What is eWebinar's integration with Chatbase and how do I set it up?

Integrate with Chatbase, easily train their AI agent (in minutes), and have it respond to chat messages from attendees in your webinars on your behalf.

eWebinar's chat system already gives you the ability to respond to messages from attendees in real-time or later by email. Even so, integrating with an AI-powered chatbot, like Chatbase, may be the best way for you to ensure attendees always get real-time responses while maximizing the scale and reach of your webinars.

Below is an example of a Chatbase AI agent answering questions in our own demo, so you can get a sense of what it is capable of:

How to set up the integration

The initial setup process between eWebinar and Chatbase – which can take as little as 10 minutes – is divided into three parts:

• Part 1: Create and train the AI agent in Chatbase
• Part 2: Integrate Chatbase with eWebinar
• Part 3: Enable the AI agent in specific webinars

Once the integration is set up, it is important to know:

• How the integration works for attendees
• How the integration works in the chat admin

Part 1: Create and train the AI agent in Chatbase

1. Login to Chatbase and create a new AI agent. 

2. Next, train the agent. To do that, you can upload files, enter text directly, add links to your website and/or knowledge base, and enter specific answers to FAQ.

To give you an idea of how easy it is to train the agent watch this 3 minute video to demonstrate how we trained it for our own product demo:

There are two ways you can train the agent that are specific to eWebinar:

A. Download a webinar's chat history as a TXT file and upload it to Chatbase

If there is a particular webinar's chat history you want to use to train your agent, go to the webinar's Chat tab, download it as a TXT file, and upload it to Chatbase.

B. Copy and paste your Saved replies into Chatbase

If you have collected Saved replies in eWebinar, you can add them into Chatbase as individual questions and answers under Q&A. If you do, the AI agent will use your wording when answering these specific questions.

3. Once you have added all relevant training sources, click to create your AI agent. It make take a few minutes to complete the training process.

4. Once the agent is created, you are able to configure various settings. The most important of these is the System prompt (Click on the icon below to access the system prompt).

 

If the agent you are creating is for a SaaS onboarding webinar, for example, you might choose Customer support agent as your base system prompt. If you are using it for a sales demo, you might choose Sales agent. Whichever option you choose, however, you can always customize the prompt afterwards, as needed.

Once you've adjusted you AI agent's settings to your liking, save your changes.

5. The next step is to use the interface in Chatbase's Playground to test how your AI agent responds to messages. If you are unsatisfied with anything, you can update your training sources (under Sources) or your agent's settings (by clicking the icon indicated below).

When you are happy with the results you are getting from your AI agent, you are ready to move on to...

Part 2: Integrate Chatbase with eWebinar

1. In Chatbase, go to your team's Dashboard → Settings → API Keys → Create API Key → Copy it.

2. Next, login to eWebinar, go to Integrations in the Settings section of the Profile menu, and click Chatbase under Available integrations.

3. In the modal that opens, paste the Chatbase API key into the field provided, under Account credentials. and click Connect.

Chatbase is now integrated with eWebinar!

Part 3: Enable the AI agent in specific webinars

Repeat the following steps for every webinar you would like to add the AI agent to:

1.  Edit the webinar in question, go to the Chat tab and, under Moderators, click Add AI moderator. (If you created more than one AI agent in Chatbase, you will be prompted to choose one.)

2. In the modal that opens, configure the AI moderator. You can:

• Upload an avatar and/or specify a Display name.
• Decide if you would like to auto archive conversations handled exclusively by the AI moderator. (Conversations that do not get auto archived include those where any non-AI moderator joined the conversation or where the attendee requested to "connect with a person" but no one responded.)
• Enable the ability for attendees to "connect with a person". (This feature is explained in detail below under How the integration works.)

If you enable "connect with a person", we recommend you update the language of your Auto response so it makes sense in the context of this feature, to something like:

"We are still looking for someone to chat with you. If no one becomes available before the webinar ends, we'll reach out to you later by email."

NOTE: You may also want to reduce the time before the auto response gets triggered to one minute, since attendees will have already received real-time responses.

