How to set up ai chatbots

Setting up AI-powered chatbots with advanced configuration in eCourtDate lets you train chatbots on your agency's knowledge base, connect them to live agents, enable multilingual support, and monitor...

Setting up AI-powered chatbots with advanced configuration in eCourtDate lets you train chatbots on your agency's knowledge base, connect them to live agents, enable multilingual support, and monitor conversations in real time.

Overview

While the basic chatbot setup covers naming, tone, and fallback behavior, advanced configuration unlocks the full potential of AI chatbots in eCourtDate. This guide covers connecting chatbots to knowledge bases, training them on your help articles, configuring handoff to live agents, setting up multilingual support, and monitoring chatbot conversations. These advanced features let your chatbot handle more complex inquiries, serve a broader client population, and seamlessly escalate when human support is needed.

  • What it does: Provides advanced AI chatbot configuration including knowledge base training, live agent handoff, multilingual responses, and conversation monitoring.
  • Why it matters: Maximizes chatbot effectiveness by ensuring accurate responses, smooth escalation paths, and accessibility for non-English-speaking clients.
  • Who uses it: Agency administrators and IT staff responsible for deploying and optimizing client-facing AI tools.
  • Expected outcome: A highly capable AI chatbot that handles complex inquiries accurately, serves multilingual clients, and hands off gracefully to live agents when needed.

Prerequisites

Before setting up advanced AI chatbot features, make sure you have:

  • An active eCourtDate agency on staging or production
  • Super admin access, or the Manage Automations permission
  • A chatbot already created (see How to Create a Chatbot)
  • Help articles published in the categories you want the chatbot to reference
  • Staff users available for live agent handoff (if configuring escalation)
  • Multilingual content prepared if you plan to enable multilingual support

How-To Steps

Step 1: Connect to a Knowledge Base

  1. Go to Admin > Automations in the top navigation bar.
  2. Click the Chatbots tab and click Edit on the chatbot you want to configure.
  3. Scroll to the Knowledge Base section.
  4. Select the article categories you want the chatbot to use as its source of truth:
    • Select broad categories (e.g., "Platform Overview," "Payments," "Events and Calendars") for a general-purpose chatbot.
    • Select specific categories for a focused chatbot.
  5. Click Save to apply your knowledge base selection.

Step 2: Train on Your Help Articles

  1. After selecting knowledge base categories, click the Train button in the Knowledge Base section.
  2. The system indexes all published articles in the selected categories and makes their content available to the chatbot's AI model.
  3. Training typically completes within a few minutes, depending on the number of articles.
  4. Review the training summary to confirm the number of articles indexed and any warnings about articles that could not be processed.
  5. Click Save to finalize the training.

Whenever you add or update articles in the selected categories, click Retrain to update the chatbot's knowledge.

Step 3: Configure Live Agent Handoff

  1. Scroll to the Handoff Settings section.
  2. Enable the Live Agent Handoff toggle.
  3. Configure the handoff triggers:
    • Client Request: The client types a phrase like "speak to a person" or "connect me to an agent."
    • Confidence Threshold: The chatbot's response confidence drops below a specified threshold (e.g., 60%).
    • Max Turns: The conversation reaches a specified number of back-and-forth exchanges without resolution.
  4. Select the Handoff Team from the dropdown. This is the group of staff users who will receive handoff requests.
  5. Enter a Handoff Message that the chatbot displays when transferring the conversation (e.g., "I am connecting you with a team member who can help. Please hold on.").
  6. Click Save to apply handoff settings.

Step 4: Set Up Multilingual Support

  1. Scroll to the Language Settings section.
  2. Enable the Multilingual Support toggle.
  3. Select the languages your chatbot should support from the Enabled Languages dropdown. Available languages match those enabled on your agency.
  4. Configure language detection:
    • Auto-Detect: The chatbot detects the client's language from their first message and responds in that language.
    • Client Preference: The chatbot uses the language set on the client's profile.
    • Ask on Start: The chatbot asks the client to select their preferred language at the beginning of the conversation.
  5. Ensure your knowledge base articles are available in the selected languages for the most accurate responses.
  6. Click Save to apply language settings.

