Learn how to automate customer support with CrewAI’s multi-agent AI system. Improve accuracy, efficiency, and response quality using AI-driven automation.
Introduction
Customer support is a critical aspect of any business, and AI-driven automation is transforming the way companies handle inquiries. Multi-agent AI systems enable multiple AI agents to collaborate efficiently, ensuring quick, accurate, and high-quality responses to customer inquiries.
In this tutorial, we will learn how to automate customer support using CrewAI, leveraging six key elements that enhance agent performance:
- Role Playing
- Focus
- Tools
- Cooperation
- Guardrails
- Memory
By the end of this guide, you will have a fully functional AI-powered customer support system that can handle inquiries, ensure quality, and improve customer satisfaction.
Complete Code
import os
import warnings
from crewai import Agent, Task, Crew
from utils import get_openai_api_key
from crewai_tools import SerperDevTool, ScrapeWebsiteTool
# Suppress warnings
warnings.filterwarnings('ignore')
# Set up API key and environment variables
openai_api_key = get_openai_api_key()
os.environ["OPENAI_MODEL_NAME"] = 'gpt-3.5-turbo'
# Define AI Agents
support_agent = Agent(
role="Senior Support Representative",
goal="Be the most friendly and helpful support representative in your team",
backstory=(
"You work at CrewAI and are providing support to {customer}, an important client."
"Your role is to ensure they receive complete and accurate answers, leaving no room for confusion."
),
allow_delegation=False,
verbose=True
)
support_quality_assurance_agent = Agent(
role="Support Quality Assurance Specialist",
goal="Ensure the highest quality of customer support responses",
backstory=(
"You work at CrewAI, ensuring that all support responses meet the company's standards."
"Your job is to review responses for accuracy, completeness, and a friendly tone."
),
verbose=True
)
# Import and instantiate AI tools
search_tool = SerperDevTool()
scrape_tool = ScrapeWebsiteTool()
docs_scrape_tool = ScrapeWebsiteTool(
website_url="https://docs.crewai.com/how-to/Creating-a-Crew-and-kick-it-off/"
)
# Define Tasks
inquiry_resolution = Task(
description=(
"{customer} just reached out with an important question: {inquiry}."
"Your job is to provide a complete and informative response."
),
expected_output=(
"A detailed response that addresses all aspects of the customer's inquiry."
"Ensure references are included and no details are left out."
),
tools=[docs_scrape_tool],
agent=support_agent,
)
quality_assurance_review = Task(
description=(
"Review the response drafted by the Support Representative for {customer}."
"Ensure the answer is comprehensive, accurate, and professional."
),
expected_output=(
"A fully reviewed and improved response, ready for the customer."
),
agent=support_quality_assurance_agent,
)
# Create the Crew with Memory Enabled
crew = Crew(
agents=[support_agent, support_quality_assurance_agent],
tasks=[inquiry_resolution, quality_assurance_review],
verbose=2,
memory=True
)
# Define Input Values
inputs = {
"customer": "DeepLearningAI",
"person": "Andrew Ng",
"inquiry": "How can I add memory to my CrewAI setup?"
}
# Execute the Support System
result = crew.kickoff(inputs=inputs)
# Display Final Output
from IPython.display import Markdown
Markdown(result)
Step 1: Setting Up the Environment
Before we begin, install the necessary libraries if you’re running this on your machine:
pip install crewai==0.28.8 crewai_tools==0.1.6 langchain_community==0.0.29We also set up OpenAI’s GPT-3.5 Turbo as our default model:
import os
from utils import get_openai_api_key
openai_api_key = get_openai_api_key()
os.environ["OPENAI_MODEL_NAME"] = 'gpt-3.5-turbo'Step 2: Defining AI Agents
Agent: Senior Support Representative
The Senior Support Representative is responsible for answering customer inquiries in a detailed and professional manner.
from crewai import Agent
support_agent = Agent(
role="Senior Support Representative",
goal="Be the most friendly and helpful support representative in your team",
backstory=(
"You work at CrewAI and are providing support to {customer}, an important client."
"Your role is to ensure they receive complete and accurate answers, leaving no room for confusion."
),
allow_delegation=False,
verbose=True
)Agent: Support Quality Assurance Specialist
The Quality Assurance (QA) Specialist reviews responses to ensure they meet customer service standards.
support_quality_assurance_agent = Agent(
role="Support Quality Assurance Specialist",
goal="Ensure the highest quality of customer support responses",
backstory=(
"You work at CrewAI, ensuring that all support responses meet the company's standards."
"Your job is to review responses for accuracy, completeness, and a friendly tone."
),
verbose=True
)Step 3: Enhancing Agent Capabilities
Role Playing, Focus, and Cooperation
- Agents are given defined roles and goals.
- They stay focused on their specific tasks.
- The QA Specialist can cooperate with the Support Agent by requesting changes if needed.
Step 4: Using Tools to Enhance Customer Support
Importing CrewAI Tools
from crewai_tools import SerperDevTool, ScrapeWebsiteToolExample Custom Tools
- Customer data retrieval
- Accessing past conversations
- Fetching data from CRM
- Checking existing bug reports
- Verifying ongoing support tickets
Instantiating Search and Scraper Tools
search_tool = SerperDevTool()
scrape_tool = ScrapeWebsiteTool()To scrape specific documentation pages:
docs_scrape_tool = ScrapeWebsiteTool(
website_url="https://docs.crewai.com/how-to/Creating-a-Crew-and-kick-it-off/"
)Step 5: Defining Tasks
Task: Resolving Customer Inquiries
The Support Agent handles incoming customer inquiries.
from crewai import Task
inquiry_resolution = Task(
description=(
"{customer} just reached out with an important question: {inquiry}."
"Your job is to provide a complete and informative response."
),
expected_output=(
"A detailed response that addresses all aspects of the customer's inquiry."
"Ensure references are included and no details are left out."
),
tools=[docs_scrape_tool],
agent=support_agent,
)Task: Quality Assurance Review
The QA Specialist reviews responses before they reach the customer.
quality_assurance_review = Task(
description=(
"Review the response drafted by the Support Representative for {customer}."
"Ensure the answer is comprehensive, accurate, and professional."
),
expected_output=(
"A fully reviewed and improved response, ready for the customer."
),
agent=support_quality_assurance_agent,
)Step 6: Creating the Crew and Enabling Memory
Memory helps agents retain context, improving their responses over time.
from crewai import Crew
crew = Crew(
agents=[support_agent, support_quality_assurance_agent],
tasks=[inquiry_resolution, quality_assurance_review],
verbose=2,
memory=True
)Step 7: Running the AI Customer Support System
Defining Input Values
inputs = {
"customer": "DeepLearningAI",
"person": "Andrew Ng",
"inquiry": "How can I add memory to my CrewAI setup?"
}Executing the Support System
result = crew.kickoff(inputs=inputs)Displaying the Final Response in Markdown format
from IPython.display import Markdown
Markdown(result)Conclusion
This tutorial demonstrated how to automate customer support using CrewAI’s multi-agent system. By leveraging specialized agents, AI tools, and memory retention, businesses can provide fast, high-quality support with minimal human intervention.
Would you like to implement this AI-driven customer support system? Start building with CrewAI today! 🚀
