How to leverage the AI Agent action in Chat mode

Hi, Angie here from Tines. In this video, we'll focus on the chat mode of our AI agent action.

Chat mode creates a dedicated URL where end users can have back and forth conversations with your agent. Unlike task mode, which runs autonomously, chat mode is designed for interactive experiences where human input guides the conversation.

To configure an AI agent action in chat mode, add a new AI agent action to your story. Then select chat mode from the mode options.

In the configuration panel, you'll see the same system instructions field as in task mode. While the system instructions work the same way, you'll want to tailor your instructions for interactive conversations when using chat mode, such as define the agents persona with a conversational tone, include guidance on how the agent should respond to different types of questions, set boundaries for what topics the agent should or should not discuss, as well as consider including example dialogues to guide the agents conversation style.

Just like in task mode, you can add tools, custom tools, and send a story for your chat interface. To do so, navigate to the tools section in the configuration panel and add tool. Here you can add relevant tools including send to story and custom tools that end users might need during conversations.

The agent will suggest using these tools when appropriate during the chat. Once configured, you can share your agent chat interface via copying the generated URL manually from the copy link option in the toolbar. Or if you need to reference the URL in a connected upstream action, utilize our page function to share the URL more autonomously.

Depending on your access control settings of your AI agent action, end users can access the chat interface directly through their browser via this URL. You can review chat interactions to improve your AI agent actions performance. In this example, we'll take a look at an AI agent action that is set up to chat with users on how they can improve the write-ups of their existing Jira issues. Once the chat is complete, you can view the conversation history in the AI agent actions events panel.

Here, analyze which tools were used during the conversations and use these insights to refine your system instructions and available tools.

Now, some general best practices. Keep system instructions clear and conversational for chat mode.

Only add tools that are relevant to the types of conversations end users will have.

Consider the end user experience when designing your agents persona and responses.

Regularly review conversations to improve your agents effectiveness.

Finally, a few key takeaways from today's video. Chat mode enables interactive conversations through a dedicated URL. System instructions should be tailored for conversational interactions. Tools can enhance the chat experience by allowing the agent to perform actions. Monitoring conversations helps improve your agents performance over time. See you in the next video and happy building.