
   Claim reward miles	claim reward miles
   Baggage policies	ask about baggage
   Upgrade my seat	upgrade seat
   Cancel flight	cancel flight
   Change flight	change flight
   Reschedule disrupted flight	reschedule disrupted flight
   Book a flight	book flight

   Next, you test out the matching of articles to consumer queries using the Answer Tester.

6. In the Answer Tester, enter a few consumer queries, click Get answers, and evaluate the results. Enter queries such as:

   • I'm going on my honeymoon. How do I book a flight?
   • I just got a bonus at work, so I want to upgrade my seat.

   

   

   You can ignore the warning in the UI about the default value for the Enrichment prompt. For the purpose of this tutorial, the default prompt will work just fine. In a real world scenario, you should select and use one of your own prompts as soon as possible. (You can copy one of LivePerson's if you choose.)

   The Enrichment prompt is the prompt to send to the LLM service when the consumer’s query is matched to articles in the knowledge base. It instructs the LLM service on how to use the matched articles to generate an enriched answer.

Step 4: Create the users and skills

Every bot and human agent requires a user profile to be deployed and accessible from within the Conversational Cloud. This section provides the details needed to update the logged-in human agent with an associated skill, as well as the details needed to create bot agents for the two bots that will be uploaded. For more details on this procedure, please see this article in the Conversation Builder Getting Started with Bot Building tutorial.

Customize your POC: This guide provides two bots that Conversation Assist can source recommendations from. These demonstration bots do not provide much functionality out of the box, so bring in your own domain-specific bots to fit the needs of your POC. Keep in mind that any additional bots created will need their own corresponding bot user agents and skills.

1. Open the menu on the left side of the page, and select Manage > Users & Skills to open the User Management section of the Conversational Cloud.

2. In the list of users, select your logged-in human user. In the Edit user form, scroll to the bottom. In the Skills field, enter “Agent,” and click the Create skill link that appears. Then click Save.

   

3. Back on the Users tab, click + Add user in the lower-left corner. Fill out the Add user form with the following user details:

   • User type: Bot
   • Login name: Booking Bot
   • Email: Your email address
   • Nickname: Booking Bot
   • Name: Booking Bot
   • Choose login method: API key
   • API key: Generate API key
   • Assignment → Assign Profile: Agent

   Click Save.

4. Repeat the same process to create a second bot user with the following details:

   • User type: Bot
   • Login name: Seating Bot
   • Email: Your email address
   • Nickname: Seating Bot
   • Name: Seating Bot
   • Choose login method: API key
   • API key: Generate API key
   • Assignment → Assign Profile: Agent

   

Step 5: Import the Conversation Builder bots

In this step, you create two, new bots in Conversation Builder using the JSON files that are in the ZIP file that you downloaded earlier:

• Booking Bot
• Seating Bot

These bots have been created with dialog starters matched to several of the LP_Airlines intents, and they will be used to recommend actions to agents engaged in conversations with users.

1. Open the menu on the left side of the page, and select Automate > Conversation Builder to open Conversation Builder.

2. Click Import Bot in the upper-right corner, and select the "Booking Bot" JSON file that you downloaded. Click Open.

   Note: If you have completed the previous step of creating the LivePerson domain, the intents will be automatically associated with the correct dialogs in the bots. However, if you have not yet done so, you will need to connect them manually to the appropriate dialog starters using Conversation Builder's Assist tool.

3. Within the Booking Bot (now opened), click Agent Connectors on the menu bar at the top.

   If you don't see this menu option, it's in menu to the right of the options that you do see. Just click the three-dot menu icon.

4. Click Add Agent Connector, and in the resulting form, select your "Booking Bot" user agent from the Agent User ID dropdown. Ensure that the Conversation Type is set to "Messaging," and click Save.

5. Click Start to start the agent connector.

   

6. Click Back to return to the bot listing page.

7. Follow the same process to upload the "Seating Bot" that you downloaded, to connect the Seating Bot user agent via an agent connector, and to start the agent connector.

   Once finished, you will have two bots up and running with agent connectors.

   These bots are for demonstration purposes, so they don't do much beyond serving a dialog to a user based on the intent. No booking or seating functionality is built in. As such, they can be modified to meet your use case.

What's next?

Continue on to the next step in the tutorial.