Step 5: Configure Conversation Monitoring

  1. Scroll to the Monitoring section.
  2. Enable the Real-Time Monitoring toggle to allow staff to observe live chatbot conversations.
  3. Configure notification settings:
    • Notify on Handoff: Send a notification to the handoff team when a conversation is escalated.
    • Notify on Low Confidence: Send a notification when the chatbot's confidence drops below the threshold.
    • Notify on Negative Feedback: Send a notification when a client rates the chatbot interaction negatively.
  4. Select the notification channel (email, Slack, or in-app notification).
  5. Click Save to apply monitoring settings.

Step 6: Review Chatbot Conversations

  1. Go to Chats in the top navigation bar.
  2. Filter by the chatbot name to see conversations handled by this specific chatbot.
  3. Click on a conversation to view the full transcript.
  4. Review the chatbot's responses for accuracy and relevance.
  5. Note any recurring questions that the chatbot struggles with and add or update knowledge base articles accordingly.

Step 7: Fine-Tune Response Behavior

  1. Return to the chatbot's edit page under Admin > Automations > Chatbots.
  2. Scroll to the Response Settings section.
  3. Adjust the following settings based on your conversation review:
    • Response Length: Increase or decrease the maximum response length based on client feedback.
    • Citation Style: Choose whether the chatbot includes links to source articles in its responses.
    • Greeting Message: Customize the message clients see when they first open the chat widget.
    • Suggested Questions: Add a list of common questions that display as clickable options when the chat starts.
  4. Click Save to apply changes.

Step 8: Test the Full Configuration

  1. Click the Test button to open the test chat window.
  2. Test in English first to verify knowledge base responses are accurate.
  3. If multilingual support is enabled, test in each enabled language.
  4. Trigger a live agent handoff by typing "speak to a person" or asking a question outside the knowledge base.
  5. Verify notifications are received by the handoff team.
  6. Check the Chats section to confirm the test conversations are logged.

What to Expect

After completing the advanced setup, your AI chatbot handles a wider range of client inquiries with greater accuracy. Multilingual clients receive responses in their preferred language. When the chatbot cannot resolve a question, it seamlessly transfers the conversation to a live agent. Your team receives notifications for escalations and low-confidence responses, allowing proactive intervention. All conversations are logged in the Chats section for review and continuous improvement.

Best Practices

  • Retrain your chatbot whenever you add, update, or archive articles in the knowledge base categories. Stale training data leads to outdated or inaccurate responses.
  • Set the confidence threshold for live agent handoff between 50% and 70%. Too low and the chatbot escalates too often; too high and clients may receive inaccurate responses without escalation.
  • Review chatbot conversations weekly and use recurring client questions to identify gaps in your knowledge base.
  • Add suggested questions based on the most common client inquiries to guide conversations and improve the client experience.
  • Test multilingual responses with native speakers or professional translators to ensure accuracy and cultural appropriateness.

Frequently Asked Questions

Q: How often should I retrain the chatbot? A: Retrain whenever you add or update articles in the chatbot's knowledge base categories. At minimum, retrain monthly to keep the chatbot current with your latest content.

Q: Can the chatbot handle conversations in multiple languages within a single session? A: The chatbot sets the language at the start of the conversation and maintains it throughout. If a client switches languages mid-conversation, the chatbot may not detect the change automatically. The client can start a new conversation in the desired language.

Q: What happens when a live agent is not available during handoff? A: If no agents in the handoff team are online, the chatbot displays the fallback message and offers alternative contact methods (phone number, email, or ticket creation), based on your fallback configuration.

Q: Can I see which articles the chatbot referenced in its responses? A: Yes. If Citation Style is enabled, the chatbot includes links to the source articles in its responses. You can also view referenced articles in the conversation transcript in the Chats section.

Q: Does multilingual support require translated articles? A: The chatbot can respond in multiple languages using AI translation even if your articles are only in English. However, responses are more accurate and natural when your knowledge base includes articles written in the target language.

Troubleshooting

Issue: Chatbot responses are inaccurate after adding new articles. Symptoms: The chatbot references outdated information or does not use newly published articles. Solution:

  1. Open the chatbot and scroll to the Knowledge Base section.
  2. Click the Retrain button to re-index all articles in the selected categories.
  3. Wait for training to complete and review the training summary.
  4. Test the chatbot with questions related to the new articles to verify updated responses.

