AI Agents in Business: Five Safe Uses to Get Started

AI Agents Invade French Companies

The integration of AI agents into French companies is progressing cautiously. Currently, these technologies are primarily used in sectors such as supply chain, demand forecasting, and logistics. However, operational complexity hinders their large-scale adoption. Anthony Cirot, Vice President EMEA South at Google Cloud, notes that companies are only deploying 2 to 5 use cases, while the identified potential is nearly 200.

To facilitate the adoption of AI agents, five areas have been identified as easier to implement: onboarding, internal FAQs, repetitive requests, document preparation, and team support.

Onboarding: An Ally for Human Resources

Onboarding, or the integration of new employees, is a process that can benefit from automation through AI agents. This area relies on clear rules and pre-existing documents, reducing the risk of errors that directly impact clients or production. AI agents can relieve human resources of repetitive administrative tasks, thus providing a smooth experience for newcomers.

Implementing these agents requires clearly defining their roles and objectives. For example, an inspector agent can verify the presence of administrative documents, while a secretary agent transforms this information into welcoming messages.

To implement this system, it is essential to configure the agents correctly. This begins with installing the necessary tools via the terminal, such as PowerShell on Windows or Terminal on Mac, using the command:

pip install crewai langchain-[openai](/dossier/openai)

This command installs the essential engines for the project: crewai, which facilitates the creation and coordination of agents, and langchain-openai, which allows CrewAI to communicate with OpenAI's AI.

Next, create a folder named MonProjetRH on the desktop, with a specific structure inside:

• onboarding_jean/ (with two empty files: "RIB.txt" and "contrat_signe.txt")
• verif_onboarding.py (where the code will be written)
• .env (for the API key)

The API key is obtained from "https://platform.openai.com/" under "API keys" > "Create new secret key" and must be added to the ".env" file:

OPENAI_API_KEY=YOUR_API_KEY

The following code is then written in verif_onboarding.py using a text editor like Notepad or VS Code:

from crewai import Agent, Task, Crew
from crewai_tools import DirectoryReadTool
from dotenv import load_dotenv

# 1. The tool to look in the folder
folder_tool = DirectoryReadTool(directory='./onboarding_jean')

# 2. FIRST AGENT: The Inspector (the technical "brain")
inspector = Agent(
    role="HR Verifier",
    goal="Check if RIB.txt, contrat_signe.txt, and mutuelle.txt are present.",
    backstory="You are a precise and rigorous administrative assistant.",
    tools=[folder_tool],
)

# 3. SECOND AGENT: The Secretary (the communication "heart")
secretary = Agent(
    role="HR Secretary",
    goal="Transform the technical report into a polite message for the employee.",
    backstory="You are welcoming and enjoy helping newcomers.",
)

# 4. FIRST TASK: Verification (raw)
verification_task = Task(
    description="Look in the folder and list the missing documents.",
    expected_output="A raw list of absent documents.",
    agent=inspector
)

# 5. SECOND TASK: Formatting (polite)
polite_task = Task(
    description="Take the list of missing documents and draft a very polite message asking the new employee to send them.",
    expected_output="A warm welcome message including the list of documents to provide.",
    agent=secretary,
    context=[verification_task]  # <--- This is where the Secretary "reads" the Inspector's work
)

# 6. LAUNCHING THE TEAM (Putting the 2 agents and the 2 tasks together)
agents=[inspector, secretary],
tasks=[verification_task, polite_task]
result = crew.kickoff()
print("\n--- FINAL MESSAGE FROM YOUR AGENT TEAM ---")

In this system, the inspector agent inventories and drafts a raw report. CrewAI retrieves it and passes it to the secretary agent, who drafts a polite message.

To run this program, simply navigate to the folder via PowerShell and type:

python verif_onboarding.py

The tasks are thus set up. The "HR Verifier" agent uses the list_files_in_directory tool to examine the employee's digital folder. It notes the absence of the mutual insurance document. Then, the "HR Secretary" agent drafts a structured and professional email, ready to be sent. It requests the employee's mutual insurance certificate while welcoming them.

