Home  >   Calculus AI   >  Vtiger Agent Builder
Articles in this section

Vtiger Agent Builder

This article discusses the AI Agent Builder feature in Vtiger.
S
Sreenivas K
20 Jan, 2025 - Updated 2 days ago
Table of Contents

Feature Availability: Calculus AI Add-on, One AI Edition.

Introduction

The Agent Builder in Vtiger is an advanced feature that empowers administrators to build, test, and deploy autonomous AI-powered Agents for performing various jobs.

These AI Agents are equipped to answer questions, create records, update records, add custom fields, and provide intelligent assistance in real time for users and customers. By integrating LLMs and information within the CRM, Agent Builder allows organizations to improve customer experiences, and reduce manual effort.

Building an Agent

To build an AI Agent

  1. Open Generative AI Designer from Main Menu > Platform (or search)
  2. Click on Bot Setup tab and select Agents
  3. Click on Add Agent
  4. Enter Bot Name, Bot Description, and select the Bot Image (optional)
    • Please see the Guidance for Bot Description section (below) to see what makes a good bot description.​​​​​​​
  5. Review the Topics and Actions suggested and make necessary modifications 
  6. Click Save

Bot Description

The text you enter in Bot Description will be used to suggest the Topics and Actions . In the description field, outline your bot’s purpose, key features, and the specific tasks it automates.

  1. Define the Agent's Purpose and Domain:
    1. Clear Objective: What is the AI agent supposed to achieve? (e.g., "Schedule meetings", "Provide customer support", "Summarize research papers")
    2. Domain of Knowledge: What subject matter will it operate within? (e.g., "calendar events", "product information", "scientific literature")
  2. Specify Topic Categories:
    1. Provide Examples: Give the AI a starting point by listing potential topic categories. (e.g., "Meeting time", "Meeting attendees", "Meeting location", "Customer order status", "Product features", "Research findings", "Study methodology")
    2. Explain Relationships: Describe how topics might relate to each other. (e.g., "Meeting time" influences "attendee availability", "Product features" determine "customer needs")
  3. Detail Action Types:
    1. List Possible Actions: Provide a range of actions the AI can take. (e.g., "Create calendar event", "Send email", "Query database", "Generate summary", "Ask clarifying questions")
    2. Link Actions to Topics: Explain how actions connect to topics. (e.g., "If the topic is 'Meeting time', possible actions are 'Suggest times', 'Check availability', 'Create calendar event'")
  4. Provide Data Structures (Optional):
    1. Schemas or Examples: If you have specific data formats (JSON, XML, etc.) or example interactions, include them. This helps the AI understand how to structure its output.
  5. Set Constraints and Guidelines:
    1. Safety and Ethics: Emphasize any restrictions on the generated topics or actions (e.g., "Do not disclose personal information", "Avoid biased or harmful language").
    2. Preferred Style: Guide the AI's tone and language (e.g., "Formal and concise", "Friendly and helpful").

Bot Description Samples

Sample 1: Travel Planning Agent

You are an AI agent designed to assist users with travel planning.

  • Topics: Your topics should include 'Destination', 'Travel dates', 'Budget', 'Accommodation preferences', 'Activities', and 'Transportation'.
  • Actions: You can 'Search for flights', 'Find hotels', 'Suggest itineraries', 'Provide destination information', 'Estimate costs', and 'Ask clarifying questions'.
  • Guidelines: Ensure your suggestions are safe and family-friendly. Be informative and enthusiastic in your responses."

Sample 2: E-Commerce Customer Support Agent

You are an AI Agent designed to help customers with product inquiries, orders, and returns.
Domain: E-commerce, product information, order tracking, customer service policies.
Topics:

  • "Product details" (features, specifications, availability)
  • "Order status" (shipping, delivery, tracking)
  • "Returns and refunds" (policy, process, eligibility)
  • "Payment and billing" (methods, issues, invoices)
  • "Account management" (login, profile, order history)

Actions:

  • "Answer questions from knowledge base"
  • "Track order status"
  • "Initiate return request"
  • "Provide shipping updates"
  • "Generate support ticket"
  • "Escalate to human agent"

Guidelines:

  • Maintain a polite and helpful tone.
  • Prioritize customer satisfaction.
  • Adhere to company policies.

Sample 3 - Personal Productivity Assistant

You are an AI agent designed to help users manage tasks and goals.

  • Domain: Task management, time management, goal setting, productivity techniques.
  • Topics:
    • "Tasks" (deadlines, priorities, progress)
    • "Goals" (short-term, long-term, milestones)
    • "Calendar events" (appointments, meetings, reminders)
    • "To-do lists" (organization, categorization, prioritization)
  • Actions:
    • "Create and schedule tasks"
    • "Set reminders and notifications"
    • "Track progress and provide feedback"
    • "Suggest time management strategies"
    • "Offer motivational support"
  • Guidelines:
    • Be supportive and encouraging.
    • Respect user preferences and priorities.
    • Help users achieve their goals effectively.
 

Agent Playground

Playground helps admins view the sequence of actions performed by agent to respond to a query. Administrators can use this information to troubleshoot and update the actions and topics and improve the responses.  


To open the playground
  1. Open Agent Builder to view the list of agents
  2. Click on Playground icon for the agent you want to troubleshoot.
Was this article helpful?
0  out of  0  found this helpful.
Comments 0
Be the first to comment
© Copyright 2023 Vtiger. All rights reserved.