3. Once you are finished configuring your AI moderator (and updating your Auto response, if relevant), save and publish your changes.

Your Chatbase AI agent will now respond to chat for you for that webinar! (Repeat these steps for as many webinars as you like.)

A note about the "Private message" interaction

Private message interactions continue to function as expected when you add an AI moderator. The setting (shown below) to skip the message if moderator chat is ongoing applies to the AI moderator as well.

If the webinar has no non-AI moderators assigned to it, private messages will appear to come from the AI moderator.

How the integration works

To understand how the Chatbase integration works, let's look at it from two points of view, from the attendee's and from a non-AI moderator's in the chat admin.

How the integration works for attendees

When the AI moderator sends its first message, the attendee will be able to see that it is "Powered by Chatbase", as indicated by their logo.

Provided "connect with a person" has been enabled, a number of attendee responses will cause it to be triggered, such as the attendee:

• Asked something out of the AI moderator's trained scope
• Had an unusually complex question
• Repeated a similar question
• Did not understand the AI moderator's response
• Expressed negative sentiment (e.g., frustration, dissatisfaction, confusion, anger)
• Expressed the desire to interact with a real person
• Asked or wondered if the moderator is AI

Once triggered, a link appears beneath the chat, "👋 Connect me with a person..."

If the attendee clicks the link, the AI moderator is disabled, moderator notifications are triggered (by email, SMS, browser notification, and/or Slack), and the message beneath the chat changes to, "👀 Finding someone to chat with you..."

Moderators can then jump in and chat with the attendee!

If no moderator is able to respond within the set timeframe, the Auto response is triggered so the attendee knows they can expect to hear back from you later. 

How the integration works for (human) moderators in the chat admin

There are a few different scenarios for how the Chatbase integration might work for moderators in the chat admin depending on how the integration is set up.

Here are three main scenarios:

1. Hybrid AI / Non-AI

In this first scenario, both AI and non-AI moderators actively participate in responding to attendee questions, though non-AI moderators only jump in when attendees request  to "connect with a person". 

This is a nice approach when you want to strike a balance between giving immediate responses and the need for a human touch.

Así es como funciona:

• The AI moderator handles all responses to chat messages unless something the attendee says triggers the "👋 Connect me with a person" link to appear and the attendee then clicks it.
• When this happens, the AI moderator is disabled and non-AI moderators are notified by email, SMS, browser notification, and/or Slack (depending on how notifications are set up).
• When the moderator clicks the notification and is taken to the chat admin, they can see how much time is left before the Auto response is triggered.
   
  
• When the moderator sends a response, the Auto response is automatically canceled. (Or they can manually cancel it, if they need more time.)
• When the conversation is complete, the moderator manually archives it.

NOTE: If moderators are not able to respond to a request to "connect with a person" before the webinar ends, the conversation does not get automatically archived. This makes it easy for moderators to see which conversations in the inbox still need a response.

2. Monitored AI

In this scenario, the AI moderator is effectively responsible for responding to all chat messages. However, non-AI moderators still get notified of attendee messages, so they can observe how well the AI is responding to messages and proactively jump in as needed.

As in the first scenario, the AI moderator gets automatically disabled if the moderator sends a message, or the moderator can manually disable it.

AI-only conversations do not get auto archived, so moderators can review them in the inbox and manually archive them. (If they find things that need to be improved, that can make changes to the AI agent in Chatbase.)

This approach is good if your ultimate goal is to implement the next scenario, unmonitored AI, but you want to feel confident that the AI moderator is handling things well before making the switch.

3. Unmonitored AI

In this scenario, the AI moderator is fully responsible for responding to all chat messages from attendees, without any (or very little) oversight. (Someone must proactively go the chat archive to review how well the AI moderator is doing.)

This approach is good if your goal is to be completely free from the responsibility of answering chat and you are satisfied with the responses from the AI moderator.

The disadvantage to this approach is there could be significant attendee dissatisfaction with their chat experience and you would never know it.