Repository Public
1[CrewAI](https://docs.crewai.com/introduction) is a framework for orchestrating2autonomous AI agents that collaborate to achieve complex goals. It lets you3define agents by specifying roles, goals, and backstories, and then define tasks4for them.56This example demonstrates how to build a multi-agent system for analyzing7customer support data to identify issues and propose process improvements using8Gemini 3 Flash, generating a report intended to be read by a Chief Operating9Officer (COO).1011The guide will show you how to create a "crew" of AI agents that can do the12following tasks:13141. Fetch and analyze customer support data (simulated in this example).152. Identify recurring problems and process bottlenecks.163. Suggest actionable improvements.174. Compile the findings into a concise report suitable for a COO.1819You need a Gemini API key. If you don't already have one, you can [get one in20Google AI Studio](https://aistudio.google.com/apikey).2122 pip install "crewai[tools]"2324Set your Gemini API key as an environment variable named `GEMINI_API_KEY`, then25configure CrewAI to use the Gemini model.2627 import os28 from crewai import LLM2930 gemini_api_key = os.getenv("GEMINI_API_KEY")3132 gemini_llm = LLM(33 model='gemini/gemini-3.8-flash',34 api_key=gemini_api_key,35 temperature=1.0 # Use the Gemini 3 recommended temperature36 )3738## Define components3940Build CrewAI applications using **Tools** , **Agents** , **Tasks** , and the41**Crew** itself. The following sections explain each of these components.4243### Tools4445Tools are capabilities that agents can use to interact with the outside world or46perform specific actions. Here, you define a placeholder tool to simulate47fetching customer support data. In a real application, you would connect to a48database, API or file system. For more information on tools, see the [CrewAI49tools guide](https://docs.crewai.com/concepts/tools).5051 from crewai.tools import BaseTool5253 # Placeholder tool for fetching customer support data54 class CustomerSupportDataTool(BaseTool):55 name: str = "Customer Support Data Fetcher"56 description: str = (57 "Fetches recent customer support interactions, tickets, and feedback. "58 "Returns a summary string.")5960 def _run(self, argument: str) -> str:61 # In a real scenario, this would query a database or API.62 # For this example, return simulated data.63 print(f"--- Fetching data for query: {argument} ---")64 return (65 """Recent Support Data Summary:66 - 50 tickets related to 'login issues'. High resolution time (avg 48h).67 - 30 tickets about 'billing discrepancies'. Mostly resolved within 12h.68 - 20 tickets on 'feature requests'. Often closed without resolution.69 - Frequent feedback mentions 'confusing user interface' for password reset.70 - High volume of calls related to 'account verification process'.71 - Sentiment analysis shows growing frustration with 'login issues' resolution time.72 - Support agent notes indicate difficulty reproducing 'login issues'."""73 )7475 support_data_tool = CustomerSupportDataTool()7677### Agents7879Agents are the individual AI workers in your crew. Each agent has a specific80`role`, `goal`, `backstory`, assigned `llm`, and optional `tools`. For more81information on agents, see the [CrewAI agents82guide](https://docs.crewai.com/concepts/agents).8384 from crewai import Agent8586 # Agent 1: Data analyst87 data_analyst = Agent(88 role='Customer Support Data Analyst',89 goal='Analyze customer support data to identify trends, recurring issues, and key pain points.',90 backstory=(91 """You are an expert data analyst specializing in customer support operations.92 Your strength lies in identifying patterns and quantifying problems from raw support data."""93 ),94 verbose=True,95 allow_delegation=False, # This agent focuses on its specific task96 tools=[support_data_tool], # Assign the data fetching tool97 llm=gemini_llm # Use the configured Gemini LLM98 )99100 # Agent 2: Process optimizer101 process_optimizer = Agent(102 role='Process Optimization Specialist',103 goal='Identify bottlenecks and inefficiencies in current support processes based on the data analysis. Propose actionable improvements.',104 backstory=(105 """You are a specialist in optimizing business processes, particularly in customer support.106 You excel at pinpointing root causes of delays and inefficiencies and suggesting concrete solutions."""107 ),108 verbose=True,109 allow_delegation=False,110 # No tools needed, this agent relies on the context provided by data_analyst.111 llm=gemini_llm112 )113114 # Agent 3: Report writer115 report_writer = Agent(116 role='Executive Report Writer',117 goal='Compile the analysis and improvement suggestions into a concise, clear, and actionable