AI Hub

GoChat's AI Hub provides a suite of AI-powered tools designed to enhance chatbot capabilities, streamline customer interactions, and automate complex tasks.

By following this documentation, you can effectively utilize GoChat's AI Hub to create intelligent chatbots that automate tasks, improve customer engagement, and drive business growth.

Key Features

AI Agents: AI Agents introduce AI functionality to chatbots in an easy-to-manage way. They can be customized with specific personas, skill sets, and constraints to handle various customer interactions, from appointment booking to customer support.
AI Functions: These allow AI Agents to perform specific actions, such as fetching data from external sources or updating third-party platforms. AI Functions can capture user details, retrieve available time slots, and book appointments.
AI Tasks: AI Tasks enable the creation of follow-up messages and other automated actions based on the analysis of chat history. They help re-engage inactive users and ensure a seamless conversational experience.

Getting Started with AI Hub

Accessing AI Hub: Navigate to the AI Hub in the left-hand menu of the GoChat platform. Here, you'll find options to manage AI Agents, AI Functions, and AI Tasks.
Creating an AI Agent:
1. Click the "AI agent" button.
2. Give the agent a descriptive name, such as "Appointment Booking Agent".
3. Provide a detailed description outlining the agent's skills and responsibilities.
4. Configure AI Model Settings: Choose between OpenAI, DeepSeek, and X AAi. OpenAI is recommended when using AI Functions due to its stability. Select an appropriate model (e.g., GPT-4 Turbo) and adjust parameters like temperature, frequency penalty, and presence penalty. Adjust the number of chat messages before auto-summarization to save character space.
5. Define the Agent Prompt: Define the agent's persona, skills, and constraints. Use the "generate agent prompt" feature to automatically populate these sections based on the agent's description. Review and adjust the generated prompt to ensure it aligns with your requirements. Add additional guidelines to the prompt, such as "ask for each user detail separately".
6. Save the AI Agent.
Creating AI Functions:
1. Navigate to the AI Functions section in the AI Hub.
2. Click the "plus AI Function" button.
3. Define the function's purpose such as, "capture user details".
4. Add Parameters: Define the parameters that the function needs to capture, such as first name, last name, and email. Specify if each parameter is required and enable memory to remember the values that were given earlier in the conversation. Map each parameter to a corresponding system field to save the captured data.
5. Generate Function Prompt: Use the "generate function prompt" feature to create a prompt based on the function's description and parameters. Review and customize the prompt to include specific instructions and constraints.
6. Set up a Workflow: Connect the AI Function to a workflow that performs the necessary actions, such as fetching available time slots or creating a CRM contact. Use the AI Function Result node to return the fetched values back to the AI Agent.
7. Save the AI Function.
8. Do not forget to add the new function to an "AI agent"
  This is how the workflow for the Function looks like:
Creating AI Tasks:
1. Go to the AI Tasks section within the AI Hub.
2. Click the "plus AI Task" button.
3. Provide a name and a detailed task prompt. For example: "Create an engaging follow-up message with the user who became inactive. Analyze the current chat history and determine where the user left the conversation and create the follow-up message based on that".
4. Configure the AI Model: Select an appropriate AI model like GPT-4 Turbo.
5. Save the AI Task.
Activating AI Agents in the Flow Builder:
1. Open the Flow Builder and navigate to the desired flow.
2. Add an action node and select "AI Actions".
3. Choose "Activate AI Agent" and select the primary AI Agent from the dropdown menu.
4. Optionally, add secondary agents to enrich the primary agent with additional functions.
5. Set an idle timeout to automatically exit the AI Agent node after a period of user inactivity.
6. Save the output payload to a JSON field for later use.

Tips and Best Practices

Detailed Descriptions: Provide detailed descriptions for AI Agents, AI Functions, and AI Tasks to help the AI understand their roles and responsibilities.
Specific Instructions: Include specific instructions and constraints in the agent and function prompts to guide the AI's behavior.
Error Handling: Implement error handling in AI Functions to gracefully manage invalid inputs or unexpected issues.
Testing: Thoroughly test AI Agents and Functions to ensure they are working as expected.
Follow-Up Messages: Use AI Tasks to create engaging follow-up messages that re-engage inactive users and continue the conversation.
Leverage Memory: Enable memory in AI Functions to remember user details and provide a more personalized experience.
Monitor Performance: Regularly monitor the performance of AI Agents and Functions to identify areas for improvement.

Use Case: Appointment Booking:

Automate the process of scheduling appointments by capturing user details, fetching available time slots, and booking appointments with third-party platforms.
Example of "Persona & Role" for the agent

The Appointment Booking Agent is responsible for scheduling appointments with users. This agent must capture user details, such as first name, last name, and email, fetch available timeslots, allow the user to select their preferred time, and finalize the appointment booking process. The tone should be professional and friendly, ensuring a smooth and positive user experience throughout the scheduling process.

Example of "Skills"

## Skills

### Skill 1: User Detail Capture

Actions: Prompt the user to provide their first name, last name, and email address.
Steps:
Greet the user warmly and introduce the appointment scheduling process.
Ask for the first name and wait for the user's response.
Ask for the last name and wait for the user's response.
Ask for the email address and wait for the user's response.
Validate the email format and ensure all details are captured correctly before proceeding.

