The Audience Chatbot draws from all published handbooks an employee (aka audience member) has access to, so they can ask about benefits, time off, expense policies, or anything else and the AI will find the relevant content.

The Chatbot is designed to be direct and accurate. If it doesn't have enough information to answer the question, it will say so rather than guess.

1. Open the Chatbot. Click the "Ask AI" chat button in the top right corner of a live document to open the chat drawer.

2. Type a question. ... into text field at the bottom and press [Enter] or click the send button.

3. Read the response. The Chatbot streams its answer in real time. After it responds, it may show a source card linking to the specific section of the document the answer came from. Employees can click to jump there directly.

4. Keep the conversation going. The chatbot will answer follow-up questions or requests for more detail. Conversation history is saved for 30 days so a convo can be picked up where it was left off.

Audience members can manage conversations in a few ways:

Start a new chat. Users can click New chat in the top right of the AI Chat sidebar at any time to start a new conversation.

View past conversations. Click the clock icon to see a list of past conversations. Click any thread to open it. The most recent conversations appear at the top.

Delete a conversation. To do so, hover over any thread in the history list and click the trash icon to delete it. Deleted conversations are not saved in our logs and are not backed up in any way, so they cannot be recovered. Once they're deleted, they're gone.

Conversations expire automatically.

Chats are private to each audience member and are automatically deleted after 30 days. There is no copy saved in logs or backed up somewhere; once they're gone, they're gone. For more on how the Chatbot handles data, see AI Governance.

The Chatbot shouldn't make things up. It's specifically instructed to only explain based on the content available in a user's documents. If something isn't covered, it should say so.

Answers are as personalized as possible. The Chatbot only uses the content each user has access to. If you're using personalized content, each employee will only receive answers based on the content they can access. Additionally, if any policies are using a merge field dynamic variable (such as {{person.workState}}), the value of that variable for that employee will be sent to the Chatbot before the answer is generated. We also send the employee's first name to the bot.

If an employee has access to multiple policies that look the same (for example they can access the Crime Victim Leave policy for employees in California, Maryland, and Arizona, we'll do our best to show the correct information, but employees may see conflicting information. The Chatbot will do its best to cite its source for the answer or report that there's conflicting information and the user should take a look at the policy text for more information.

This isn't legal or HR advice. For questions about how a policy applies to someone's specific situation, or when someone asks a legal question, the Chatbot will advise them to reach out to their HR team or manager.

You can set the Chatbot to one of three modes on your AI Settings page:

Use Test (Admins Only) mode to try the Chatbot yourself before enabling it for your whole organization. In preview mode, the AI responds based on draft content so you can verify how it will answer questions about policies you haven't published yet. You can also use the preview bar dropdown to preview how different audience members (employees) will experience the Chatbot. Different employees will see different answers if they have access to different content. Preview conversations aren't saved.

For information on governance, usage limits, and data handling, see AI Governance & Privacy in Blissbook.

Don't see AI Settings in your dashboard? This feature may not be included in your current plan. Contact us to learn more.