Internal FAQ: An Inspector and a Secretary at Your Service

The AI agents in the FAQ use RAG ("Retrieval Augmented Generation"), which almost entirely eliminates the risk of hallucinations. Simply modifying the FAQ documents allows the agents to update automatically. This system can notably free experts (HR, IT, General Services) from repetitive questions, such as:

• "What is the wifi code?"
• "How do I report a claim?"

To set this up, we first create a "FAQ Inspector" agent. Its sole role is to "read" the documents. These can, for example, be located in the onboarding_jean folder created in the previous example. After that, the "FAQ Secretary" agent takes the raw information and provides an appropriate response to the employee.

The technical setup is "stripped down." A single line in the terminal:

pip install crewai langchain-openai

is used to install the AI engines. A single folder on the desktop (MonProjetRH) is sufficient to store the documents and the code. The FAQ documents are the only "keys to the office" given to the agent. It has access to nothing else, ensuring better security. As seen in the verif_onboarding.py code, CrewAI links these agents. It takes the response found in your documents and passes it to the communication agent.

In practice, to get started, identify frequently asked questions. Then, place the answers in a folder. And launch your agent with the command:

python verif_onboarding.py

You can use a simple Slack, Teams, or a small internal chat window as an interface for the employee to ask a question.

Repetitive Requests: Logistics and IT Support to Save Time

Requests for IT equipment (laptops, monitors, mice) can be frequent. Deploying AI agents can be beneficial to address this without too much risk. The IT inventory typically does not rely on sensitive data. Stock management is binary: either the item is there (1) or it is not (0). An error is generally minimal.

You can first create a logistics agent to check if the equipment is in stock. Then, the "IT Support" agent confirms the order or suggests an alternative. The installation is still simple, via the terminal:

pip install crewai langchain-openai

Create a folder on the desktop with this organization:

• inventory/ (place a stock.txt file listing, for example: "2 Laptops, 5 Mice, 0 Monitors". You can use a Python function update_stock to physically subtract the equipment from stock as soon as an order is validated.)
• manage_equipment.py (The code for your agents)
• .env (the API key)

In the manage_equipment.py file, the Logistics agent will scan the stock.txt file. If an employee requests a monitor, the agent will see "0 Monitors." CrewAI will relay the information to the IT Support agent. For example, it might draft: "Hello, I have received your request for a monitor. Unfortunately, we are out of stock. I will notify you as soon as we receive new ones next week."

Document Preparation: Analyzer and Writer Hand in Hand

Document preparation (contracts, reports, template letters) offers the opportunity to provide the AI with its own document templates. Human final control also exists. As with onboarding, we define two complementary roles. The "Technical Writer" agent collects raw information (name, date, subject) and fills out a document template. The "Proofreader" agent checks the tone, spelling, and ensures that the message is ready to be sent.

To set this up, we use the same procedure as for our HR project. The installation is done in the terminal via:

pip install crewai langchain-openai

A folder MonProjetDocs with a subfolder templates/ contains the standard Word or text files. The "Inspector" agent verifies the source data, and the "Secretary" agent drafts the final document.

The prepa_docs.py file replaces the previous verif_onboarding.py. Here is the code:

from crewai import Agent, Task, Crew
from dotenv import load_dotenv

# 1. FIRST AGENT: The Analyzer (it extracts the data)
analyzer = Agent(
    role="Client Data Analyzer",
    goal="Extract the client's name and the document model to fill out.",
    backstory="You are an assistant who collects the necessary information for drafting.",
)

This code allows the agent to extract the necessary information to fill out a document. Then, the "Writer" agent uses this data to create a final document ready to be sent. This approach ensures that documents are prepared accurately and efficiently, while allowing for human verification before final dispatch.