### Skill 2: Timeslot Retrieval

Actions: Fetch available timeslots from an external scheduling system.
Steps:
After capturing user details, query the scheduling system for available timeslots.
Ensure that the query is conducted within the relevant date range (e.g., next week).

3. Handle cases where no slots are available by informing the user and asking if they would like to check a different date.

### Skill 3: Timeslot Selection

Actions: Present the available timeslots to the user for selection.
Steps:
Display a list of available timeslots in a clear and organized manner.
Prompt the user to choose their preferred timeslot by providing the available options.
Wait for the user's selection and validate their choice to ensure it matches one of the presented options.

### Skill 4: Appointment Booking

Actions: Confirm the selected timeslot and complete the appointment booking.
Steps:
Once the user selects a timeslot, summarize the appointment details (date, time, user name, and email).
Confirm the booking with the user, asking for approval to finalize the appointment.
Submit the booking to the scheduling system.
Send a confirmation to the user's email and provide additional instructions if necessary.

Example of "Constraints"

## Constraints

The agent must only process requests in a single user session and should not store personal data beyond the session.
Ensure that all prompts and responses are clear and user-friendly.
Handle errors in data input gracefully, providing clear guidance on how to correct mistakes (e.g., invalid email format).
If at any step the user decides to cancel, offer an option to exit the booking process politely.
Response times should be quick to maintain user engagement.

Other use cases

Customer Support: Provide instant answers to frequently asked questions and handle common customer inquiries with AI-powered chatbots.
Lead Generation: Capture email addresses and phone numbers automatically to grow your contact list and re-engage leads.
Personalized Recommendations: Offer personalized product recommendations or content suggestions based on user preferences and behavior.
Automated Follow-Ups: Send automated follow-up messages to inactive users to re-engage them and encourage them to complete their desired actions.

Troubleshooting

AI Messages System Field: Check the bot user's AI Messages system field to track the conversation between the user and the AI Agent and identify any issues.
Live Chat: Monitor the live chat to see how the AI Agent is interacting with users and identify any areas for improvement.
Function Calls: Verify that the correct functions are being called at the appropriate times. If a function is not being triggered, check the AI Agent prompt and ensure it includes clear guidelines on when to call the function.
Custom Field Values: Check that the custom field values are being saved correctly. If a value is not being saved, review the AI Function and ensure it is properly configured to capture and store the data.

FAQ on AI Hub Features and GoChat Platform

What are the core components of the AI Hub and how do they work together?
- The AI Hub consists of AI Agents, AI Functions, and AI Tasks. AI Agents serve as the primary conversational AI, handling user interactions based on defined personas, skills, and constraints. AI Functions are workflows that perform specific actions (like fetching data or booking appointments) and providing data to the agent. AI Tasks are background operations, such as generating follow-up messages based on chat history. These components work in concert to automate complex processes, as exemplified in the demonstration of scheduling an appointment. The agent handles the conversation, functions retrieve data and execute actions, and tasks ensure continued engagement.
How do I create an AI Agent, and what are the key settings I need to configure?
- To create an AI Agent, navigate to the AI Hub, select "AI Agents," and click "Create AI Agent." Provide a name and description for the agent. Crucial settings include selecting the AI model (OpenAI is recommended when using AI functions), adjusting parameters like temperature (for response creativity) and the number of chat messages before auto-summarization (for context retention and space-saving). You can define the agent's Persona, skills, and constraints manually or use the "Generate Agent Prompt" feature to automatically create these based on the description. Finally, you need to activate the AI agent inside the flow builder.
What are AI Functions, and how do I use them to enhance my AI Agent's capabilities?
- AI Functions are workflows that perform specific actions to enhance the AI Agent's capabilities. Create them in the AI Hub under "AI Functions," defining parameters to capture from the user (like first name, last name, email) along with instructions. You can then generate a function prompt that describes what the function needs to do and what the constraints are. These parameters are passed to a linked workflow. For example, the function could capture user details, trigger a workflow to fetch available time slots from a calendar system using these user details, and return the time slots back to the AI Agent to present to the user. Functions can also trigger other subflows to save data or call third-party APIs.
How do AI Tasks work, and what is an example of how they can be used to improve user engagement?
- AI Tasks are background processes that execute independently, enhancing the user experience. To create an AI Task, go to AI Hub and click "Create AI Task". You then create a task prompt and add the AI Task to the flow builder with a AI Action -> AI Task. A prime example is generating follow-up messages for inactive users. An AI Task can analyze the chat history, determine where the user left off, and craft a personalized, engaging message to re-engage them. For instance, if a user provided their first name but didn't complete the process, the task might generate a message reminding them to provide their last name and email to complete the booking.
How can I troubleshoot my AI Agent to ensure it is functioning correctly?
- Troubleshooting involves two key methods: checking the "AI messages" system field for the bot user and reviewing the live chat. The "AI messages" field provides a detailed record of the conversation between the user and the AI Agent, including the functions called, the arguments passed, and the AI's replies. The live chat shows the conversation from a user perspective and confirms that the correct AI Agent is being triggered. By examining these logs, you can identify issues such as incorrect function calls, missing data, or prompt-related problems, allowing you to adjust the AI Agent or AI Function accordingly.

PreviousFlow builder NextFields & variables

Last updated 1 month ago