Issue: Live agent handoff is not working. Symptoms: The chatbot does not transfer conversations to a live agent when it should. Solution:

  1. Open the chatbot and scroll to the Handoff Settings section.
  2. Verify the Live Agent Handoff toggle is enabled.
  3. Check that a Handoff Team is selected and has active staff members.
  4. Review the handoff triggers (client request, confidence threshold, max turns) to ensure they match the expected behavior.
  5. Test by typing "speak to a person" in the test chat window.

Issue: Multilingual responses are inaccurate or incomplete. Symptoms: The chatbot responds in the wrong language or provides poor-quality translations. Solution:

  1. Open the chatbot and scroll to the Language Settings section.
  2. Verify the correct languages are selected in the Enabled Languages dropdown.
  3. Check the language detection mode (Auto-Detect, Client Preference, or Ask on Start) and confirm it matches your needs.
  4. Add or improve articles in the target language within your knowledge base for higher accuracy.
  5. Click Retrain to re-index articles after adding multilingual content.

Issue: Notifications are not being received for chatbot events. Symptoms: Staff do not receive notifications for handoffs, low confidence, or negative feedback. Solution:

  1. Open the chatbot and scroll to the Monitoring section.
  2. Verify the Real-Time Monitoring toggle is enabled.
  3. Check that the correct notification types are enabled (Notify on Handoff, Notify on Low Confidence, Notify on Negative Feedback).
  4. Verify the notification channel (email, Slack, in-app) is configured correctly.
  5. Check your Slack integration or email settings under Admin > Settings if using those channels.

Important: If you continue to experience issues after following these troubleshooting steps, sign in to eCourtDate and create a ticket by clicking the ticket button in the bottom right corner of the screen.

Related Articles

Setting up AI-powered chatbots with advanced configuration in eCourtDate lets you train chatbots on your agency's knowledge base, connect them to live agents, enable multilingual support, and monitor conversations in real time.

### Overview

While the basic chatbot setup covers naming, tone, and fallback behavior, advanced configuration unlocks the full potential of AI chatbots in eCourtDate. This guide covers connecting chatbots to knowledge bases, training them on your help articles, configuring handoff to live agents, setting up multilingual support, and monitoring chatbot conversations. These advanced features let your chatbot handle more complex inquiries, serve a broader client population, and seamlessly escalate when human support is needed.

- **What it does:** Provides advanced AI chatbot configuration including knowledge base training, live agent handoff, multilingual responses, and conversation monitoring.
- **Why it matters:** Maximizes chatbot effectiveness by ensuring accurate responses, smooth escalation paths, and accessibility for non-English-speaking clients.
- **Who uses it:** Agency administrators and IT staff responsible for deploying and optimizing client-facing AI tools.
- **Expected outcome:** A highly capable AI chatbot that handles complex inquiries accurately, serves multilingual clients, and hands off gracefully to live agents when needed.

### Prerequisites

Before setting up advanced AI chatbot features, make sure you have:

- An active eCourtDate agency on staging or production
- Super admin access, or the **Manage Automations** permission
- A chatbot already created (see [How to Create a Chatbot](/how-to-create-a-chatbot))
- Help articles published in the categories you want the chatbot to reference
- Staff users available for live agent handoff (if configuring escalation)
- Multilingual content prepared if you plan to enable multilingual support

### How-To Steps

#### Step 1: Connect to a Knowledge Base

1. Go to **Admin** > **Automations** in the top navigation bar.
2. Click the **Chatbots** tab and click **Edit** on the chatbot you want to configure.
3. Scroll to the **Knowledge Base** section.
4. Select the article categories you want the chatbot to use as its source of truth:
   - Select broad categories (e.g., "Platform Overview," "Payments," "Events and Calendars") for a general-purpose chatbot.
   - Select specific categories for a focused chatbot.
5. Click **Save** to apply your knowledge base selection.