report for the COO.',118 backstory=(119 """You are a skilled writer adept at creating executive summaries and reports.120 You focus on clarity, conciseness, and highlighting the most critical information and recommendations for senior leadership."""121 ),122 verbose=True,123 allow_delegation=False,124 llm=gemini_llm125 )126127### Tasks128129Tasks define the specific assignments for the agents. Each task has a130`description`, `expected_output`, and is assigned to an `agent`. Tasks are run131sequentially by default and include the context of the previous task. For more132information on tasks, see the [CrewAI tasks133guide](https://docs.crewai.com/concepts/tasks).134135 from crewai import Task136137 # Task 1: Analyze data138 analysis_task = Task(139 description=(140 """Fetch and analyze the latest customer support interaction data (tickets, feedback, call logs)141 focusing on the last quarter. Identify the top 3-5 recurring issues, quantify their frequency142 and impact (e.g., resolution time, customer sentiment). Use the Customer Support Data Fetcher tool."""143 ),144 expected_output=(145 """A summary report detailing the key findings from the customer support data analysis, including:146 - Top 3-5 recurring issues with frequency.147 - Average resolution times for these issues.148 - Key customer pain points mentioned in feedback.149 - Any notable trends in sentiment or support agent observations."""150 ),151 agent=data_analyst # Assign task to the data_analyst agent152 )153154 # Task 2: Identify bottlenecks and suggest improvements155 optimization_task = Task(156 description=(157 """Based on the data analysis report provided by the Data Analyst, identify the primary bottlenecks158 in the support processes contributing to the identified issues (especially the top recurring ones).159 Propose 2-3 concrete, actionable process improvements to address these bottlenecks.160 Consider potential impact and ease of implementation."""161 ),162 expected_output=(163 """A concise list identifying the main process bottlenecks (e.g., lack of documentation for agents,164 complex escalation path, UI issues) linked to the key problems.165 A list of 2-3 specific, actionable recommendations for process improvement166 (e.g., update agent knowledge base, simplify password reset UI, implement proactive monitoring)."""167 ),168 agent=process_optimizer # Assign task to the process_optimizer agent169 # This task implicitly uses the output of analysis_task as context170 )171172 # Task 3: Compile COO report173 report_task = Task(174 description=(175 """Compile the findings from the Data Analyst and the recommendations from the Process Optimization Specialist176 into a single, concise executive report for the COO. The report should clearly state:177 1. The most critical customer support issues identified (with brief data points).178 2. The key process bottlenecks causing these issues.179 3. The recommended process improvements.180 Ensure the report is easy to understand, focuses on actionable insights, and is formatted professionally."""181 ),182 expected_output=(183 """A well-structured executive report (max 1 page) summarizing the critical support issues,184 underlying process bottlenecks, and clear, actionable recommendations for the COO.185 Use clear headings and bullet points."""186 ),187 agent=report_writer # Assign task to the report_writer agent188 )189190### Crew191192The `Crew` brings the agents and tasks together, defining the workflow process193(such as "sequential").194195 from crewai import Crew, Process196197 support_analysis_crew = Crew(198 agents=[data_analyst, process_optimizer, report_writer],199 tasks=[analysis_task, optimization_task, report_task],200 process=Process.sequential, # Tasks will run sequentially in the order defined201 verbose=True202 )203204## Run the crew205206Finally, kick off the crew execution with any necessary inputs.207208 # Start the crew's work209 print("--- Starting Customer Support Analysis Crew ---")210 # The 'inputs' dictionary provides initial context if needed by the first task.211 # In this case, the tool simulates data fetching regardless of the input.212 result = support_analysis_crew.kickoff(inputs={'data_query': 'last quarter support data'})213214 print("--- Crew Execution Finished ---")215 print("--- Final Report for COO ---")216 print(result)217218The script will now execute. The `Data Analyst` will use the tool, the `Process219Optimizer` will analyze the findings, and the `Report Writer` will compile the220final report, which is then printed to the console. The `verbose=True` setting221will show the detailed thought process and actions of each agent.222223To learn more about CrewAI, check out the [CrewAI224introduction](https://docs.crewai.com/introduction).