#### Step 2: Train on Your Help Articles

1. After selecting knowledge base categories, click the **Train** button in the **Knowledge Base** section.
2. The system indexes all published articles in the selected categories and makes their content available to the chatbot's AI model.
3. Training typically completes within a few minutes, depending on the number of articles.
4. Review the training summary to confirm the number of articles indexed and any warnings about articles that could not be processed.
5. Click **Save** to finalize the training.

Whenever you add or update articles in the selected categories, click **Retrain** to update the chatbot's knowledge.

#### Step 3: Configure Live Agent Handoff

1. Scroll to the **Handoff Settings** section.
2. Enable the **Live Agent Handoff** toggle.
3. Configure the handoff triggers:
   - **Client Request**: The client types a phrase like "speak to a person" or "connect me to an agent."
   - **Confidence Threshold**: The chatbot's response confidence drops below a specified threshold (e.g., 60%).
   - **Max Turns**: The conversation reaches a specified number of back-and-forth exchanges without resolution.
4. Select the **Handoff Team** from the dropdown. This is the group of staff users who will receive handoff requests.
5. Enter a **Handoff Message** that the chatbot displays when transferring the conversation (e.g., "I am connecting you with a team member who can help. Please hold on.").
6. Click **Save** to apply handoff settings.

#### Step 4: Set Up Multilingual Support

1. Scroll to the **Language Settings** section.
2. Enable the **Multilingual Support** toggle.
3. Select the languages your chatbot should support from the **Enabled Languages** dropdown. Available languages match those enabled on your agency.
4. Configure language detection:
   - **Auto-Detect**: The chatbot detects the client's language from their first message and responds in that language.
   - **Client Preference**: The chatbot uses the language set on the client's profile.
   - **Ask on Start**: The chatbot asks the client to select their preferred language at the beginning of the conversation.
5. Ensure your knowledge base articles are available in the selected languages for the most accurate responses.
6. Click **Save** to apply language settings.

#### Step 5: Configure Conversation Monitoring

1. Scroll to the **Monitoring** section.
2. Enable the **Real-Time Monitoring** toggle to allow staff to observe live chatbot conversations.
3. Configure notification settings:
   - **Notify on Handoff**: Send a notification to the handoff team when a conversation is escalated.
   - **Notify on Low Confidence**: Send a notification when the chatbot's confidence drops below the threshold.
   - **Notify on Negative Feedback**: Send a notification when a client rates the chatbot interaction negatively.
4. Select the notification channel (email, Slack, or in-app notification).
5. Click **Save** to apply monitoring settings.

#### Step 6: Review Chatbot Conversations

1. Go to **Chats** in the top navigation bar.
2. Filter by the chatbot name to see conversations handled by this specific chatbot.
3. Click on a conversation to view the full transcript.
4. Review the chatbot's responses for accuracy and relevance.
5. Note any recurring questions that the chatbot struggles with and add or update knowledge base articles accordingly.

#### Step 7: Fine-Tune Response Behavior

1. Return to the chatbot's edit page under **Admin** > **Automations** > **Chatbots**.
2. Scroll to the **Response Settings** section.
3. Adjust the following settings based on your conversation review:
   - **Response Length**: Increase or decrease the maximum response length based on client feedback.
   - **Citation Style**: Choose whether the chatbot includes links to source articles in its responses.
   - **Greeting Message**: Customize the message clients see when they first open the chat widget.
   - **Suggested Questions**: Add a list of common questions that display as clickable options when the chat starts.
4. Click **Save** to apply changes.

#### Step 8: Test the Full Configuration

1. Click the **Test** button to open the test chat window.
2. Test in English first to verify knowledge base responses are accurate.
3. If multilingual support is enabled, test in each enabled language.
4. Trigger a live agent handoff by typing "speak to a person" or asking a question outside the knowledge base.
5. Verify notifications are received by the handoff team.
6. Check the **Chats** section to confirm the test conversations are logged.

### What to Expect

After completing the advanced setup, your AI chatbot handles a wider range of client inquiries with greater accuracy. Multilingual clients receive responses in their preferred language. When the chatbot cannot resolve a question, it seamlessly transfers the conversation to a live agent. Your team receives notifications for escalations and low-confidence responses, allowing proactive intervention. All conversations are logged in the **Chats** section for review and continuous improvement.

### Best Practices

- Retrain your chatbot whenever you add, update, or archive articles in the knowledge base categories. Stale training data leads to outdated or inaccurate responses.
- Set the confidence threshold for live agent handoff between 50% and 70%. Too low and the chatbot escalates too often; too high and clients may receive inaccurate responses without escalation.
- Review chatbot conversations weekly and use recurring client questions to identify gaps in your knowledge base.
- Add suggested questions based on the most common client inquiries to guide conversations and improve the client experience.
- Test multilingual responses with native speakers or professional translators to ensure accuracy and cultural appropriateness.

### Frequently Asked Questions

**Q: How often should I retrain the chatbot?**
A: Retrain whenever you add or update articles in the chatbot's knowledge base categories. At minimum, retrain monthly to keep the chatbot current with your latest content.

**Q: Can the chatbot handle conversations in multiple languages within a single session?**
A: The chatbot sets the language at the start of the conversation and maintains it throughout. If a client switches languages mid-conversation, the chatbot may not detect the change automatically. The client can start a new conversation in the desired language.

**Q: What happens when a live agent is not available during handoff?**
A: If no agents in the handoff team are online, the chatbot displays the fallback message and offers alternative contact methods (phone number, email, or ticket creation), based on your fallback configuration.

**Q: Can I see which articles the chatbot referenced in its responses?**
A: Yes. If **Citation Style** is enabled, the chatbot includes links to the source articles in its responses. You can also view referenced articles in the conversation transcript in the **Chats** section.

**Q: Does multilingual support require translated articles?**
A: The chatbot can respond in multiple languages using AI translation even if your articles are only in English. However, responses are more accurate and natural when your knowledge base includes articles written in the target language.

### Troubleshooting

**Issue:** Chatbot responses are inaccurate after adding new articles.
**Symptoms:** The chatbot references outdated information or does not use newly published articles.
**Solution:**

1. Open the chatbot and scroll to the **Knowledge Base** section.
2. Click the **Retrain** button to re-index all articles in the selected categories.
3. Wait for training to complete and review the training summary.
4. Test the chatbot with questions related to the new articles to verify updated responses.

**Issue:** Live agent handoff is not working.
**Symptoms:** The chatbot does not transfer conversations to a live agent when it should.
**Solution:**

1. Open the chatbot and scroll to the **Handoff Settings** section.
2. Verify the **Live Agent Handoff** toggle is enabled.
3. Check that a **Handoff Team** is selected and has active staff members.
4. Review the handoff triggers (client request, confidence threshold, max turns) to ensure they match the expected behavior.
5. Test by typing "speak to a person" in the test chat window.

**Issue:** Multilingual responses are inaccurate or incomplete.
**Symptoms:** The chatbot responds in the wrong language or provides poor-quality translations.
**Solution:**

1. Open the chatbot and scroll to the **Language Settings** section.
2. Verify the correct languages are selected in the **Enabled Languages** dropdown.
3. Check the language detection mode (Auto-Detect, Client Preference, or Ask on Start) and confirm it matches your needs.
4. Add or improve articles in the target language within your knowledge base for higher accuracy.
5. Click **Retrain** to re-index articles after adding multilingual content.

**Issue:** Notifications are not being received for chatbot events.
**Symptoms:** Staff do not receive notifications for handoffs, low confidence, or negative feedback.
**Solution:**

1. Open the chatbot and scroll to the **Monitoring** section.
2. Verify the **Real-Time Monitoring** toggle is enabled.
3. Check that the correct notification types are enabled (Notify on Handoff, Notify on Low Confidence, Notify on Negative Feedback).
4. Verify the notification channel (email, Slack, in-app) is configured correctly.
5. Check your Slack integration or email settings under **Admin** > **Settings** if using those channels.

> **Important:** If you continue to experience issues after following these troubleshooting steps, sign in to eCourtDate and create a ticket by clicking the ticket button in the bottom right corner of the screen.

### Related Articles

- [Automations Overview](/automations-overview)
- [How to Create a Chatbot](/how-to-create-a-chatbot)
- [How to Create Chatbots](/how-to-create-chatbots)
- [How to Create Availability Templates](/how-to-create-availability-templates)
- [How to Manage Automations](/how-to-manage-automations)
Share: